DEV Community: Sourceable

When Someone Asks AI “What Should I Buy?”, Which Products Get Named?

Sourceable — Wed, 29 Jul 2026 06:26:13 +0000

Shoppers are starting to ask assistants for product recommendations instead of scrolling search results. For e-commerce brands, that turns a shelf of options into a shortlist of one or two. Here's how to be on it.

A shopper used to type "best running shoes for flat feet" into Google, open six tabs, read a couple of reviews, and decide. Increasingly, they just ask: "What running shoes should I get for flat feet under 150 dollars?" And the assistant answers with two or three specific products and a sentence on why each fits.

For an e-commerce brand, that's a profound change in how the shelf works. The old shelf was long: page one of Google, a marketplace with fifty listings, a category page a shopper could scroll. The new shelf is short. The assistant does the scrolling, forms an opinion, and hands over a tiny recommended set. Your product is in it or it isn't, and the shopper may never see the fifty options you were competing against.

Most e-commerce teams are optimizing hard for the long shelf and haven't noticed the short one forming next to it.

Short answer: how does AI decide which products to recommend?

An assistant recommends products by drawing on structured product information, reviews, comparison and "best of" content, and its broader knowledge of a category, then matching that against the specifics a shopper asks for (budget, use case, constraints). Products with clear, consistent, well-described data and strong third-party validation are far more likely to be named. If your product information is thin, inconsistent, or trapped in images, the assistant struggles to confidently recommend it.

Key takeaways

The shelf got short. AI names a few products, not fifty. Being in the recommended set is close to being chosen.
Structured product data is the raw material. Clear attributes, specs, prices, and use cases are what a model matches against a shopper's request.
Reviews and third-party content decide ties. What the web says about your product often matters more than what your product page says.
Specificity wins. Shoppers ask specific questions. Products described specifically, by use case and attribute, get matched to them.

Why this is different from ranking on a marketplace

On a marketplace or a search results page, you compete for position in a long list. A shopper can scroll, filter, and discover you even if you're not first. The list is forgiving; there's room to be found.

An AI product recommendation is not a list. It's a curated answer. The assistant has already done the filtering and formed the shortlist before the shopper sees anything. There's no scrolling to a lower position, because there is no lower position, there's just "named" and "not named." A product that would have been a respectable tenth result on a marketplace is simply absent from an AI answer that recommends three.

This also changes what you're optimizing for. Marketplace ranking rewards things like sales velocity and keyword-stuffed titles. AI recommendation rewards being understandable and well-matched: a model has to grasp what your product is, who it's for, and why it fits a specific request. Those are different games, and the second one is barely being played by most brands.

What an AI reads to recommend a product

When an assistant decides which products to name, it's assembling an answer from a few specific inputs. Optimizing for AI product recommendations means feeding each of them.

Structured product data. Your specs, attributes, price, materials, sizes, compatibility, use cases, stated clearly and, ideally, marked up so a machine can read them without guessing. This is the backbone. A model matching "waterproof, under 150, good for wide feet" against your catalog can only match on information it can actually parse.

Reviews. As with every other category, what customers say is powerful signal. Reviews tell the model not just that your product exists but that it's good, and good for what. Detailed reviews that mention specific use cases ("great for flat feet," "held up in heavy rain") are gold, because they map directly to how shoppers ask.

Comparison and best-of content. The "best running shoes for X" articles, round-ups, and buying guides are exactly what an assistant reads to answer a "best product for X" question. If your product is included and well-positioned in that third-party content, you're in the source material for the recommendation.

Clear, readable product descriptions. Descriptions that state plainly what the product is and who it's for, in real text, not just an image or a wall of marketing adjectives. "Premium. Elevated. Iconic." tells a machine nothing. "Lightweight trail running shoe with a wide toe box and waterproof membrane" tells it everything.

The specificity advantage

Here's the strategic key, and it's a little counterintuitive. Shoppers ask AI specific questions, more specific than they'd ever type into a search box. Not "running shoes" but "running shoes for flat feet, wide, waterproof, under 150." That specificity is an opportunity, because a product that is clearly described for a specific use case gets matched to those specific requests, while a generically described product competes for nothing in particular.

Most product pages are written to sound impressive to humans, which means they're vague. The brands that will win AI product recommendations are the ones that describe their products in the concrete, attribute-rich, use-case-specific language that shoppers actually ask in. If your product genuinely is the best waterproof wide-toe-box trail shoe under 150, say exactly that, clearly, and make sure your reviews and the round-ups say it too. Vagueness is invisibility.

A practical checklist for e-commerce brands

In rough priority order:

Make your product data clean and structured. Accurate, complete attributes and specs, consistent across your site and any marketplaces, ideally with product schema markup so machines read it precisely.

Write descriptions in real, specific, readable text. State what it is, who it's for, and what it's good at, in concrete terms. Don't trap the substance in images or marketing fog.

Invest in reviews, especially specific ones. Encourage genuine reviews that mention real use cases. They double as human persuasion and machine signal.

Get into the buying guides. Earn placement in the "best of" and comparison content your category's shoppers, and their assistants, rely on.

Then check what AI actually recommends. Ask the assistants the specific questions your shoppers ask and see whether your products appear.

You have to see the recommendation to improve it

That final step is the one e-commerce teams skip. You can perfect your product data and gather thousands of reviews, but the only way to know whether an assistant now recommends your product, for which queries, and against which competitors, is to look at the actual answers.

That's what Sourceable does, tracking how AI assistants respond when shoppers ask about products in your category, whether yours get named, how they're described, and which competitors are recommended alongside or instead of you. For an e-commerce brand, that's the difference between hoping your catalog is understood and knowing you're the answer when someone asks what to buy.

The shelf got short. Make sure your product is on it.

FAQ

How is AI product recommendation different from marketplace ranking?
Marketplace ranking places you in a long, scrollable list where lower positions can still be found. AI recommendation produces a short curated set of a few products; you're either named or absent, with no lower position to occupy.

What's the single most important thing for product visibility in AI?
Clear, structured, specific product data. A model can only recommend your product for a shopper's specific request if it can actually parse what your product is, who it's for, and what it's good at.

Do reviews matter for AI product recommendations?
Very much. Reviews signal that a product is good and, crucially, good for what. Detailed reviews mentioning specific use cases map directly onto how shoppers phrase their questions to an assistant.

Should I use product schema markup?
Yes. Product and Offer schema lets machines read your attributes, price, and details precisely rather than inferring them, which makes accurate recommendation more likely.

How do I know if AI recommends my products?
Ask the assistants the specific questions your shoppers ask, like "best [product type] for [use case] under [price]," and see if your products appear. A monitoring tool like Sourceable tracks this across engines systematically.

See which products AI recommends in your category

Check how AI represents your products with Sourceable

The Honest Case Against the AEO Hype (From a Company That Sells AEO)

Sourceable — Tue, 28 Jul 2026 06:00:08 +0000

Half of what you're reading about AI search optimization is overselling a real thing. Here's what's genuinely true, what's exaggerated, and how to invest without getting burned - written by people with an obvious incentive to tell you otherwise.

Let's start with the awkward part. We sell an AI visibility tool. We have every commercial reason to tell you that AEO is the most urgent thing in marketing, that you're already losing, and that you need to act this instant.

A lot of companies in our space are saying exactly that, in exactly those words. And some of it is true. But the breathless version of this story is doing real damage, because it's pushing teams to make expensive, panicked decisions based on stats that don't mean what they're told they mean. So here is the version we'd give a friend who asked whether this is worth their time: the honest one, hype removed.

Short answer: is AEO real or hype?

Both. The underlying shift is real: people increasingly ask AI assistants instead of searching, and being recommended in those answers matters. But much of the surrounding advice exaggerates the current scale, misuses statistics, and sells certainty about mechanics nobody fully understands yet. The right response is to take it seriously and invest modestly and early, not to panic and overspend.

Key takeaways

The trend is real. The current scale is often oversold. AI referral is the fastest-growing channel, but still small next to traditional search.
The scary stats are frequently misread. Growth rates, share figures, and conversion numbers get quoted in misleading ways.
Nobody fully understands the mechanics. Anyone promising a guaranteed formula for AI citations is overselling.
The correct move is measured, not frantic. Position early, cheaply, and keep doing the fundamentals. Don't set money on fire out of FOMO.

What's genuinely true

Let's give the real thing its due, because it is real.

People are changing how they find things. A growing share of buyers, and consumers, now ask an assistant and act on its answer, sometimes without ever clicking a traditional link. AI referral traffic is the fastest-growing acquisition channel on the web right now. That's not marketing spin; it shows up across independent data sets.

Being the named answer matters more than being one of ten links, because AI answers name far fewer options. Visibility is genuinely compounding, so early presence has real advantages. And most brands are doing nothing about it, which means the opportunity for the ones who act is real too. If we didn't believe all of that, we wouldn't have built what we built.

So this is not a "it's all nonsense" post. The foundation is solid. It's the second story on top of it that gets shaky.

What's oversold

Now the uncomfortable part, the claims that need an asterisk.

"AI is taking over search." Not yet, not close. AI platforms combined still drive a small fraction of the traffic traditional search does, often a fraction of a percent of global web traffic in the studies people cite. It's growing fast from a small base. "Fastest growing" and "biggest" are different claims, and the hype constantly blurs them. Abandoning your organic search strategy for AEO today would be a serious mistake.

The giant growth percentages. You'll see "AI referrals grew 300, 400, 800 percent." Those are almost always describing growth from a tiny base, and often describing referral volume while implying market share. A number can quadruple and still be small. When a stat sounds astonishing, check what it's actually a percentage of.

The precise conversion numbers. "AI visitors convert 11x better" and similar figures get repeated as established fact. They're usually single-vendor estimates from small samples, and the qualitative story, AI-referred visitors are more pre-qualified, is more reliable than any specific multiplier. Treat the direction as real and the decimal as decorative.

The guaranteed formulas. Anyone who tells you they know exactly how to get cited by ChatGPT, with a repeatable formula and a guarantee, is overselling. The models are opaque, they change often, and they behave differently from each other. Good AEO improves your odds meaningfully. It does not hand you a lever that reliably produces citations on demand.

Why the hype is actually harmful

You might think exaggeration is harmless if the underlying trend is real. It isn't, for two reasons.

First, it causes bad allocation. A team that believes AI search is already the dominant channel might slash proven organic-search investment to chase it. That's trading a channel driving most of their traffic for one driving a sliver, based on where things might be in a few years. Getting the timing right matters, and hype destroys timing judgment.

Second, it breeds cynicism. When the promised flood of AI traffic doesn't arrive next quarter, the honest conclusion, "this is an early, compounding channel worth a modest, patient investment," gets replaced with "AEO was overblended nonsense," and the team disengages entirely, right before it starts to matter. Overselling today manufactures the backlash that makes people miss the real thing tomorrow.

How to invest in AEO without getting burned

Here's the measured approach, which is genuinely what we'd recommend even though it's less exciting than "panic now."

Keep doing the fundamentals. Do not abandon SEO or organic search. The same crawlable, well-structured, authoritative content that wins traditional search also feeds AI answers. Most good AEO is just good web fundamentals aimed at a new reader. This is the opposite of a rip-and-replace.

Do the cheap, high-certainty things first. Make sure you're not accidentally blocking AI crawlers. Publish an llms.txt. Keep your facts consistent and machine-readable. These are low cost, low risk, and useful regardless of how fast AI search grows. There's no reason to wait and no way to overspend on them.

Position early, but proportionally. Establish presence now because it compounds, but size the investment to the channel's current reality, a promising, growing, still-small channel, not the dominant one. Early and modest beats late and frantic.

Measure instead of believing. This is the antidote to hype in both directions. Don't take anyone's word for how big or urgent this is, including ours. Look at whether AI actually mentions your brand, and whether that presence is growing. Real data about your own situation beats every scary industry statistic.

The genuinely honest pitch

So here's the version with no spin. AI search is a real, growing, compounding channel that most of your competitors are ignoring. It is also smaller today than the loudest voices imply, wrapped in statistics that are frequently misused, and governed by mechanics nobody has fully cracked. The correct response is neither to dismiss it nor to panic about it. It's to do the cheap fundamentals now, position early but proportionally, and, above all, measure your own reality instead of trusting anyone's narrative.

That last point is the one we'll happily make even against our own interest to hype: the smartest thing you can do about AI search is not to believe a blog post about how urgent it is. It's to check your own numbers. Sourceable exists to give you exactly that, an honest read on how AI actually represents your brand across ChatGPT, Claude, Gemini, and Perplexity, so you can decide how much this matters for you based on evidence instead of FOMO.

Take AEO seriously. Just don't take the hype at face value, ours included.

FAQ

Is AEO just hype?
No, but a lot of the marketing around it is. The underlying shift, people asking AI and acting on the answer, is real and growing. The exaggeration is in overstating its current scale, misusing statistics, and promising guaranteed results.

Should I move my budget from SEO to AEO?
Generally not by much, and not yet. Traditional search still drives far more traffic for most sites, and the same fundamentals serve both. Add AEO as a modest, early investment rather than a replacement.

Why are AEO growth stats so high?
Usually because they measure growth from a very small base, and sometimes because they conflate referral volume with market share. A number can grow several hundred percent and still represent a tiny slice of total traffic. Always check what the percentage is of.

Can anyone guarantee I'll get cited by AI?
No. The models are opaque and change frequently. Good AEO meaningfully improves your odds, but anyone promising a guaranteed formula for citations is overselling.

What's the safest way to start?
Do the cheap, certain things: make sure AI crawlers aren't blocked, keep your content readable and consistent, publish an llms.txt, and measure whether AI actually mentions you. All low cost, all useful regardless of how fast the channel grows.

Skip the hype. Check your own numbers.

See how AI actually represents your brand with Sourceable

"Hey ChatGPT, Find Me a Good Plumber" - What AI Search Means for Local Businesses

Sourceable — Mon, 27 Jul 2026 06:05:02 +0000

AI search isn't just a big-brand problem. When someone asks an assistant for a good dentist, contractor, or cafe nearby, one business gets named and the rest don't exist. Here's how local businesses win that answer.

For a long time, the advice for local businesses was simple and stable: claim your Google Business Profile, gather reviews, rank in the map pack, and you'll get found. It worked because that's how people searched. They typed "dentist near me," scanned the three-pack, and picked one.

That behavior is quietly changing. More people now just ask: "Who's a good dentist in this area?" or "I need an emergency plumber tonight, who should I call?" And the assistant does what it does everywhere else. It doesn't return ten options. It names one or two, with a sentence about why. The rest of the businesses in town might as well not exist for that query.

Local businesses tend to assume AI search is a problem for software companies and national brands. It isn't. It might hit them harder, because a local recommendation is often a decision, not a starting point.

Short answer: does AI search matter for local businesses?

Yes, and increasingly so. When someone asks an AI assistant for a local service or place, the assistant recommends a short list of specific businesses, drawn from reviews, directory listings, local mentions, and its knowledge of the area. If your business isn't part of what it reads, it won't be named, and the customer will simply go with whoever was. For high-intent local queries, being the named answer is close to being the chosen business.

Key takeaways

Local recommendations are high-intent. Someone asking for a plumber "tonight" is ready to act. Being named is nearly being hired.
Reviews and consistency are the raw material. AI leans heavily on what third-party sources say about local businesses, especially reviews.
Consistency across listings matters enormously. Conflicting hours, addresses, or names make an assistant hesitate to recommend you.
The basics still apply, just aimed at a new reader. The work that wins local AI recommendations overlaps with good local SEO, done for a machine that summarizes instead of lists.

Why this hits local businesses differently

For a national SaaS brand, an AI recommendation is one touchpoint in a long buying journey. For a local business, it's often the whole thing. Someone with a burst pipe is not going to research five plumbers, read comparison articles, and book a demo. They ask, they get a name, they call. The distance between "recommended by the AI" and "money in the till" is a single phone call.

That compresses the stakes. In the old map-pack world, being the fourth-ranked plumber still meant a motivated customer could scroll and find you. In an AI answer that names one or two businesses, fourth place is invisible. The winner-take-most dynamic that's reshaping every category is even sharper locally, because local intent is so immediate and the shortlist so short.

The flip side is that the opportunity is just as concentrated. Most local competitors have no idea this is happening. The business that gets this right first, in a given town and category, can own the answer for a long time.

What an AI actually reads about a local business

When an assistant recommends a local business, it's drawing on a specific pool of information, and knowing that pool tells you exactly what to work on.

Reviews, heavily. What customers say about you on the major platforms is some of the strongest signal an assistant has. A steady stream of genuine, recent, detailed reviews doesn't just build trust with humans; it's raw material the model reads to decide who's actually good. Quantity, recency, and specificity all matter.

Directory and listing data. Your presence and accuracy across business directories, maps, and local listings. This is where consistency becomes critical, and it's where most local businesses quietly fail.

Local mentions and coverage. Being written about, recommended in community forums, listed in local "best of" round-ups, or discussed in neighborhood groups. The same third-party consensus that drives AI recommendations for brands drives it for the corner cafe too.

Your own site, if it has one and it's readable. Clear, plain information about what you do, where, when, and for whom. Many local businesses either have no real site or bury the essentials in an image or a PDF menu the machine can't read.

The consistency problem that quietly sinks local businesses

If there's one thing that disproportionately hurts local businesses in AI search, it's inconsistent information.

Your restaurant is listed with one set of hours on Google, different hours on an old directory, and a third version on a review site. Your business name appears three slightly different ways. Your address changed but two listings never updated. To a human, these are minor annoyances. To an AI trying to decide whether to confidently recommend you, they're a reason to hesitate, or to recommend the competitor whose information is clean and agrees with itself everywhere.

Remember that models trust consensus. When the web can't agree on your own basic facts, you've given the assistant a reason to doubt you. Cleaning up and aligning your listings, hours, name, address, phone, and category, across every place they appear is unglamorous, cheap, and one of the highest-return things a local business can do for AI visibility.

A practical starting list for local businesses

You don't need a big budget. In rough order:

Get your basic facts identical everywhere. Name, address, phone, hours, and category, aligned across Google, maps, directories, and your site. Fix the contradictions first.

Make reviews a habit, not an afterthought. Ask happy customers, consistently. Recent, specific, genuine reviews are among the strongest signals an assistant reads. Never fake them; it backfires and it's obvious.

Have a simple, readable web presence. Even one clean page that states plainly what you do, where, when, and who you serve, in real text, not trapped in an image. That gives the machine something clear to extract.

Earn local mentions. Get into local round-ups, community recommendations, and neighborhood discussions where you can, honestly. This is the local version of building consensus.

Then check whether it's working. Ask the assistants the questions your customers ask, "best [your service] in [your town]", and see whether you're named. If you're not, you now know where to push.

You still have to check the answer

That last step is the one local businesses skip, and it's the one that closes the loop. You can align every listing and gather a hundred reviews, but the only way to know whether an assistant now recommends you is to look at what it actually says.

That's what Sourceable does, tracking how AI assistants respond when people ask about businesses like yours, whether you're named, how you're described, and who's recommended instead. For a local business, that's the difference between hoping the work paid off and knowing you're the answer when someone in your area asks.

"Find me a good one nearby" is becoming one of the most common things people ask an AI. Make sure the answer is you.

FAQ

Do small local businesses really need to worry about AI search?
Increasingly, yes. As more people ask assistants for local recommendations instead of scrolling a map, being named in that answer becomes as important as ranking once was, and often more decisive, because local intent is so immediate.

What's the single most important thing to fix first?
Consistency of your core facts, name, address, phone, hours, category, across every listing and your site. Contradictory information gives an assistant a reason to hesitate or recommend a competitor.

How much do reviews matter for AI recommendations?
A lot. Reviews are among the strongest third-party signals an assistant reads about a local business. Recent, specific, genuine reviews carry real weight. Faking them backfires.

I don't have a website. Am I invisible to AI?
Not necessarily, since assistants also read listings and reviews, but a simple, readable page describing what you do and where helps considerably. Even one clean page in real text beats nothing or an image-only presence.

How do I know if an AI recommends my business?
Ask the assistants the questions your customers would, like "best [your service] in [your town]," and see if you're named. A monitoring tool like Sourceable tracks this systematically so you're not spot-checking by hand.

See if AI recommends your business

Check how AI represents your business with Sourceable

AI Is Telling Your Customers Something False About Your Brand. Here's How to Fix It.

Sourceable — Sun, 26 Jul 2026 05:00:00 +0000

Outdated pricing, a discontinued product, a founder who left years ago - AI states it all with total confidence. You can't edit the model, but you can change what it says. Here's the playbook.

Somewhere today, a potential customer asked an AI assistant about your brand and got an answer that was confidently, fluently wrong.

Maybe it quoted pricing you changed last year. Maybe it described a product you discontinued, mentioned a founder who left in 2023, or repeated a "it's hard to set up" criticism absorbed from a stale review. The customer had no reason to doubt it. AI states its mistakes in exactly the same calm, authoritative tone it uses for facts. And unlike a bad review you can at least see and respond to, this happened in a private conversation you will never know about.

The reflex is to feel powerless. You can't call OpenAI and ask them to fix it. But that reflex is wrong. You have more control than you think, once you understand why the model got it wrong in the first place.

Short answer: can you fix what AI says about your brand?

Yes, indirectly. You can't edit the model, but you can change the inputs it draws from. AI describes your brand using a mix of what it learned in training and what it retrieves live from the web. By correcting and reinforcing your facts across the sources it reads, you can shift what it says, often faster than you'd expect for the engines that search live.

Key takeaways

AI errors come from stale or conflicting inputs, not malice. Fix the inputs, fix the output.
Live-search engines update fastest. For Perplexity, ChatGPT search, and Gemini, fresh, crawlable, consistent facts can change the answer quickly.
Consensus beats correction. One updated page rarely overrides a web full of old information. You have to update the sources, plural.
You can't fix what you can't see. Most brands don't even know what AI is getting wrong, because the errors happen in conversations with no trace.

First, understand why it's wrong

An AI doesn't invent falsehoods about you out of nowhere. Its errors trace to one of two places.

The first is stale training knowledge. The model learned a snapshot of the web up to a cutoff date. If your pricing, leadership, or product lineup changed after that, the model may still "remember" the old version and state it confidently, because from its perspective that's what it knows.

The second is bad or conflicting retrieval. For engines that search live, the answer reflects whatever they fetch in the moment. If an outdated third-party page, a years-old review, or an inconsistency between your own pages is what they find, that's what they repeat. Sometimes the wrong information is more prominent on the web than the correct version, so the model trusts it.

Both are fixable, but they're fixed differently, which is why the first step is always diagnosis: find out what's wrong and, ideally, which source it's echoing.

The remediation playbook

Once you know what's wrong, here's the order of operations that actually moves the answer.

1. Correct your own canonical pages first. Make sure the accurate fact is stated clearly, plainly, and prominently on your own site, in extractable text, not buried in an image or three paragraphs deep. This is the source of truth everything else gets checked against. If your own site is ambiguous or outdated, nothing downstream will hold.

2. Fix the loudest third-party sources. Update your profiles on directories, review platforms, and any listing you control. Where you can, reach out to correct outdated third-party articles or data aggregators repeating the old fact. Remember the model triangulates: it believes what several sources agree on, so you have to move more than one.

3. Build fresh, consistent reinforcement. Publish current, dated content that states the correct facts, and keep your positioning and key details identical everywhere they appear. Consistency is what turns a correction into a new consensus. The goal is for the accurate version to become the thing the web overwhelmingly agrees on.

4. Target the live-search engines for speed. For Perplexity, ChatGPT search, and Gemini, which retrieve live pages, updated crawlable content can change the answer relatively quickly, because they're reading the current web, not just memory. Training-based recall shifts more slowly, over model updates, which is why consistent long-term presence matters for the correction to stick.

5. Verify the fix landed. A correction is a hypothesis until you check. Re-ask the questions your buyers ask and confirm the answer actually changed. If it didn't, the old source is probably still winning, and you have more reinforcing to do.

Why "just update your website" isn't enough

The most common mistake is fixing one page and assuming the job is done. It rarely is. If your site now says the right thing but a dozen third-party pages still say the wrong thing, the model is looking at a web that disagrees with itself, and it may well side with the majority or hedge.

Correcting AI's view of your brand is less like editing a document and more like changing a reputation. You're not overwriting a single record; you're shifting a consensus. That means the fix is proportional to how widespread the wrong information is. A small, recent error can move fast. A misconception baked into years of content takes sustained, consistent reinforcement to overturn.

The part most brands miss entirely

Here's the uncomfortable truth underneath all of this: most brands have no idea what AI is getting wrong about them, so they never start the process at all. The errors live in private conversations. There's no notification, no referral, no review to flag it. You find out, if you ever do, when a prospect mentions it on a sales call, by which point it's already cost you others who never spoke up.

You cannot fix what you cannot see. The first and most important move is simply to look: ask the engines what they say about you, systematically, and catch the errors before your customers do. That monitoring is exactly what Sourceable provides, tracking how ChatGPT, Claude, Gemini, and Perplexity describe your brand, including accuracy and sentiment, so a false claim becomes something you can catch and correct instead of a silent leak.

FAQ

Can I contact an AI company to fix an error about my brand?
Generally no, not for individual factual corrections. The practical path is to change the inputs the model draws from: your own canonical pages and the third-party sources it reads.

How long does it take to correct what AI says?
It varies. Live-search engines that retrieve current pages can update relatively quickly once your facts are fixed and consistent. Errors rooted in training data shift more slowly, across model updates, which is why sustained consistency matters.

Why does AI keep repeating an old fact after I updated my site?
Because it triangulates across many sources. If third-party pages still carry the old information, or your own pages are inconsistent, the model may trust the outdated majority. You usually have to correct more than one source.

What kinds of errors are most common?
Outdated pricing, discontinued products, former leadership, and stale criticisms absorbed from old reviews. Anything that changed after a model's training cutoff, or that a prominent old source still states, is a candidate.

How do I even know what AI is getting wrong?
You have to ask the engines directly and systematically, since the errors happen in private conversations with no trace. A monitoring tool like Sourceable surfaces how each engine describes your brand so you can spot inaccuracies.

Find out what AI is saying about you

Check how AI describes your brand with Sourceable

The 62% Problem: When AI Cites Your Page But Never Says Your Name

Sourceable — Sat, 25 Jul 2026 05:00:00 +0000

You can be the source behind an AI answer and get none of the credit. It's called a ghost citation, and it might be the most overlooked gap in your AI visibility.

There's a failure mode in AI search that almost nobody is measuring, because on the surface it looks like a win.

An assistant answers a buyer's question. It pulls a fact from your page. Your URL even shows up in the little list of sources. And yet the buyer walks away without ever seeing your brand name, because it was never written into the answer. You did the work. Someone else, or nobody, got the memory.

That's a ghost citation, and by one reported 2026 analysis, roughly 62% of AI citations are exactly this: your page used as a source, your name absent from the text.

Short answer: what is a ghost citation?

A ghost citation is when an AI answer uses your content as a source, sometimes even linking it, but never actually writes your brand name in the response the user reads. The buyer absorbs your information without learning it came from you. You get the extraction cost and none of the brand benefit.

It matters because the real payoff of AI visibility is the halo effect: people who see your name in an answer remember it and search for you later. A ghost citation skips that entirely. There's no name to remember, so there's no downstream demand.

Key takeaways

A source link is not a mention. Being cited as a source and being named in the answer are different outcomes with very different value.
Named mentions build brand memory. Ghost citations don't. The halo effect that makes AI visibility worth chasing depends on the reader seeing your name.
Standard tools can't tell them apart. A referral log shows a click either way. Only reading the actual answer text reveals whether you were named.
Some ghost citations are convertible. Many are near-misses where a small change to how you're described tips a bare source into a spoken recommendation.

Why a name is worth more than a link

Picture the two outcomes side by side. In the first, an assistant says "According to Sourceable, most sites block at least one AI crawler by accident," and links you. In the second, it says "Most sites block at least one AI crawler by accident," and quietly lists your URL among five sources at the bottom. Same fact, same source. Radically different result.

In the first case, the reader learns something and learns who taught them. Your name rides along with the credibility of the answer. A week later, when they're ready to act, they might search for you. In the second case, the reader learns the fact and moves on, with no idea you existed. The information became a commodity and your brand became invisible.

This is why counting raw citations misleads you. Two brands can have identical citation counts, and one is building durable brand demand while the other is donating facts to the void.

Why this is invisible in your analytics

Here's the trap. If a ghost citation includes a link and someone clicks it, you get a referral session that looks exactly like a named-mention click. Your analytics cannot tell you whether the answer said your name or not, because analytics only sees the click, never the answer that produced it.

So a brand can be getting cited constantly, feel good about the referral trickle, and completely miss that most of those citations are ghosts, quietly failing to build any name recognition. The only way to know is to look at the actual text of the answers, which is not something a dashboard of sessions and bounce rates will ever show you.

Ghost citations are often the easiest wins you have

Now the good news, and it's genuinely good. A ghost citation is not a rejection. The engine already trusts your page enough to source it. It just didn't say your name. That's a much smaller gap to close than earning a citation from scratch.

Usually the fix lives in how your content is written. If your key claims are phrased so your brand name is welded to them, the model tends to carry the name along when it lifts the fact. "Sourceable's scan found that..." travels better than a naked statistic in the middle of a paragraph. Self-contained, quotable sentences that attribute the claim to you are far more likely to be reproduced with attribution intact.

So the playbook is: find the queries where you're a ghost, and rewrite the source content so your name is inseparable from the insight. You're not chasing a new citation. You're upgrading one you already have from silent to spoken. That's the fastest kind of win in AEO, because the hardest part, earning the engine's trust, is already done.

What to do about it

Three moves.

Separate named mentions from ghost citations in how you measure. Stop treating "we got cited" as one metric. The question is what share of your citations actually name you, because that share is what predicts brand lift.

Weld your brand to your best claims. Write your most citeable facts, your original data especially, so the attribution is baked into the sentence. Make it awkward for a model to state your insight without stating your name.

Prioritize the near-misses. The queries where you're a ghost, or where a competitor is named and you're a silent source, are your cheapest available wins. The trust is already there; only the naming is missing.

You have to read the answers to see it

None of this is visible without looking at the raw AI responses, which is exactly what Sourceable does. It distinguishes named mentions from ghost citations across ChatGPT, Claude, Gemini, and Perplexity, so you can see not just whether you're cited but whether you're credited, find the queries where you're a ghost, and fix the ones worth fixing.

Getting cited feels like winning. Getting named is the actual win. The difference is worth measuring.

FAQ

What's the difference between a citation and a mention?
A citation uses your page as a source, sometimes with a link. A mention writes your brand name into the answer the user reads. A ghost citation is a citation without a mention: sourced but unnamed.

Why do ghost citations matter if I still get the click?
Because most of AI visibility's value is the halo effect, where seeing your name drives later branded search. A ghost citation gives the reader your information but no name to remember, so it skips the brand benefit even when it sends a click.

How common are ghost citations?
Very. One reported 2026 analysis put them at roughly 62% of all AI citations, meaning most of the time your page is sourced, your name is not spoken.

How do I turn a ghost citation into a named one?
Rewrite the source content so your brand name is attached to the claim, especially for original data. Self-contained, attributed sentences are more likely to be reproduced with your name intact.

Can I see my ghost citations in Google Analytics?
No. Analytics sees clicks, not the answer text that produced them. Detecting ghost citations requires reading the actual AI responses, which a monitoring tool like Sourceable does.

See whether AI credits you or just uses you

Check your named mentions vs ghost citations with Sourceable

Structured Data Is How You Explain Your Brand to a Machine That Can’t Read Between the Lines

Sourceable — Thu, 23 Jul 2026 06:36:44 +0000

Humans infer. Machines extract. Schema markup is the difference between an AI guessing what your page means and knowing it - and in the answer-engine era, that gap decides whether you get cited.

A human reading your pricing page understands, without being told, that "$49/mo" is a price, that "Acme Pro" is a product, and that the person smiling in the corner is your founder. We fill in meaning automatically. It's what reading is.

A machine does none of that for free. It sees text and shapes and has to guess what any of it means. Most of the time it guesses reasonably well, and sometimes it guesses wrong, or gives up. Structured data is how you stop it guessing. It's a way of labeling the meaning of your content explicitly, in a language machines read perfectly, and in the era of AI answers it has quietly become one of the highest-leverage things you can add to a page.

Short answer: does schema markup help AI cite my brand?

Yes. Structured data, usually written as JSON-LD, labels what your content means: this is an organization, this is a product, this is its price, this is an FAQ, this is the author. That labeling makes your facts unambiguous and easy for AI to extract with confidence, which makes correct citation and accurate description more likely. It doesn't guarantee a citation, but it removes the ambiguity that causes a model to skip you or get you wrong.

Key takeaways

Schema translates prose into facts. It turns "content a machine must interpret" into "data a machine can lift directly."
JSON-LD is the standard format. A small script in your page's head, invisible to readers, fully legible to machines.
It reduces errors and hedging. Explicit facts are safer for a model to state than inferred ones, so you get quoted more accurately.
It's foundational, not magic. Schema helps a machine understand a page it can already reach and read. It doesn't fix a blocked crawler or thin content.

What structured data actually is

Under the hood, structured data is a small block of code, almost always JSON-LD, that sits in your page and describes the page's meaning using a shared vocabulary (schema.org) that machines already understand.

Instead of hoping a model infers that your company is an organization founded in a certain year with a certain product, you state it outright: this is an Organization, here is its name, here is its founding date, here is the Product it offers, here is the price, here is the FAQ and each question's answer. The reader never sees this block. The machine reads it first. You've handed over the facts in a form that requires zero interpretation.

That's the whole idea: replace inference with declaration. And a fact a machine doesn't have to interpret is a fact it can state with confidence.

Why this matters more for AI than it did for Google

Structured data isn't new; SEO teams have used it for years to earn rich results in Google. But its role changes in the answer-engine era, and in some ways it matters more.

Classic search used schema mostly to decorate a link: star ratings, FAQ dropdowns, recipe cards. The link was still the product. In AI search, there is no link to decorate. The model reads your page, extracts facts, and generates an answer. Whether it extracts your facts, correctly, depends heavily on how legible those facts are, and schema is legibility in its purest form.

There's a caveat worth stating plainly, because honesty is the whole brand here: Google reduced FAQ rich results in search back in 2023, so don't expect schema to produce the search snippets it once did. But the machine-readability benefit, the thing that helps an AI parse and trust your facts, is exactly the benefit that now matters most. You're not marking up your page for a rich result anymore. You're marking it up for a machine that's about to describe you to a customer.

The schema types worth prioritizing

You don't need to mark up everything. A few types carry most of the value for AI visibility.

Organization. The foundational one. Your name, logo, description, founding details, and social profiles, stated as fact. This is how a machine knows who you are without reconstructing it from scattered pages.

Product and Offer. For anything you sell: what it is, what it costs, what it includes. This is precisely the information a buyer asks an assistant about, and precisely what you don't want it guessing.

FAQPage. Question-and-answer pairs, explicitly labeled. This is a strong one, because it maps directly onto how people query AI and gives the model pre-formatted, self-contained answers to lift.

Article and Author. For your content, who wrote it and when. Authorship and freshness are trust signals, and stating them explicitly helps.

BreadcrumbList and WebSite. Structural context that helps a machine understand how your site is organized and what's canonical.

Start with Organization and whatever describes your core offering. Those two alone remove most of the ambiguity a model faces when trying to describe your business.

Where schema fits in the bigger picture

A blunt reality check so you spend effort in the right order. Structured data helps a machine understand a page it can already reach and read. It does nothing if the crawler is blocked in robots.txt, and nothing if your facts are trapped in an image instead of text. Schema is the third step, not the first.

So the sequence is: make sure AI crawlers are allowed, make sure your content is real extractable text, and then add schema to remove ambiguity about what that text means. Add an llms.txt to point at your canonical sources, and you've built a genuinely machine-legible site from the ground up. Schema is a powerful layer, but it's a layer on a foundation, not a substitute for one.

FAQ

What is structured data / schema markup?
It's code, usually JSON-LD, that labels the meaning of your content using the shared schema.org vocabulary, so machines can extract facts like your organization, products, prices, and FAQs without having to interpret prose.

Does schema markup guarantee AI will cite me?
No. It removes ambiguity and makes your facts easy to extract accurately, which makes correct citation and description more likely. It's a strong enabling factor, not a guarantee, and it can't compensate for a blocked crawler or thin content.

Isn't schema just an SEO thing from years ago?
It began as an SEO tactic for rich results, and Google has since reduced some of those (FAQ rich results in 2023, for example). But the machine-readability it provides is exactly what AI answer engines need to parse and trust your facts, so its importance has grown, not shrunk.

Which schema types should I add first?
Organization and whatever describes your core offering (Product/Offer), then FAQPage, then Article/Author. Those cover most of what a buyer asks an assistant about your brand.

Do I need schema if my content is already clear?
Clear writing helps, but schema removes the remaining ambiguity by stating facts explicitly rather than leaving them to inference. For the facts that matter most, price, product, identity, explicit is safer than implicit.

Make sure machines can actually read you

Schema is one piece of being legible to AI. Sourceable's agent-readiness scan checks whether machines can find, read, and understand your site, including your structured data, and returns the specific fixes to close the gaps.

Scan your site's agent-readiness, free

The Uncomfortable Truth About AI Visibility: It's Decided on Websites You Don't Own

Sourceable — Wed, 22 Jul 2026 05:53:56 +0000

You can perfect every page on your site and still be invisible in AI answers. Because when an assistant decides who to recommend, it's mostly reading Reddit, reviews, and comparison articles, not your homepage.

Here is the sentence that reorders a marketing strategy: the single biggest factor in whether an AI recommends your brand is almost entirely outside your website.

For twenty years, "controlling your narrative online" meant controlling your own pages. You wrote the copy, you owned the domain, you optimized the funnel. Your website was the thing you could perfect, so you perfected it. That instinct is now working against you, because AI assistants don't build their opinion of you from your site. They build it from everyone else's.

Short answer: where does AI actually learn about your brand?

Mostly from third-party sources you don't control: community threads like Reddit, review platforms, comparison and "best of" articles, editorial coverage, and reference sites like Wikipedia. Your own website matters, but as one voice among many. When independent sources agree about you, the model states it with confidence. When they're silent or contradictory, the model hedges or leaves you out, no matter how good your homepage is.

Key takeaways

AI triangulates. It trusts what many independent sources agree on, not what you say about yourself.
The sources differ by engine. Reported analyses show ChatGPT leaning on encyclopedic references, Perplexity heavily on community signal like Reddit, and Google's AI answers on multimodal content. Optimizing one site can't cover all of that.
Being known is not being recommended. A model can recognize your brand and still never suggest it, if you're rarely named alongside the category leaders.
The work moves off your domain. The highest-leverage AEO is now earning mentions in the places AI reads, not polishing pages you already control.

Why your homepage isn't the source of truth anymore

Think about how an assistant answers "what's the best tool for X." It does not open your website, read your value proposition, and get persuaded the way a human visitor would. It draws on a vast store of what the web collectively says about your category, plus whatever it retrieves live, and synthesizes a recommendation from the consensus.

Your homepage is a single, obviously self-interested source in that pool. Of course you say you're the best; every competitor's homepage says the same. So the model weights independent corroboration far more heavily than self-description. What a hundred Reddit users, a dozen review pages, and three comparison articles say about you moves the needle. What your hero headline says barely registers.

This is why brands with beautiful sites are routinely invisible in AI answers, and why brands with modest sites but heavy third-party presence get recommended constantly. The decision was made off your property.

The sources aren't the same across engines

It gets harder, because the engines don't even read the same corner of the web. Reported source-diversity analyses in 2026 paint a fragmented picture:

ChatGPT leans heavily on encyclopedic and reference sources, with Wikipedia reportedly making up a large share of its top citations.
Perplexity leans hard on community signal, reportedly citing Reddit at a strikingly high rate.
Google's AI answers favor multimodal content, with YouTube reportedly a significant citation source.

Overlap between engines is thin. Studies this year put shared citation domains between ChatGPT and Perplexity in the low double digits. The practical consequence is brutal: there is no single off-site channel that wins all of them. A Reddit-heavy strategy might lift Perplexity while doing little for an engine that leans on reference sites. Winning broadly means showing up credibly across several different kinds of sources, not one.

Being recognized is not the same as being recommended

Here's the finding that should unsettle anyone who thinks a strong Knowledge Graph entry is enough. A reported 2026 analysis compared two well-known athleisure brands with nearly identical structured descriptions. One appeared in a large majority of AI recommendations for the category. The other, with near-identical on-paper credentials, appeared in essentially none.

The difference was not on-site optimization. It was co-mention: how often each brand gets named alongside the category leaders in the third-party content models learn from. The recommended brand lived in the comparisons, the round-ups, the "X vs Y" articles, the community threads where the category gets discussed. The invisible one was known but not associated. The model recognized it and still didn't suggest it, because nothing in its training taught it that this brand belonged in the answer.

AI visibility, it turns out, is built on semantic association across the open web. You want the model to have read your name in the same breath as the leaders in your space, over and over, from sources it trusts.

What this means you should actually do

If the decision happens off your site, your strategy has to move off your site too. Four shifts, in rough priority order.

Earn co-mentions in category content. Get named in the comparison articles, best-of round-ups, and expert lists your category already trusts. This is the single highest-leverage move, because it directly feeds the co-mention signal that decides recommendations. It is also the work on-page SEO structurally cannot do for you.

Take community presence seriously. Reddit, niche forums, Slack and Discord communities, and Q&A sites are not beneath a serious brand anymore; they are a primary source for some of the biggest engines. That does not mean astroturfing, which backfires. It means being genuinely present and useful where your category is discussed, so authentic mentions accumulate.

Cultivate reviews and third-party validation. Review platforms and independent editorial coverage are exactly the corroboration a model looks for. A steady stream of honest reviews does citation work your landing pages never will.

Keep your own facts consistent, then stop over-investing there. Your site still matters as the canonical reference the others get checked against, so keep your category, positioning, and facts clean and machine-readable. But once that's solid, additional homepage polish has sharply diminishing returns compared to earning one more credible third-party mention.

The uncomfortable reframe: your marketing team's job is shifting from writing your story to getting the web to tell it consistently.

You can't influence what you can't see

There's a catch that makes all of this harder than classic PR. When the web's consensus about you shifts, and when an engine starts or stops recommending you, none of it shows up in your analytics. You can earn a great comparison article or a viral Reddit thread and have no idea whether it moved your standing in ChatGPT or Perplexity.

That's the visibility Sourceable is built to give you: how often each engine actually names your brand, which competitors get named alongside you, and how that presence moves as your off-site footprint grows. It turns the invisible, third-party consensus into something you can watch and act on, so earning a mention becomes a measurable cause rather than a hopeful gesture.

FAQ

Does my own website still matter for AI visibility?
Yes, but as one source among many. It's the canonical reference other sources get checked against, so keep it accurate, consistent, and machine-readable. It just isn't where the recommendation decision is primarily made.

Why does AI weight third-party sources over my site?
Because your own site is an obviously self-interested source. Models weight independent corroboration more heavily, so what many outside sources agree about you carries more decision-making weight than your own claims.

Is Reddit really important for AI visibility?
For some engines, yes. Reported analyses show Perplexity in particular citing community sources like Reddit at a very high rate. The right approach is authentic, useful presence, not manipulation, which tends to backfire.

What is co-mention and why does it matter?
Co-mention is how often your brand is named alongside the category leaders in third-party content. A reported 2026 study found brands with near-identical profiles differed enormously in AI recommendations based largely on this, because the model learns which brands "belong" in an answer from these associations.

If I can't control these sources, what can I do?
You can influence them: earn mentions in category comparisons and round-ups, be genuinely present in relevant communities, cultivate honest reviews, and keep your own facts consistent so everything checks out. Then measure which engines respond.

See who AI names in your category

You can do all the off-site work in the world and still be guessing about whether it landed. Sourceable shows you how often ChatGPT, Claude, Gemini, and Perplexity actually name your brand, which competitors appear beside you, and how it shifts over time, so the consensus you're trying to shape stops being invisible.

Your AI visibility is being decided on websites you don't own. The least you can do is watch the result.

See how AI represents your brand with Sourceable

The Agent Web Is Being Built Inside a Folder Called /.well-known/

Sourceable — Mon, 20 Jul 2026 06:45:38 +0000

While marketers argue about llms.txt, a much larger machine-readable layer is quietly taking shape at your domain root. Here's an honest map of what exists, what's real, and what's still a draft.

Open a terminal and try this against almost any website:

curl https://example.com/.well-known/agent.json

You will almost certainly get a 404. Try it against a few hundred sites and you will get a few hundred 404s. That emptiness is the most interesting thing happening on the web right now.

There is a directory convention, /.well-known/, that has quietly existed for years holding boring infrastructure files. Over the last eighteen months it has become the staging ground for something bigger: the place where websites declare, in machine-readable form, what an AI agent is allowed to do with them. Not read. Do.

Most of it is empty. Some of it is not even finalized. But the shape of it is now clear enough to map, and knowing the map is worth more than rushing to fill it in.

Short answer: what is the agent web layer?

It's a set of small, machine-readable files published at predictable paths, mostly under /.well-known/, that tell AI agents what your site is, what services it exposes, how to authenticate, and how you want your content used. Where robots.txt says who may crawl and llms.txt says what your content means, this layer says what an agent can do with you.

It matters because AI is shifting from reading the web to acting on it. A model that summarizes your page needs your text. An agent that books, buys, compares, or calls your API needs a contract. That contract lives here.

Key takeaways

Three distinct jobs. Permission (robots.txt), meaning (llms.txt), and capability (the /.well-known/ layer). They are complementary, not competing.
Maturity varies wildly. Some of these are settled conventions. Others are IETF drafts. At least one has no ratified path at all and is being probed by convention.
Adoption is near zero. Which is exactly why the map is more useful than the checklist right now.
Do the boring parts first. Sitemaps, JSON-LD, crawlable text, and sane robots rules still carry almost all of your actual AI visibility today.

Three layers, three different questions

It helps to separate what these files actually do, because they get lumped together constantly.

Permission: robots.txt. The oldest layer. It answers "which crawlers may access which paths." It is a gate, not a description. It has also grown a new dialect: Content-Signal: directives, which let you state how your content may be used (training versus indexing versus live answers) rather than simply whether it may be fetched. That distinction matters, because "don't train on me" and "don't cite me" are very different wishes.

Meaning: llms.txt. A Markdown file at your root describing what your site is, which pages are canonical, and what to ignore. It gives a model context so it stops guessing which of your pages speaks for you.

Capability: /.well-known/. The newest and least populated layer. Instead of describing your content, these files describe your services: here are my APIs, here is how to authenticate against them, here is what my agent can do, here is how to pay me.

The first two serve a reader. The third serves an actor. Almost every site on the web has addressed the first, some have addressed the second, and virtually none have addressed the third.

What actually lives in /.well-known/

Here's the honest inventory, with maturity stated plainly, because a lot of writing on this subject presents drafts as settled standards.

A2A Agent Card at /.well-known/agent.json. From the Agent-to-Agent protocol. It declares that your organization runs an agent, gives it a name, and describes its capabilities and skills so other agents can discover and talk to it. The probe path is stable enough to rely on, though the spec is still settling.

MCP Server Card at /.well-known/mcp/server-card.json, defined in SEP-1649. Model Context Protocol is how assistants connect to external tools and data. A server card announces that you expose an MCP server and describes it. Worth knowing: the schema is an early draft and real-world implementations already disagree with it. Some cards put name and version at the top level; older drafts nest them under serverInfo. Anyone telling you this is locked down has not looked at production cards.

OAuth discovery and protected-resource declarations. These are the grown-ups in the room. If your APIs require authentication, agents need to discover how to authenticate, and which authorization servers can mint valid tokens for your protected endpoints. This is well-trodden OAuth/OIDC machinery, not speculative. If you run an authenticated API, this is the least experimental item on the list.

API catalog and agent skills index. Machine-readable listings of what your APIs are and what your agent can actually do. Without these, an agent has to guess at your capabilities or rely on a human having documented them somewhere in prose.

Web Bot Auth at /.well-known/http-message-signatures-directory. An IETF draft addressing a real problem: today, a bot claiming to be GPTBot is just a string in a header, and anyone can send it. This publishes cryptographic keys so bots can prove who they are. Genuinely important direction, genuinely still a draft.

x402 at /.well-known/x402. The payments layer, built on the long-dormant HTTP 402 "Payment Required" status code, so agents can pay for access programmatically. Be clear-eyed here: there is no ratified well-known path for x402 yet. That path is a convention that tools probe by agreement, not a standard you can cite. It is the furthest from settled of anything on this list.

Markdown content negotiation. Not a file at all. It's a request: an agent sends Accept: text/markdown to your homepage, and a well-prepared site returns a clean Markdown representation instead of a wall of rendering HTML. Cheap to implement, immediately useful, and almost nobody does it.

Alongside these sit the unglamorous fundamentals that still do most of the work: your sitemap, your JSON-LD structured data, your HTTP link headers, and your robots rules.

Why almost none of this is implemented

Three reasons, and they are all rational.

The specs are moving. Building against a draft that shifts under you is genuine engineering cost with uncertain payoff. Teams that waited on early web standards were often vindicated.

There is no immediate traffic reward. Publishing an agent card today does not send you visitors tomorrow. The payoff is positional, not immediate, and positional bets are the first thing cut from a roadmap.

And most importantly, almost nobody is asking for it yet. The agents that would consume these files are still being built. This is infrastructure for a demand curve that has not arrived.

That is a completely reasonable case for waiting. It is also exactly what the case for waiting sounded like in 1996, and in 2007, and in 2010.

What to actually do about it, in order

Here is the honest priority order, and note that the top of it is deliberately unexciting.

First, fix the boring layer. Crawlable server-rendered text, a working sitemap, correct JSON-LD, and robots rules that do not accidentally block the assistants you want citing you. This is where nearly all of your real AI visibility lives today. A site with a beautiful agent card and a stray Disallow: / is a site nobody can see. Do not skip to the interesting part.

Second, publish an llms.txt. Cheap, useful now, and it makes your canonical sources explicit rather than inferred.

Third, do OAuth discovery properly if you have authenticated APIs. This is stable, standard, and useful regardless of how the agent web shakes out. It is not a speculative bet.

Fourth, publish an agent card or MCP server card if you actually run an agent or an MCP server. Note the condition. Publishing a card describing an agent you do not have is worse than publishing nothing, because it advertises a capability that will fail when something tries to use it.

Fifth, watch the drafts. Web Bot Auth and x402 solve real problems and are worth understanding. They are not worth building against this quarter unless you have a specific reason.

The pattern is straightforward: implement what is stable and useful now, understand what is coming, and resist the urge to build against specifications that are still arguing with themselves.

The map is the point

The reason to know this layer exists is not so you can rush out and populate every path tomorrow. It is so that when a client, a competitor, or a vendor says "you need an agent card," you can ask the right question: is that a settled standard, or a draft, and does it describe something we actually run?

Right now, on this subject, the confident people are usually the ones who have not read the specs. The paths are conventions in flux. The schemas disagree with their own implementations. The honest position is that this layer is real, it is coming, most of it is not ready, and the fundamentals still matter more than any of it.

Knowing precisely which is which is the whole advantage.

FAQ

What is /.well-known/?
A standardized directory at your domain root where services publish machine-readable metadata at predictable paths. It has existed for years for infrastructure purposes and is now being used to describe what AI agents can do with a site.

Is this the same as llms.txt?
No. llms.txt describes your content: what your site is about and which pages are canonical. The /.well-known/ layer describes your capabilities: your APIs, your agent, your authentication, your payment endpoints. One serves a reader, the other serves an actor.

Do I need an A2A agent card?
Only if you actually run an agent that other agents should be able to discover and call. Publishing a card for an agent that does not exist advertises a capability that will fail on first contact.

Is x402 a standard I should implement?
Not yet. x402 is the agent-payments direction built on HTTP 402, but there is no ratified well-known path for it. Tools that probe /.well-known/x402 are following a convention, not a finalized spec. Understand it; do not build production infrastructure against it today.

What is Web Bot Auth solving?
Bot identity. Today any client can send a user-agent string claiming to be GPTBot, and there is no cryptographic way to verify it. Web Bot Auth, an IETF draft, publishes signing keys so bots can prove their identity. Real problem, draft solution.

What should I do first if I only have an hour?
Check that you are not accidentally blocking AI crawlers in robots.txt, confirm your important content is server-rendered text rather than trapped in JavaScript or images, and publish an llms.txt. That hour will do more for your AI visibility than every file in /.well-known/ combined, at least for now.

See where your site actually stands

Sourceable's agent-readiness scanner probes this entire layer, from the fundamentals through to the experimental drafts, and returns a 0-100 score across five pillars with the specific fixes to reach the next level. It tells you which of these files you have, which you are missing, and which ones actually matter for your situation.

Fair warning, and we say this from experience: most sites score well on the first three pillars and near zero on Protocol Discovery. When we ran it on our own site, we scored 78 out of 100, with a 10 on exactly this layer. The gap is normal. Knowing its exact shape is what lets you close it deliberately instead of guessing.

Scan your site's agent-readiness, free

Only 36 Brands Show Up Across Every AI Platform. Here's What the "Universal 36" Do Differently.

Sourceable — Sun, 19 Jul 2026 05:00:00 +0000

Semrush just ran the numbers on 126 million AI search prompts. The headline finding should stop any marketer cold: out of the top 100 brands, only 36 stayed visible across ChatGPT, Gemini, Google AI Overviews, and Google AI Mode in every month of the study. Everyone else won in one AI engine and vanished in another.

If you assumed that ranking well in ChatGPT means you're covered in Gemini too, this is your wake-up call. AI visibility doesn't travel. You earn it platform by platform, or you don't have it.

What is the "Universal 36"?

The Universal 36 is Semrush's name for the 36 global brands that held top-100 visibility across all four major AI search platforms during every month of its 2026 AI Visibility Index — a study of 126 million U.S. AI search prompts run from January through April 2026. Names on the list include YouTube, Google, Reddit, Amazon, Facebook, Apple, Walmart, Disney, and Nintendo. What they share is broad consumer reach, sustained recognition, and a clear role in helping people compare options or finish a purchase.

In other words, 64 of the 100 most visible brands couldn't hold their spot everywhere at once. Consistency across engines is the exception, not the rule.

Why doesn't AI visibility carry from one platform to the next?

Because the engines read different webs. Semrush found that the four platforms pull from strikingly different source pools when they build an answer.

ChatGPT cites an average of 15 sources per response and leans heavily on community and reference sites like Reddit and Wikipedia. Gemini cites an average of just 3 sources per response, drawing from a tighter set that includes Wikipedia, Reddit, and YouTube. Fewer citations means less room for your brand to sneak in — and a different bar for what counts as a trustworthy source.

There's a second wrinkle most teams miss: being mentioned and being cited are not the same thing. A brand can get named in an answer without its own website appearing as the supporting source. On Gemini, Semrush found the overlap between mentioned brands and cited domains can drop to 30%. So you might be the answer while a competitor's blog, a Reddit thread, or a review site gets the click and the credibility.

That gap is exactly what tools like Sourceable exist to close — showing you not just whether you're named, but which sources each AI engine trusts to back up that mention, across ChatGPT, Claude, Gemini, and Perplexity.

Your category decides how hard this fight is

Not every industry is equally locked up. Semrush measured how concentrated visibility is per category, and the spread is wide.

In News and Media, the three most visible brands soaked up 82.9% of total category visibility. Consumer Electronics was nearly as top-heavy at 76.9% for the top three. But Finance was far more open — the top three held just 41.4% — and Industrial sat at 42.2%.

The read for you: if you're in a concentrated category, you're fighting entrenched giants for scraps of AI attention, and precision matters more than ever. If you're in a fragmented one, there's genuine daylight. The brands that move first to understand how each engine cites their space can claim ground before it consolidates.

Third parties are writing your AI story

Here's the part that stings for teams used to controlling their message: your AI narrative isn't built only from your own website anymore. AI platforms pull from customer reviews, community threads, independent publishers, retailers, and industry sources to decide how to describe and recommend you.

Semrush's example is Patagonia, which held an AI visibility score around 79–80 for the whole study. That consistency wasn't just Patagonia.com talking. It was reinforced by matching descriptions across OutdoorGearLab, REI, Switchback Travel, GearJunkie, and Reddit. When independent sources tell the same story about you, AI engines trust it. When they contradict each other, or say nothing, you drift.

Most teams can't even see the problem

The scariest stat isn't about visibility — it's about blindness. Semrush's companion survey found that 45% of marketing leaders can't accurately measure their brand's visibility inside AI-generated answers, and only 9% have the tools to track every relevant metric across platforms.

That measurement gap has real cost. Among organizations that fully integrate SEO and AI visibility into one workflow, 81% reported more traffic or leads from AI platforms. Among those managing the two separately, only 36% said the same. Treating AI visibility as a side project, rather than wiring it into your existing SEO and content work, leaves most of the upside on the table.

And the channel is only getting bigger. Adobe reports AI-driven traffic to U.S. retail sites climbed 1,324% between October 2024 and May 2026, with travel sites up 2,215% over the same stretch. This isn't a niche you can afford to leave unmeasured.

How to work toward cross-platform AI visibility

You won't join the Universal 36 next quarter. But you can stop guessing and start closing gaps:

Measure each engine separately. Your ChatGPT visibility tells you nothing about Gemini. Track ChatGPT, Gemini, Perplexity, and Google's AI answers as distinct scoreboards.
Separate mentions from citations. Know when you're named versus when your domain is the cited source. The second is what builds durable authority.
Audit your third-party story. Check whether reviews, community threads, and publishers describe you consistently. Contradictions confuse AI engines.
Pick your battles by concentration. In locked-up categories, go deep on the sources that actually get cited. In open ones, move fast.
Track over time, not once. The Universal 36 earned their spot by holding visibility every month. A single audit is a snapshot; visibility is a trend line.

Sourceable is built for exactly this — continuous tracking of when and how your brand shows up across the major AI models, so you can see which engines cite you, which ignore you, and where the gaps are before they cost you customers. See how your brand shows up across AI search →

FAQ

What is answer engine optimization (AEO)?
AEO is the practice of improving how your brand is mentioned and cited inside AI-generated answers from tools like ChatGPT, Gemini, Perplexity, and Google's AI Overviews. Unlike traditional SEO, which targets ranked links, AEO targets the synthesized response itself.

Does ranking #1 on Google mean I'll appear in AI answers?
No. Semrush's data shows AI platforms cite different source pools, and only 36 of the top 100 brands held visibility across all four platforms at once. A strong Google ranking helps but doesn't guarantee an AI citation.

Why does my brand appear in ChatGPT but not Gemini?
Because the engines cite differently. ChatGPT averages about 15 sources per answer and leans on Reddit and Wikipedia; Gemini averages roughly 3 and draws from a narrower pool. A source that gets you into one may not clear the bar in the other.

How do I track my visibility across AI platforms?
Use a dedicated AI visibility tool that monitors each engine separately. Sourceable tracks when and how your brand is mentioned across ChatGPT, Claude, Gemini, and Perplexity, so you can spot gaps engine by engine.

AI Search Sends 190x Less Traffic Than Google But It Convert up to 4x Higher

Sourceable — Sat, 18 Jul 2026 05:00:00 +0000

Google dominates referral volume, but AI assistants are creating a new class of visitors: fewer in number, far more likely to become customers.

Here's the uncomfortable math a lot of marketers are sitting on right now: ChatGPT already handles roughly 12% of Google's search volume, yet Google still sends about 190 times more clicks to websites. If you judge AI search by referral traffic alone, it looks like a rounding error. Judge it by who actually buys, and the picture flips.

That gap is the single most misunderstood number in search this year. Below is what the 2026 data actually says, why AI visitors behave differently, and how to decide whether AI search deserves a line in your budget. Sourceable exists precisely because this channel is hard to see in a standard analytics dashboard — more on that later.

Does AI search traffic convert better than Google organic?

Yes — consistently, across multiple independent 2026 studies. Ahrefs found that ChatGPT holds about 12% of Google's search query volume but Google sends roughly 190x more traffic to sites, because ChatGPT answers most questions inside its own interface without a click. But the visitors who do click through convert far above organic benchmarks: Semrush data shows AI-referred visitors converting at about 4.4x the rate of standard organic search. So AI search is a low-volume, high-intent channel, not a traffic firehose.

The traffic gap is real — and it's structural

The 190x figure comes straight from Ahrefs' clickstream analysis. Google makes up nearly 40% of referral traffic to the open web; ChatGPT sits around 0.21%. ChatGPT's click-through rate to external sites runs roughly 96% lower than Google's (Ahrefs).

Why so lopsided? Two different business models. Google was built to route you out to the web. ChatGPT was built to keep you in the conversation. When an AI answers your question in three tidy paragraphs with a couple of inline citations, most people never click. That's not a bug you can optimize away — it's the product working as designed.

So if you're waiting for AI referral traffic to rival Google in raw volume, you'll be waiting a long time. That's the wrong scoreboard.

The conversion story is where it gets interesting

Traffic measures attention. Conversion measures intent. And on intent, AI visitors punch far above their weight.

A few verified data points from 2026:

Semrush: AI-driven visitors convert at roughly 4.4x the rate of standard organic search across industries (reported via AirOps).
Adobe: AI-referred shoppers to U.S. retail sites converted 42% better than non-AI traffic in March 2026 — a record high for the channel (Digital Applied summary of Adobe data).
Ahrefs: In one analysis, AI-referred visitors made up just 0.5% of sessions but drove 12.1% of signups — roughly a 23x conversion differential (via Averi).

The numbers vary by source and by industry, but the direction never does: people arriving from an AI answer convert well above the organic baseline.

Why the intent is so much higher

Think about how someone reaches your site through ChatGPT or Perplexity. They didn't type two keywords and skim ten blue links. They asked a real, often specific question — "best HRIS for a 40-person remote team" — read a synthesized answer, and then clicked your name because the model surfaced you as a credible option.

That click is pre-qualified. The visitor has context, a shortlist, and a reason to be on your page. It's closer to a warm referral than a cold search. Low volume, but the volume that shows up is already halfway down the funnel.

What this means for your strategy

Stop grading AI search on traffic volume. Grade it on pipeline. A channel that sends 500 visits a month and converts at 14% can out-earn one that sends 50,000 and converts at 2%.

Three practical moves:

Measure the right metric. Segment AI referral traffic in GA4 (ChatGPT, Perplexity, Gemini, Claude referrers) and track conversion rate and revenue per session — not just sessions. If you only look at volume, you'll underrate the channel and defund it.
Get cited, not just crawled. Since most AI answers never produce a click, the citation itself is the impression. Being named in the answer builds consideration even when nobody clicks through. That means the real KPI is how often you show up in AI answers, not just how much traffic trickles back.
Track share of voice across models. ChatGPT, Perplexity, Gemini, and Claude each cite different sources for the same question, and their answers shift week to week. You can't manage what you can't see.

That third point is exactly the visibility gap Sourceable is built to close. It monitors when and how your brand gets mentioned across ChatGPT, Claude, Gemini, and Perplexity, so you can watch your presence in AI answers the way you already watch your Google rankings — and prove the channel's worth before the traffic ever shows up in analytics.

The bottom line

AI search is not a traffic play. It's a conversion play. The clicks are scarce because the models keep users inside the chat, but the clicks that escape are among the highest-intent visitors you'll get anywhere. Ignore the channel because the traffic looks small, and you're leaving your most qualified prospects to whichever competitor the model decides to name instead.

FAQ

Is AI search traffic worth tracking if it's such a small share of visits?
Yes. Even at under 1% of sessions, AI referrals routinely convert several times higher than organic — one Ahrefs analysis tied 0.5% of sessions to 12.1% of signups. Value per visit matters more than visit count.

How much traffic does ChatGPT actually send compared to Google?
Google sends roughly 190x more clicks to websites, per Ahrefs, even though ChatGPT holds about 12% of Google's query volume. ChatGPT answers most questions in-app, so click-through is far lower.

Why do AI referrals convert better than organic search?
Intent. AI visitors arrive after reading a synthesized answer that already surfaced you as an option, so they land with context and a shortlist — closer to a warm referral than a cold keyword search.

How do I know if my brand is showing up in AI answers?
Standard analytics only captures the clicks, not the citations. Tools like Sourceable track how often and in what context your brand appears in AI-generated responses across the major models.

Want to see how often AI actually recommends your brand? Start tracking your AI visibility with Sourceable and measure the channel that converts before it ever shows up in your traffic reports.

There's No Single "AI SEO": Why ChatGPT, Perplexity, and Google Cite Three Different Webs

Sourceable — Fri, 17 Jul 2026 06:50:01 +0000

You optimized one page. You're being judged by at least three referees, and they don't agree on the rules.

A new benchmark from the AI-visibility platform CiteLens ran 320 real buyer queries through ChatGPT, Perplexity, Claude, and Google's AI Mode, then checked every source each engine cited against that same query's Google and Bing results. The finding is blunt: SEO gets you into Google's AI and Perplexity, but it barely touches ChatGPT. If you've been treating "get cited by AI" as one project, that's the problem. Tools like Sourceable exist because you can't fix what a single search dashboard can't even see.

The short answer: is optimizing for AI search one strategy or several?

It's at least three. In the CiteLens study, Google's AI Mode pulled 93% of its citations from Google's top-10 organic results and Perplexity pulled 89% — so for those two, classic SEO is the entry ticket. Claude pulled only 53%, and ChatGPT just 30%, meaning 70% of what ChatGPT recommended ranked in neither Google's nor Bing's top 10. Winning across all of them takes SEO and brand authority, not one or the other.

Three engines, three rulebooks

The most useful thing in the CiteLens data isn't a single number — it's that the engines cluster into three distinct behaviors. Here's how each one decides who gets named.

Search-ranking machines: Google AI Mode and Perplexity

These two behave like AI wrappers around the classic index. Citation frequency tracked Google rankings almost perfectly — a correlation of 0.92 for Google AI Mode and 0.87 for Perplexity (MarTech Series). If you rank on page one for a query, you have a strong shot at being cited when someone asks the AI version of that same question.

The practical read: your existing SEO program is doing double duty here. Keep earning top-10 positions for the questions your buyers actually ask, because on these two engines, rank is the qualifier.

The brand machine: Claude

Claude broke the pattern. It was the only engine whose citations tracked brand search demand more than search ranking — it leaned toward well-known, global names. And 58% of its citations went to sites with a Wikipedia presence.

That tells you something concrete. On Claude, being a recognized entity — the kind of brand that has a Wikipedia page, consistent mentions, and name recognition — matters more than which position you hold for a given keyword. You don't win Claude with a better meta description. You win it by being a brand the model already knows.

The black box: ChatGPT

ChatGPT followed neither rulebook. Its correlation with Google ranking was near zero, and its correlation with brand popularity was near zero too. Only 21% of its citations went to Wikipedia-backed sites — versus Claude's 58%. It kept surfacing a handful of niche domains that rank nowhere and command little search demand.

And the popular theory that ChatGPT just mirrors Bing? It didn't hold. Fewer than 4% of ChatGPT's citations showed up in Bing's top 10. Whatever ChatGPT is doing, it's baked into the model's own preferences, not borrowed from a search index you can reverse-engineer.

This is the engine that keeps founders up at night: the one where a competitor you've never heard of gets recommended and you can't figure out why. The honest answer is that you can't optimize your way in with keywords alone. You measure where you stand, watch which niche sources keep appearing, and work to get mentioned in those places.

Why this breaks the "AI SEO" playbook

Most advice still treats AI visibility as a single funnel: write good content, structure it well, get cited everywhere. The CiteLens numbers say that's only true for the engines that lean on Google's index — and even those two aren't identical.

Here's the strategic split the data supports:

For Google AI Mode and Perplexity: double down on traditional SEO. Rankings convert directly into citations. This is the most measurable, most controllable slice of AI visibility you have.
For Claude: invest in entity authority. Wikipedia presence, consistent brand mentions across reputable sites, and recognizable name equity. This is slower PR-and-brand work, not on-page tweaks.
For ChatGPT: treat it as its own research problem. Track your presence continuously, identify the niche domains it favors in your category, and pursue mentions there. Assume there's no ranking shortcut.

A second CiteLens study reinforces how little of this the old tools can see. Across 500 commercial prompts in 126 categories on Google AI Overviews, 60% of the domains an AI answer cited didn't appear anywhere in Google's organic top 10 for that query (CiteLens via National Law Review). User-generated content dominated: 74% of answers cited YouTube and 84% cited forums or other UGC. Your rank-tracking dashboard is blind to most of that.

The measurement problem underneath all of this

There's a catch that makes single checks useless. In that same 500-prompt study, asking Google's AI the identical question three times returned the full set of cited sources only 81% of the time — meaning roughly three sources shift on every repeat. Different engines already disagree with each other; each engine also disagrees with itself run to run.

So "I asked ChatGPT and we showed up" is not data. It's one sample from a noisy, engine-specific distribution. To know where you actually stand, you need to measure the same queries across ChatGPT, Claude, Gemini, and Perplexity, repeatedly, and track how your share of citations moves over time. That's the gap Sourceable is built to close — showing when and how your brand gets named across the major AI models, so you can tell a real trend from a lucky run.

FAQ

Does ranking on Google get me cited by AI?
On some engines, strongly. CiteLens found Google's AI Mode drew 93% of citations from Google's top 10 and Perplexity 89%. But ChatGPT pulled only 30% and Claude 53%, so Google rank is a partial signal at best — and near useless for ChatGPT specifically.

Why does ChatGPT recommend brands that don't rank on Google?
Because its citation behavior correlated near zero with both Google ranking and brand search demand in the CiteLens benchmark. It favors a set of niche sources tied to the model's own preferences, not to any search index — fewer than 4% of its citations even appeared in Bing's top 10.

What actually helps me show up in Claude?
Entity authority. Claude's citations tracked brand recognition over ranking, and 58% went to sites with a Wikipedia presence. Building a recognizable, well-referenced brand matters more here than optimizing an individual page.

Can one audit tell me my AI visibility?
No. Repeat runs of the same query changed roughly 3 sources out of the set even on a single engine, and the four engines disagree sharply with each other. You need continuous, cross-engine measurement.

The takeaway

Stop asking "how do I rank in AI search" as if it's one question. Ask three: how do I keep winning Google's index (for AI Mode and Perplexity), how do I build entity authority (for Claude), and how do I earn mentions in the niche sources ChatGPT trusts. Each needs different work — and all three need continuous measurement, because these engines change their answers on you, sometimes within the same afternoon.

Want to see which of the three rulebooks is already working for your brand — and which one is leaving you invisible? Check your visibility across every major AI model with Sourceable.

Google Search Console Now Shows Your AI Overview Impressions — But It Can't See the Half of AI Search That Matters Most

Sourceable — Tue, 14 Jul 2026 06:07:04 +0000

Google finally gave SEOs a native way to measure AI visibility. On June 3, 2026, it rolled out Search Generative AI performance reports inside Search Console, showing how often your pages appear in AI Overviews and AI Mode. It's a real step forward. It's also only part of the picture — and if you treat it as the whole scoreboard, you'll miss where most of your AI referral traffic actually comes from.

What does Google's new AI performance report actually show?

Google's Search Generative AI performance report shows impressions for AI Overviews and AI Mode: how often URLs from your site appeared inside those features, which pages surfaced, and breakdowns by country, device, and date. It launched June 3, 2026, and is rolling out to a subset of sites first. Critically, it does not include click data, and it only covers Google's own AI surfaces — nothing about ChatGPT, Claude, Gemini, or Perplexity.

That's the report in two sentences. Now here's why the two things it leaves out matter so much.

The two blind spots: no clicks, and no other engines

Start with the missing clicks. Google's own announcement confirms the report includes impressions, pages, countries, devices, and dates — but not clicks. So you can see that your page was pulled into an AI Overview, but not whether that appearance sent anyone to your site. Given what we already know about AI Overviews suppressing clicks, that's a meaningful gap. Studies compiled through 2026 show AI Overviews now appear on roughly 47–64% of queries and cut organic click-through rates by anywhere from 15% to 46%, with one analysis of 68,000 queries finding a 46.7% relative decline in clicks. Impressions without clicks can look like a win while your traffic quietly erodes.

The bigger blind spot is scope. Search Console only measures Google. But Google isn't where the standalone AI referral traffic is concentrated. According to Previsible's 2026 AI Traffic Report, ChatGPT accounts for 92.4% of all trackable referral traffic from standalone AI platforms. Other methodologies put the split differently — SE Ranking's study has ChatGPT at 74.78%, Gemini at 11.56%, Perplexity at 7.23%, and Claude at 2.62% — but every one of them agrees on the headline: a large majority of AI referrals now come from engines Google's report will never show you.

So the new report answers "how am I doing in Google's AI features?" while leaving the more urgent question — "how am I doing across the AI tools people actually ask?" — completely unanswered.

Why one dashboard was never going to be enough

There's a structural reason no single native tool can cover this. Each AI engine is its own walled garden with its own index, its own citation logic, and its own idea of who the authoritative sources are. Ask ChatGPT, Perplexity, and Gemini the same buying question and you'll routinely get different brands cited. Google can report on Google. OpenAI doesn't hand you a Search Console for ChatGPT. Perplexity won't email you when it stops citing you.

This is compounded by ownership confusion inside companies. A Muck Rack study released July 9, 2026 found that 73% of PR professionals now see AI search visibility as at least somewhat important — yet 29% say no one at their organization owns it. A native Google report doesn't fix an ownership vacuum. It just gives one team one slice of one engine.

This is the gap Sourceable was built for. It tracks how and when your brand shows up across ChatGPT, Claude, Gemini, and Perplexity — the surfaces Search Console can't touch — so "our AI visibility" stops meaning "our AI Overview impressions in Google" and starts meaning your presence everywhere buyers are actually asking.

How to use the new Search Console report without being misled

The report is genuinely useful. Just scope your expectations. A few practical moves:

Treat AI Overview impressions as a coverage signal, not a traffic number. Rising impressions tell you Google trusts your pages enough to synthesize them into answers. Pair that with your actual organic clicks to see whether that coverage is helping or cannibalizing.

Check the content-blocking toggle while you're in there. As Search Engine Land reported, Google shipped a control letting sites opt out of appearing in AI features. Most brands chasing visibility should leave it on — but know it exists before someone flips it by accident.

Don't over-engineer schema for AI. Google's own guidance says there's no special markup required for AI Overviews or AI Mode, and it just deprecated FAQ rich results entirely as of May 7, 2026. Clean, well-structured, genuinely useful content still wins.

Then measure the other 90%. Whatever Search Console shows you about Google, assume there's a larger story playing out in ChatGPT and Perplexity that you can only see with a cross-engine tool.

Frequently asked questions

Does Google's AI performance report show clicks from AI Overviews?
No. The report includes impressions, pages, countries, devices, and dates, but not click data. You can see that your page appeared in an AI Overview or AI Mode answer, but not how many people clicked through.

Does Search Console track my visibility in ChatGPT or Perplexity?
No. Search Console only measures Google's own surfaces, including AI Overviews and AI Mode. It shows nothing about ChatGPT, Claude, Gemini, or Perplexity — which is where the majority of standalone AI referral traffic now originates.

Do I need special schema to appear in AI Overviews?
No. Google states there is no special structured data required for AI Overviews or AI Mode. FAQ rich results were deprecated on May 7, 2026, though FAQPage markup won't harm your pages if it stays in place.

How do I measure brand visibility across all the major AI engines?
Use a tool built to monitor multiple engines at once. Sourceable tracks when and how your brand is mentioned across ChatGPT, Claude, Gemini, and Perplexity, filling the gap Google's native reporting leaves open.

The takeaway

Google's new report is worth checking weekly — it's the first native window into how AI reshapes your Google presence. But a window into one room isn't a view of the house. With ChatGPT and its rivals driving the bulk of AI referrals and each engine citing a different set of sources, measuring only Google means flying half-blind. See where your brand actually shows up across every major AI engine with Sourceable.