DEV Community: Bulti

How much does a GEO agency cost? An outsource-vs-DIY calculator

Bulti — Wed, 26 Aug 2026 01:01:00 +0000

⏰ The 30-second version

GEO agency cost usually combines strategy, content production, and ongoing AI-visibility monitoring.
"GEO agency" is a search term that actually exists in autocomplete — the move to look for someone to hire has already begun.
The outsourcing math: usually per-article rate × number of articles + a strategy fee. Per-article rates in the content-marketing market range widely, from tens to hundreds of dollars.
The DIY math: instead of money, time — question research + writing + verification. Learn the formula and it drops to 1–2 hours per piece.
The deciding factor is one thing — who holds the question list and the priorities. Outsourcing without them is expensive shooting with no target.

The structure of agency rates: know what you're buying first

Whether it flies a GEO banner or not, a content agency quote usually breaks into three chunks — ① strategy/keyword planning ② article production (per-article rate) ③ monthly operation. Rates vary widely with skill, expertise, and industry, so it's hard to state a single figure — but just knowing the structure gives you the eyes to read a quote.

What to check matters more than the price: does it present, as data, which questions the content answers? A quote that promises "8 posts a month" with no question list fills the article count but never the coverage.

The real cost of DIY: doing the math on your time

DIY isn't free — it's time. Do the honest math:

The time cost of one DIY GEO post

[ ] Question selection & fact-checking 20–30 min
[ ] First draft 40–60 min
[ ] Structure check (the 5 moves) 10 min
[ ] FAQ & schema 10 min

At first, budget about two hours per piece — that's realistic. But once the formula (the 5 moves) is second nature it halves, and above all you gain a strength no agency can buy: content written by someone who knows the product. In the AI-citation contest, specificity is a weapon — and the source of specificity is product understanding.

The middle path: direction by data, hands by tool

The space between outsourcing and full DIY isn't empty. Automate question discovery and prioritization with a tool, and do only the writing yourself (or take a draft and refine it). For reference, Bulti is designed as that middle path — the audit pulls 30 questions and priorities, and if you want, produces a draft for the seller to review and publish. It starts at $9 for a single-use pass, so you can set your whole direction for a fraction of the price of one outsourced article.

When to choose what

Outsourcing fits when: you need a high article volume, you have an internal eye to review, and you can hand over a question list you control.
DIY fits when: product knowledge is the weapon in your category (fights over ingredients, specs, usage), and you're early enough that exploring direction matters more.
The middle path fits when: you're short on time but want to keep your own voice — which is most solo sellers.

FAQ

Q. What's the difference between a GEO specialist and a regular blog agency?
Judge by deliverables, not the sign out front. If there's a question list, coverage measurement, and a citation report, it's the GEO kind; if there's only ranking language like "top-placement guaranteed," it's the old SEO kind.

Q. If I outsource, is there still work I have to do myself?
Approving the question list and fact-checking can't be delegated. Product figures and efficacy claims are territory only the seller can own.

Q. How is a Bulti draft different from an outsourced article?
It starts from question-coverage data, and it doesn't invent numbers that need verifying — it leaves them flagged as [needs verification]. Final review and publishing are designed to be the seller's job.

Originally published at en.bultibulti.com. We run a free GEO audit there — it checks one of your pages against 30 buyer questions, no signup.

A GEO case study — what we're doing with this very series

Bulti — Wed, 26 Aug 2026 01:00:35 +0000

⏰ The 30-second version

Let's start honest — instead of inventing someone else's verified case study, we're opening up what we're doing right now.
What we did: picked topics from real search demand · GEO structure across every post · FAQ schema · a public learning library connected to a report-first product.
Every topic maps 1:1 to a search term that actually exists in autocomplete — not a single topic was picked on gut feel.
This page is the lab. As results accumulate, we'll update this post.

Why our own case? Because we can't fake it

"GEO optimization case study" is a search term that actually exists in autocomplete. But the market is early, and there are almost no published e-commerce cases. Writing a case that doesn't exist as if it did violates our principle (we don't publish numbers we can't verify), so we picked the most honest case available — ourselves. Bulti sells a GEO audit tool, and it runs this series with the formula that tool teaches. If it doesn't work, there's no hiding it.

Execution 1 — we picked topics by search volume, not gut feel

The first 18 topics in this series came from keyword research. We measured search volume and trends for 25 candidate keywords with Google Keyword Planner, confirmed real questions with an autocomplete sweep ("what is GEO optimization," "cited in Google AI Overviews"), then mapped each search term to a post. The library has since expanded to 36 articles as of 2026-08-11. The principle remains the same: answer a real practitioner question on one canonical page.

Execution 2 — we applied the formula we teach to every post

Every post in this series follows the 5 GEO moves exactly. The "30-second version" up top is the answer-first lead, the subheads are question-style, the paragraphs are self-contained, every post carries three FAQs, and those FAQs have schema embedded. View the source of this very page and you'll see the FAQPage structured data for real.

Execution 3 — we made the learning library public and moved conversion to the report

All 36 Learn GEO articles are public. Readers, search crawlers, and AI systems receive the same complete article, so there is no ambiguity about what is accessible or citable. Commercial intent lives in a separate report-first flow: a marketer can generate and review a brand report at /a, then decide whether to buy the complete report and 60-day operating program. The content earns discovery and trust; the report demonstrates product value before checkout.

The honest scorecard so far

The series has just been published. So this post has no "traffic multiplied in 3 months" numbers — because those numbers don't exist yet. Instead, we make a promise: as soon as we confirm AI citations, search traffic, and Overview source adoption, we'll add them here, with dates. This post itself becomes a time-lapse of GEO.

FAQ

Q. Can you call it a case study with no results yet?
"What we did and why" is all we can show you right now, and we judged that opening it up honestly is more useful than an invented success story. Results will keep getting added to this page.

Q. To apply the same approach to my brand, where do I start?
The order is the same — question list (how to find them) → measure coverage (self-check) → write for the empty questions first (the formula). You can automate the measurement with the free audit.

Q. How do I measure keyword data?
Google Keyword Planner (needs an ad account) is the baseline. If you don't have an account, an autocomplete check alone can validate real demand.

Originally published at en.bultibulti.com. We run a free GEO audit there — it checks one of your pages against 30 buyer questions, no signup.

Is my content for SEO or for GEO? A 3-minute test

Bulti — Tue, 25 Aug 2026 01:00:35 +0000

⏰ The 30-second version

SEO content and GEO content can look similar but are designed to different standards — keyword density vs. question coverage.
Five tests: is the title a search term or a question · where's the conclusion · are paragraphs self-contained · what kind of subheads · is there an FAQ.
Happily, they're not mutually exclusive — fix a piece into GEO structure and you keep the SEO performance while gaining the AI surface.
You don't have to fix everything. Start with the posts that sit on buying-decision questions.

The 3-minute test: ask your content five things

Open one existing post and check.

The telltale signs of SEO content

[ ] Title is a keyword pileup (Best 7 OO Recommendations)
[ ] Conclusion is at the very end
[ ] Pronoun links like as mentioned above / this product
[ ] Subheads are just numbers (1. 2. 3.)
[ ] No FAQ

Three or more, and it was designed to rank. That doesn't make it bad content — it was just that era's optimization. But it loses out at the AI extraction stage: with the conclusion at the back, no direct-answer sentence gets pulled, and pronoun links go meaningless the moment they're cut. The mechanics are covered in how AI decides what to quote.

The rewrite order: don't rewrite from scratch — rearrange first

You don't have to throw out existing content. Follow the order and one piece takes 30 minutes.

Conclusion to the top — move the last paragraph of the body to the first. This alone gets you halfway.
Subheads into questions — "2. Ingredients" → "Which ingredients should sensitive skin avoid?"
Unpack pronouns — "this product" → the product name. Make each paragraph stand alone.
Add 3 FAQs — pull real questions from reviews and inquiries and attach them at the end.
Polish the title — keep the keyword but reshape it into a form that answers a question.

Because the keywords don't disappear, your existing search traffic holds. Add an extractable structure on top and the AI surface opens — one piece running in two arenas.

Priorities: which content first?

Rewriting everything is a trap. The rule is one thing — start with posts on questions closest to a buying decision. Comparisons ("A vs B"), selection criteria ("what do I look at to choose"), and use conditions ("can someone like me use this") are first priority. If you're not sure which questions your posts cover, the free audit maps it across 30 questions.

FAQ

Q. Won't rewriting drop my existing search ranking?
A structural improvement that keeps the keyword and URL is generally safe. If anything, putting the conclusion up top helps dwell time and comprehension. Just keep your core keyword when you change the title.

Q. For new content, should I just write in GEO structure from the start?
Yes. Writing new posts with the 5 GEO moves from the start is cheaper than rewriting. Apply rewrites to your proven, traffic-earning posts first.

Q. How many do I have to fix before results show?
Coverage beats count. Five posts that precisely cover five core questions show results faster than twenty scattered ones.

Originally published at en.bultibulti.com. We run a free GEO audit there — it checks one of your pages against 30 buyer questions, no signup.

The questions you're losing to competitors — there's a way to find them

Bulti — Tue, 25 Aug 2026 01:00:11 +0000

⏰ The 30-second version

Every customer question has a fixed set of seats in the AI answer, and if you're not in them, a competitor is using that seat.
The diagnostic is a four-quadrant grid — lay out your question list and sort into both answer / only me / only competitor / no one.
The priorities are "only competitor" (being lost — reclaim) and "no one" (open territory — claim).
Do it without the emotional drain — this is map-making, not envy.

What "losing" actually means

If AI mentions only a competing brand for "Can a protein shake replace a meal?", the competitor takes the first impression of every customer who asks that question. There's no second or third place like in search rankings — if you're not in the answer, you're at zero. That's why checking, question by question, "who's using that seat" becomes the new baseline of competitive analysis.

Drawing the four-quadrant grid: 30 minutes does it

Prepare a list of 30 questions (how to build it), throw each at the AI, and mark whether your brand and your competitor's brand appear. Then every question falls into one of four boxes.

Quadrant	Meaning	Action
Both answer	Head-to-head zone	Win with a more specific answer
Only me	An asset to defend	Hold it with fresh content
Only competitor	Being lost	Reclaim with a better answer to the same question
No one	Open territory	Claim it by writing first

Where to start: claiming beats reclaiming

Counterintuitively, it's more efficient to hit the "no one" box before the "only competitor" box. Reclaiming means fighting content that's already entrenched, while open territory is yours just by walking in. Grow your coverage through claiming to build momentum, then reclaim the lost questions closest to a buying decision first.

One tip when writing a reclaim piece — write content that answers more precisely, not content that attacks the competitor. AI has no interest in brand fights; it just picks the document that answered the question better.

By hand vs with a tool

Drawing the grid once is fine by hand (30 questions × 1–2 AIs is a half-day). The problem is maintenance — the map shifts every time you publish content and every time a competitor moves. Bulti's competitive comparison audit draws this grid automatically; just enter your page and your competitor's brand name. You don't even need the competitor's URL — the brand name is enough.

FAQ

Q. Shouldn't I compare against several competitors?
One — the one most on your mind — is enough to start. The shape of the grid doesn't change much when the rival changes, and the "no one" box holds regardless of who the rival is.

Q. Isn't the reason we're not mentioned that we're a small brand?
More often the cause is missing content, not size. Without a document that answers the question, even big brands get left out — which conversely means there's room for a small brand to get in.

Q. Once I write for a lost question, when do I reclaim it?
Think in terms of weeks, given the indexing cycle. If it doesn't happen at once, revise by raising the specificity of your answer (numbers, conditions, use situations).

Originally published at en.bultibulti.com. We run a free GEO audit there — it checks one of your pages against 30 buyer questions, no signup.

FAQ schema, no coding required — a copy-and-paste guide

Bulti — Mon, 24 Aug 2026 01:00:38 +0000

⏰ The 30-second version

FAQ schema is writing your page's questions and answers a second time, in a format machines understand.
The payoff: search and AI recognize your FAQ as precise question-answer pairs. A plain-text FAQ already delivers GEO value, but schema makes that signal crisp.
The method is copy-paste — swap the questions and answers in the template below into your page, done.
One caution: don't put content in the schema that isn't on the visible page (it violates search-engine guidelines).

What schema does: a machine-readable copy of the same content

If a human-readable FAQ already exists on your page, schema is writing that same content once more in a structured format called JSON. It's essentially labeling for machines — "the question starts here, the answer starts here." "Schema markup" gets 210 searches a month — a card only the informed play, but the signal is clear relative to the effort of adding it.

Copy-paste template

Put the code below into your page's HTML and change only the bold parts to your own questions and answers.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "Question 1 here",
      "acceptedAnswer": { "@type": "Answer", "text": "Answer 1 here" }
    },
    {
      "@type": "Question",
      "name": "Question 2 here",
      "acceptedAnswer": { "@type": "Answer", "text": "Answer 2 here" }
    }
  ]
}
</script>

Where to put it, by platform — on hosted store builders (Shopify and the like), use the page settings' "Edit HTML" or "script insert" area; on WordPress, a Custom HTML block. On platforms that block script insertion, go with a plain-text FAQ and no schema (it still works).

How to verify: a 1-minute check

Paste your page URL into Google's Rich Results Test (search.google.com/test/rich-results) and it tells you immediately whether the FAQ was detected. A green check means success.

FAQ schema in 3 steps

[ ] Write a human FAQ on the page first
[ ] Fill the same content into the template and insert it
[ ] Verify with the Rich Results Test

Validation is not a rich-result guarantee

Google's Rich Results Test confirms whether the markup is valid and detected. It does not promise that Google will show an expanded FAQ result. Eligibility and display are decided by Google, and most commercial sites should treat visible, helpful answers as the primary benefit.

Keep the order: content comes first

Schema is an amplifier, not a generator. A good FAQ comes first (which questions to include), and schema just delivers it crisply. Then audit whether a page has visible FAQs and readable text. For reference, this very page you're reading has FAQ schema embedded — we practice what we teach.

FAQ

Q. Does adding schema raise my search ranking?
It doesn't raise ranking by itself. Valid markup can help machines interpret visible question-answer pairs, but it does not guarantee an expanded FAQ result in Google.

Q. How many FAQs should I include?
Three to five per page is about right. Focus on pre-purchase questions (shipping, how to use, comparisons), and always keep the on-screen FAQ and the schema content identical.

Q. Should I add other schemas (review, product) too?
For a product page, Product schema has value too. But FAQ is the easiest place to start and has the biggest overlap with GEO. Add them one at a time.

Originally published at en.bultibulti.com. We run a free GEO audit there — it checks one of your pages against 30 buyer questions, no signup.

Product page vs blog — which one does AI actually read?

Bulti — Mon, 24 Aug 2026 01:00:13 +0000

⏰ The 30-second version

Whether AI can read you is decided by the surface — only pages where the text physically exists in the HTML get read.
Most marketplace listings (Amazon, Etsy, and the like) are a single image — lush to the human eye, a blank page to AI.
GEO's home turf is your own store's product pages and your blog. One text post is more visible than ten image-based detail pages.
This isn't a call to quit marketplaces — it's a call to add one more text surface for AI to read.

Same content, different fate: image vs text

You've probably poured ingredient tables, usage instructions, and comparison charts into your listing. The problem is that most of it is built as an image. For design freedom, the industry standard became one big flat image.

To a human, it makes no difference. To AI, it's everything. AI reads a page's text and uses it as material — and letters inside an image aren't text. The result: the page packed with the most information becomes, to AI, the emptiest page of all.

**** —

AI-readability by surface

Surface	Can AI read it?	Notes
Own-store product page (text)	✅	GEO home turf — best when structured
Own-store blog	✅	Ideal for question-style content
Marketplace listing (image)	❌	Only text regions, partially
Marketplace app listing (image)	❌	Same
Hosted blog platform	✅	Read if written as text

That "marketplace visibility" searches fell -24% in three months isn't unrelated to this current. Interest in ranking inside the marketplace is fading, while exposure outside it (AI, search) has become the new variable.

The realistic prescription: don't move surfaces — add one

This isn't a call to abandon your marketplace revenue and come to your own store. Keep selling as you do — just add one text surface AI can read.

The cheapest way is to move your listing's key content into a text post. Ingredients, how to use, frequently asked questions — it's content you already have, so there's nothing new to write. That post gets cited in AI answers, and customers flow in, whether to the marketplace or your own store. Follow the GEO formula for the structure when you move it.

FAQ

Q. Do I have to rebuild my listing as text?
No, leave the listing as-is. Images are better for conversion. Just keep one text version of the same content on your blog and that's enough.

Q. I don't have my own store — where do I write?
You can start on a text blog platform. As long as you write in text, AI reads it. Long-term, building it on your own domain becomes an asset.

Q. Can't I just add alt text to the images?
Alt text is better than nothing, but it rarely gets treated as body content. To become material for extraction and citation, you need actual body text.

Originally published at en.bultibulti.com. We run a free GEO audit there — it checks one of your pages against 30 buyer questions, no signup.

Writing to get cited in Google AI Overviews — the conditions for becoming a source

Bulti — Thu, 20 Aug 2026 01:00:57 +0000

⏰ The 30-second version

The essence of getting cited in an AI Overview is one thing — becoming content that's easy for the summarizer to lift.
Four conditions: a title that matches question-style searches · a direct answer in the first paragraph · an itemized body · a text surface.
"Cited in Google AI Overviews" is real enough to appear in autocomplete — the competition has already begun.
The checking routine is simple — search your target term yourself and see whether you're in the source links.

Citation isn't luck: how the Overview picks its material

An AI Overview builds a summary from the documents in the search results. Which documents get chosen as material? Think from the Overview's point of view and the answer appears — documents already organized in a way that's easy to summarize.

To answer "recommend a moisturizer," the Overview has to lay out "by what criteria, and which products get mentioned." Content where the criteria and items are already clear can be lifted as-is; content that meanders like a diary is hard to use as material.

The four conditions of citable content

A title that matches the search — "Tips for a Dewy Summer" (✗) vs. "3 Criteria for Choosing a Summer Moisturizer for Oily Skin" (✓). The Overview starts by checking the relevance between the search term and the document.
A direct answer in the first paragraph — the conclusion has to be up front so the summarizer can grab it quickly.
An itemized body — structure the criteria with numbers or subheads. "First, ingredients. Second, texture." maps directly onto the Overview's summary format.
A text surface — image-based detail pages don't even make the candidate pool. Blogs and text product pages are home turf.

As you may have noticed, these are the same conditions as the GEO formula. Meaning: you don't have to write separate content just for Overviews.

The checking routine: once a month, 10 minutes

Pick 5–10 target terms and search them yourself once a month. Look at two things — ① is this a term the Overview appears for, and ② is your content in the source links? If the source is a competitor's blog, dissecting that post's structure alone gives you the blueprint for what to write next.

To check many terms at once, the free audit measures your coverage across 30 questions.

FAQ

Q. If I get cited in an Overview, do clicks actually go up?
Overview sources appear as links right below the answer, so the placement beats an ordinary list. More than that, the credibility of being "a source the AI chose" attaches to it.

Q. Can a marketplace product page get cited?
An image-based detail page with almost no text is hard. Moving the same content into a text blog is the realistic workaround.

Q. How long until a new post gets reflected in an Overview?
It varies with the search-indexing cycle, but usually days to a few weeks. Checking 2–4 weeks after publishing is more accurate than right away.

Originally published at en.bultibulti.com. We run a free GEO audit there — it checks one of your pages against 30 buyer questions, no signup.

Where do you find the 30 questions your customers ask AI?

Bulti — Thu, 20 Aug 2026 01:00:32 +0000

⏰ The 30-second version

GEO doesn't start with writing — it starts with building a question list. If you don't know what they're asking, you have nothing to answer.
There are three question warehouses — your reviews, customer inquiries, and search autocomplete. All free, all real data.
The frame that fills them in without gaps has five axes — selection criteria / comparisons / side effects & cautions / how to use / price.
The target is 30. Sounds like a lot, but it's just five axes × six each.

Warehouse 1 — Reviews: your customers are already writing the answers

A review is an answer to a question. Behind "less dry and chalky than I expected" sits the question "Is there a chicken breast that isn't dry?" Skim just 50 of your own reviews and ten questions surface immediately. Competitor reviews are even better — especially negative reviews, which show exactly what customers worried about before buying, which is exactly what they'll ask AI.

Warehouse 2 — Inquiries: the purest question, verbatim

Customer-support inquiries are questions in the customer's own words. "Is this safe during pregnancy?" "Can I put it in the washing machine?" Don't polish these — drop them into your question list almost as-is. Customers ask AI the same way.

Warehouse 3 — Autocomplete: real demand, straight from the search box

Type a category keyword into Google's search box and autocomplete follows. These are words people actually searched. Type "probiotics" and you get "probiotics when," "probiotics side effects," and so on. Anything in autocomplete is a pre-validated question.

Filling the blanks with five axes

Place the questions you gathered from the three warehouses onto the five axes, and the empty axes reveal themselves.

The five question axes (six each = 30)

[ ] Selection criteria — what should I look at to choose?
[ ] Comparison — how are A and B different?
[ ] Side effects & cautions — should certain people avoid this?
[ ] How to use — when and how do I use it?
[ ] Price — is this a fair price?

In my experience, sellers gather "selection criteria" and "how to use" well, but the comparison and price axes often sit empty. The more sensitive a question feels — the kind you'd rather avoid — the more your customer asks AI instead. If you don't answer it, someone else will.

Once you have 30

That list is your content roadmap. Measure how many of them your pages answer (self-check) and you get your priorities too. For reference, Bulti's free audit automates this whole process — feed it your product keywords and it builds 30 category-tailored questions and measures your coverage.

FAQ

Q. Do I really need 30 questions? Can't I start with 5 core ones?
You can start with 5. But customer situations vary, so the long tail of questions runs deep. Around 30, you'll always uncover a question you didn't know about — and that's where the competition is empty.

Q. Don't questions change with the season or trends?
They do. So a question list isn't a one-and-done — it's best to re-skim autocomplete each quarter and refresh it.

Q. I gathered the questions but don't have time to write.
You don't have to write them all. Start with the axes closest to a buying decision (comparison, selection criteria), one post at a time. One question per post is the basic unit.

Originally published at en.bultibulti.com. We run a free GEO audit there — it checks one of your pages against 30 buyer questions, no signup.

What kind of content does AI quote? Start with how the answer gets built

Bulti — Wed, 19 Aug 2026 01:01:07 +0000

⏰ The 30-second version

An AI answer is built in three stages — retrieve (find), extract (cut), synthesize (weave).
To survive stage one you need a text page that's retrievable; stage two pulls paragraphs that make sense when cut out; stage three pulls sentences that are the answer itself.
Keyword repetition helps at none of the three stages. A structure that answers the question passes all three.
Once you understand the principle, you don't have to memorize the tactics.

Stage 1 — Retrieve: no retrieval, no start

When a customer asks "When should I take a probiotic?", the AI first searches for relevant documents and gathers candidates. The first gate opens here. Image-only detail pages and login-gated pages don't make the candidate pool. The text has to physically exist in the HTML just to make the starting line.

This stage overlaps with traditional search visibility — one reason SEO doesn't become completely pointless. Being retrievable is still the ticket in.

Stage 2 — Extract: AI takes paragraphs, not articles

From the candidate documents, AI extracts the parts relevant to the question — by paragraph, not by whole article. Here's the second gate: does an extracted paragraph make sense on its own?

"As mentioned above, that product is good for this" becomes meaningless the moment it's cut out. "The ○○ product uses a low-temperature process, so the texture stays tender" is complete information wherever you cut it. Same content, different fate.

What makes a paragraph survive extraction

[ ] Product names and specifics instead of pronouns
[ ] One paragraph = one claim
[ ] Complete even read without context
[ ] Numbers and criteria contained inside the paragraph

Stage 3 — Synthesize: sentences that are the answer get pulled

When AI assembles an answer from the extracted pieces, what it pulls most often is the sentence that answers the question head-on. Something with a clear question-answer pair, like "When is the best time to take a probiotic? After a meal is generally recommended, rather than on an empty stomach."

Content that opens with an intro and hides the conclusion at the very end loses out at this stage. That's why the GEO rule — put the answer in the first paragraph (details) — is a principle, not a trick.

The conclusion the principle points to

Look at the three stages again and the common thread appears — AI isn't looking for "well-written prose," it's looking for good raw material. Material that's retrievable, material that can be cut, material that's the answer itself. Reread your content through the eyes of raw material. Which paragraphs survive being cut out? You can also check all of that across 30 questions at once with the free audit.

FAQ

Q. Doesn't every AI work differently?
The implementation details differ, but the "retrieve–extract–synthesize" skeleton is shared. That's why one formula responds to multiple engines.

Q. Are backlinks and domain scores meaningless now?
They still influence stage one (retrieval). But stages two and three are a contest of content structure, so even a weak domain has room to win with better structure.

Q. Is longer content always better?
Density beats length. Since extraction is paragraph-level, several complete paragraphs beat one long article. Even 800 words is plenty to get cited if it answers the question precisely.

Originally published at en.bultibulti.com. We run a free GEO audit there — it checks one of your pages against 30 buyer questions, no signup.

ChatGPT vs Perplexity vs Google AI Overviews — where do your customers ask?

Bulti — Wed, 19 Aug 2026 01:00:42 +0000

⏰ The 30-second version

Customers meet AI search in roughly three places — ChatGPT (conversation), Perplexity (sources), and Google AI Overviews (top of search).
All three build answers from web content, but they differ in how they show sources. Perplexity and Google expose source links, so you earn clicks; on ChatGPT, the brand mention itself is the win.
Prioritize by where your customers already are — for e-commerce, start with Google AI Overviews + ChatGPT.
Good news: all three reward the same kind of content, so write it well once and you cover all three surfaces at once.

Three surfaces, three personalities

	ChatGPT	Perplexity	Google AI Overviews
How you meet it	Customer chats directly	Customer searches directly	Appears automatically when you Google
Answer format	Conversational recommendation	Answer + source cards	Search summary + sources
The seller's win	Brand mention	Cited source → click	Cited source → click
Monthly search vol. (tool name)	—	368,000	320 (+27%)

ChatGPT is a surface where the mention is everything. The effect is less about link clicks and more about the customer learning your brand. Perplexity is the reverse — sources are the star, so getting cited drives real traffic.

Google AI Overviews operates at a different scale. The customer doesn't go looking for AI — the Google they always use pushes AI at them first. For e-commerce sellers, it's the surface where the impact hits fastest.

Where to respond first: start where the customers are

Try to cover everything and you'll never start. Here's an order:

Google AI Overviews — the surface most customers meet automatically. Detailed approach
ChatGPT — the first port of call for younger and research-minded customers.
Perplexity — volume is growing, and the source-click upside is concrete.

And here's the relief: all three engines reward content that "answers the question directly, can be cut into pieces, and exists as text." You don't have to write different content per surface — one GEO formula covers all three at once.

FAQ

Q. What about Google's AI Overview specifically?
That's exactly what "Google AI Overviews" refers to here — the AI summary Google shows at the top of search. It appears often for recommendation-style queries in e-commerce, and the same principles apply, so the formula above carries straight over.

Q. If my brand's mentions differ by engine, which do I optimize for?
The difference itself is information. If you show up in only one place, check what the source content is there, and extend that format to your other questions. That's the efficient move.

Q. Do I have to check all three by hand every time?
The principle is covered by the self-check, but repeating dozens of questions × three surfaces is real work, so a tool is the realistic option.

Originally published at en.bultibulti.com. We run a free GEO audit there — it checks one of your pages against 30 buyer questions, no signup.

What are Google AI Overviews, and why did my blog traffic drop?

Bulti — Tue, 18 Aug 2026 01:00:55 +0000

⏰ The 30-second version

Google AI Overviews is the feature that summarizes the answer for you, right at the top of the results. Customers get their answer before they ever click a link.
That's why link clicks are falling. It's the same directional change as "blog top rankings" searches dropping -39% in three months.
The way out isn't turning the Overview off — it's becoming one of its sources. Tellingly, "cited in Google AI Overviews" has started appearing in autocomplete.
The conditions for getting cited are the same as the GEO formula — content that answers the question directly, in cut-and-paste-able paragraphs.

What the AI Overview does: it answers before the click

When a customer searches "recommend a moisturizer," they used to see a list of ads and blog links. Now, above all of that, an AI Overview appears first — pulling the results together into something like, "The most frequently mentioned brands are ○○ and ○○."

For the customer, it's a nice feature. The problem is that a customer satisfied by that summary clicks the links below far less. The harder you've worked to rank a blog at the top, the more you feel it.

What the numbers show

"Google AI Overviews" searches: 320/month, +27% in three months — curiosity is climbing fast.
"Blog top rankings" searches: 390/month, -39% — demand for learning ranking techniques has cracked.
Autocomplete now lines up "Google AI Overviews cited," "how to opt out," and "how to set up" side by side — some people are already hunting for how to get cited, and others for how to avoid it.

The same change, and people split on which side of it to stand: the side losing traffic, or the side becoming the source of the answer.

How to become a source: it's the same formula in the end

An AI Overview builds its summary from web documents too. The conditions for being chosen as material are no different from ChatGPT's.

A first paragraph that answers directly — in a form the summarizer can lift as-is.
Self-contained paragraphs — units that still make sense when cut out.
Concrete criteria — not "it's nice," but "for oily skin, check for the ○○ ingredient."
A page that exists as text — image-based detail pages can't be raw material.

The detailed execution is in the 5 GEO moves. How many questions your pages currently answer, you can measure with the free audit.

FAQ

Q. Does the AI Overview appear for every search?
No. It mainly shows up for question-style searches with something to summarize — comparisons, recommendations, how-tos. And those are exactly the searches that decide a purchase, which is why the impact on sellers is large.

Q. How do I know if my content got cited in an Overview?
The basic method is to search the term yourself and check the Overview's source links. To monitor many questions regularly, a tool is the realistic option.

Q. Should I give up on blogging because of the Overview?
The opposite. Because what the Overview uses as material is, in the end, text documents like blog posts, the value of a well-written blog actually goes up. What changed is the standard — from "ranking content" to "citable content."

Originally published at en.bultibulti.com. We run a free GEO audit there — it checks one of your pages against 30 buyer questions, no signup.

Your customers are asking ChatGPT instead of Google now — what changed?

Bulti — Tue, 18 Aug 2026 01:00:31 +0000

⏰ The 30-second version

Customer search is shifting from keywords to sentences. Instead of "best probiotics," it's "Recommend a probiotic — which one should I pick?"
Keyword search hands back ten links; AI hands back one organized answer. With the work of comparing gone, a brand that isn't in the answer never gets a chance to be seen.
Perplexity draws 368,000 searches a month — AI search is no longer a niche hobby.
Sellers have one thing to prepare: content that answers sentence-form questions.

From keywords to sentences: the way people ask changed first

In the search-box era, customers trimmed their words to fit the machine: "chicken breast tender recommend." No human talks like that to another human. In front of AI, natural speech returns: "Recommend a chicken breast that isn't dry and chalky. I started working out about a month ago."

Why does this difference matter? Because a sentence carries information a keyword never did — the situation (a month into working out), the complaint (dryness), the requirement (a tender texture). AI assembles an answer that fits that context, and to do it, it looks for content that answered the context. That's exactly where content written to chase keywords alone falls behind.

From ten links to one answer: the way people look changed too

The result of a keyword search was homework — open ten links, filter the ads, compare reviews. The result of an AI search is an answer key: two or three brands, laid out with reasons.

Easier for the customer, and flipped for the seller. It used to be that landing on page two of the results meant a diligent customer could still find you. Now, if you're not on the answer key, that's it. Ten slots became two or three — that's the whole story, and the whole point, of this shift.

The numbers: the change is already here

Perplexity: 368,000 searches/month — the tool's name now out-searches most product keywords.
"AI search": 5,400/month; "ChatGPT search": 110/month, +4% — demand to learn the new way of searching itself.
"Google AI Overviews": 320/month, +27% — a change already on the front page of Google.

(Google Keyword Planner data, retrieved July 2026)

What sellers should do now

Nothing grand. Start by writing down, in full sentences, what your customers ask AI. Those sentences are already sitting in your reviews and inquiries. How to check whether your pages answer that list is in the 5-minute self-check, and how to write content that answers it is in the 5 GEO moves.

FAQ

Q. My customers skew older and don't use AI — does this apply?
Google AI Overviews is the variable. Even if a customer never opens a separate AI tool, an AI answer now appears at the top of the Google search they already use. We've gone from an era of "going to use AI" to an era of "running into AI."

Q. Is keyword search going away?
No — it'll coexist for a while. But the more a question decides a purchase (comparisons, recommendations), the sooner it moves to AI. In other words, the searches that matter most to sellers change first.

Q. What's different about content that answers sentence-form questions?
It's designed around question coverage, not keyword density. Use customer questions as subheads, answer directly in the first paragraph, and address situation-specific conditions in concrete detail.

Originally published at en.bultibulti.com. We run a free GEO audit there — it checks one of your pages against 30 buyer questions, no signup.