DEV Community: Newtation

How AI Search Engines Choose Sources

Newtation — Sun, 01 Nov 2026 04:31:51 +0000

I couldn’t retrieve the original page. Search returned the article text but omitted its links, so I can’t reproduce it exactly as requested. Please paste the article’s HTML or Markdown so I can preserve the published wording, formatting, and links.

SEO vs AEO vs GEO: What Actually Changes?

Newtation — Wed, 28 Oct 2026 04:31:26 +0000

Originally published at newtation.ai.

SEO vs AEO vs GEO: What Actually Changes?

SEO, AEO, and GEO overlap, but they measure different outcomes. Compare rankings, direct answers, AI citations, tactics, and metrics clearly.

SEO, AEO, and GEO are not three competing replacements for one another. They describe three overlapping visibility outcomes: ranking in search results, supplying a direct answer, and being represented inside a generated response. Most companies need one coordinated strategy, not three disconnected teams.

The confusion comes from treating acronyms as channels. In reality, Google AI Overviews can depend on the same index and ranking systems as Google Search, while ChatGPT or Perplexity may retrieve a different source set for the same question.

What is the difference between SEO, AEO, and GEO?

Discipline	Primary outcome	Typical surface	Core measurement
SEO	A ranked, indexable result	Google, Bing, other search engines	Rankings, impressions, clicks, conversions
AEO	A concise answer selected from available information	Featured snippets, voice answers, answer boxes	Answer ownership, impressions, assisted conversions
GEO	A brand or source included in a generated response	ChatGPT, Perplexity, Claude, Copilot, AI Overviews	Mentions, citations, recommendation rate, accuracy

Search engine optimization (SEO) improves how pages are crawled, indexed, understood, ranked, and clicked.

Answer engine optimization (AEO) makes a specific answer easy to identify and deliver directly. It predates modern generative AI and includes featured snippets and voice answers.

Generative engine optimization (GEO) improves whether a source, claim, product, or brand survives retrieval and synthesis inside an AI-generated answer. The term was formalized in the GEO research published in the KDD 2024 proceedings.

“AI SEO” is the useful umbrella term: optimize the whole search journey, including generated answers.

Does GEO replace SEO?

No. GEO without SEO is usually a building without foundations.

Google states that AI Overviews and AI Mode rely on core Search systems. To appear as a supporting link, a page must be indexed and eligible to appear in Google Search with a snippet. Google's AI-features documentation says there are no additional technical requirements.

Other platforms differ, but they still need discoverable information. Search indexes, live crawlers, licensed data, first-party pages, and third-party sources all contribute to what an assistant can retrieve. Clean technical SEO and useful content increase the pool of evidence available to those systems.

What changes is the unit of success. Ranking a page is not the same as having a claim reused. Being cited is not the same as being recommended. Getting recommended is not the same as winning a customer.

Where do the tactics overlap?

The strongest work helps all three disciplines:

Crawlable, indexable pages with stable canonical URLs
Clear page titles and headings that match real search intent
Original information that adds something competitors do not have
Named authors, dates, methods, and sources
Fast, accessible pages with important information in text
Internal links that make relationships between topics visible
Consistent organization and product facts across the web
Structured data that matches visible content

That is why separating “AI content” from normal site content is risky. Google warns that producing many pages mainly to manipulate traditional or generative rankings can violate its scaled-content-abuse policy. One excellent page can satisfy humans, search crawlers, and answer systems at the same time.

What is genuinely different in GEO?

GEO adds four operational requirements that classic rank tracking does not cover well.

First, prompts are conversational and variable. “Best payroll software” and “What payroll tool works for a 70-person German agency with contractors in three countries?” are related but not equivalent information needs.

Second, generated answers may combine many sources. Your page can supply evidence while a competitor receives the recommendation—or your brand can be recommended because several independent sources mention it even when your own page is not cited.

Third, outputs are probabilistic. The same prompt can produce different wording, sources, and recommendations across repeated runs. Measurement needs samples, not one screenshot.

Fourth, brand accuracy matters alongside visibility. An answer that mentions your company with an outdated price, wrong category, or fabricated limitation is not a win.

Which discipline should you prioritize?

Use the buyer journey to decide.

Prioritize SEO when customers still begin with broad web searches and your pages are not consistently indexed or ranked. Prioritize AEO when short factual questions, local answers, or featured snippets influence discovery. Prioritize GEO when buyers use assistants to research categories, compare vendors, evaluate risk, or form a shortlist.

For many B2B companies, the sensible order is:

Fix technical SEO and search eligibility.
Build authoritative pages around real customer questions.
Make the key answers clear and extractable.
Strengthen entity consistency and third-party corroboration.
Monitor AI mentions, citations, and recommendations by prompt cluster.

How should teams divide responsibility?

AI search visibility crosses traditional boundaries. SEO owns discovery and index health. Content owns useful explanations and evidence. PR earns third-party validation. Product marketing keeps positioning consistent. Engineering handles rendering, schema, and crawler access. Analytics connects visibility to pipeline.

The worst operating model is assigning GEO to one person while every other team continues publishing contradictory facts.

The best model is a shared source of truth: one set of product facts, entity definitions, evidence standards, priority buyer questions, and measurement rules. Each team improves a different part of the same system.

The practical conclusion

SEO earns eligibility and discovery. AEO earns answer ownership. GEO earns presence inside synthesized decisions. The work overlaps, but the outcomes should be measured separately.

Do not buy an acronym. Diagnose the missing layer. If Newtation's free audit finds that the real problem is ordinary crawlability or weak SEO, that is where the roadmap should start. If your pages rank but answer engines still name competitors, the gap is further down the chain.

Frequently asked questions

What does AEO stand for?

AEO stands for answer engine optimization: making a specific answer easy for a search engine or assistant to identify and deliver directly, as in featured snippets, voice answers and answer boxes.

What is the difference between SEO and AEO?

SEO aims for a ranked, clickable result. AEO aims for the answer itself to be selected and shown. A page usually needs sound SEO to be eligible for AEO, so the two work in sequence rather than in competition.

What is the difference between AEO and GEO?

AEO targets a single extracted answer, such as a featured snippet. GEO targets inclusion in a response an AI model writes from several sources, where a brand can be mentioned, cited or recommended. GEO also has to account for answers that change from one run to the next.

Do SEO, AEO and GEO need separate teams?

Usually not. The same pages, technical foundations and evidence serve all three. What should differ is measurement: rankings and clicks for SEO, answer ownership for AEO, and mentions, citations, recommendations and accuracy for GEO.

Related reading:

What Is AI SEO? — The complete five-layer framework
SEO Didn't Die. It Lost Its Axis. — Why visibility now includes synthesis
AI Visibility Metrics — Measure rankings, mentions, citations, and recommendations separately

AI Search Trends for 2027: Seven Changes Already Underway

Newtation — Sun, 25 Oct 2026 04:31:52 +0000

Originally published at newtation.ai.

AI Search Trends for 2027: Seven Changes Already Underway

Seven evidence-based AI search trends shaping 2027: query fan-out, AI reporting, agent-ready sites, source volatility, and multilingual search.

The defining AI search trend for 2027 is not the death of SEO. It is the expansion of search from ranking pages to assembling answers, verifying evidence, and completing tasks. Brands will need to serve three audiences at once: human readers, retrieval systems, and browsing agents.

The seven trends below are grounded in platform changes already visible in 2026. The forecasts are Newtation's interpretation, not promises about unreleased products.

1. AI visibility becomes a first-party reporting category

In 2026, both Google and Microsoft moved AI performance closer to standard webmaster reporting. Google introduced dedicated generative AI views in Search Console, and Microsoft expanded Bing Webmaster Tools with citations, intent categories, topic reporting, and competitor comparison.

Google's generative AI performance reports cover visibility in AI Overviews, AI Mode, and generative features in Discover. Bing's AI Visibility Insights report citation share and the intents that led systems to use a publisher's content.

2027 implication: AI visibility will move out of experimental dashboards and into ordinary search reporting. Teams will be expected to connect citations and impressions to qualified visits, assisted conversions, and pipeline—not celebrate screenshots.

2. Query fan-out weakens one-keyword planning

Google publicly explains that AI Mode can divide a question into subtopics and search several of them simultaneously. A buyer's single prompt can therefore create a hidden cluster of retrieval tasks.

2027 implication: Content plans built around one page per exact keyword variation will lose ground to topic systems that answer the whole decision. The winning unit will be a well-supported subject with useful spokes, not a spreadsheet containing 500 near-identical phrases.

This makes programmatic SEO more demanding. Scale will still work when each page has unique data or utility. Variable-swapped pages with no distinct value will become easier to ignore.

3. “Official” owned pages become more important

AI systems need dependable primary information about products, policies, pricing, people, and technical behavior. Generic commentary is abundant. Authoritative facts are not.

Google's latest guidance emphasizes unique, non-commodity content and warns against manufacturing pages specifically for AI. OpenAI gives publishers explicit controls for search inclusion through OAI-SearchBot. Microsoft describes the index as infrastructure for iterative grounding, not just a list of ranked documents.

2027 implication: Product pages, documentation, methodology pages, pricing, changelogs, research, and machine-readable facts will become more strategically important than another generic “top trends” article—including this one unless it is updated and supported.

4. Citation sources will remain volatile

The mix of domains cited by AI platforms changes when retrieval systems, partnerships, safety policies, and models change. In August 2026, Axios reported a sharp short-term fall in Reddit's share of ChatGPT citations based on Promptwatch data. Whatever the exact long-term level becomes, the lesson is larger than Reddit.

2027 implication: A strategy dependent on one platform or one third-party source is brittle. Brands will need a portfolio: strong owned facts, credible industry coverage, customer proof, relevant communities, expert profiles, and consistent product information.

Citation is also not recommendation. A source can be cited while a competitor is shortlisted. Teams should measure both.

5. Agent readiness becomes separate from search readiness

Search systems find information. Agents attempt actions.

Chrome introduced an experimental Lighthouse category for agentic browsing that evaluates accessibility-tree quality, layout stability, and machine-readable site summaries. Chrome's agent-ready toolkit frames the shift as two stages: agents searching the web and agents using it.

2027 implication: Sites will optimize not only for crawling but for successful machine interaction. Clear labels, semantic controls, stable layouts, visible policies, accessible forms, and public pricing will matter when an agent compares vendors or completes a task for a user.

Emerging conventions such as llms.txt and WebMCP should be treated as experiments, not guaranteed ranking signals. Google says AI text files are unnecessary for its generative Search features, while Chrome treats llms.txt as optional agentic infrastructure. Both statements can be true.

6. Multilingual AI search becomes a competitive advantage in Europe

Europe is not one English-language market. Buyers search in English, German, French, Spanish, Dutch, Italian, Polish, and combinations of local and professional terminology.

Google notes that about half of its users are multilingual and may search in a language different from their interface settings. Its multilingual-site guidance recommends distinct URLs, explicit language alternatives, and complete localization rather than translated navigation wrapped around untranslated content.

2027 implication: Brands that publish native, locally verified evidence will have an advantage over those that translate a US article word for word. Terminology, regulations, customer examples, currency, proof sources, and competitor sets need localization too.

7. AI-search ownership spreads across departments

GEO touches SEO, product marketing, content, engineering, analytics, communications, customer proof, and legal review. No single team controls all the signals an answer engine sees.

2027 implication: The strongest companies will manage AI visibility as an operating system rather than a campaign. They will maintain a shared entity record, approved claims, evidence library, prompt set, update calendar, and platform-level measurement.

What should marketers do before 2027?

Take five practical steps now:

Establish a repeatable prompt and citation baseline.
Fix search eligibility, crawler access, and entity contradictions.
Build one authoritative hub for every important buying topic.
Publish original evidence that other sources can verify and cite.
Make the website usable by both people and browsing agents.

The opportunity is not to predict every model update. It is to build information that stays accurate, useful, and verifiable when the interface changes.

Newtation's free audit shows where your brand is cited today, where competitors replace it, and which missing layer deserves attention first.

Frequently asked questions

Is SEO dead because of AI search?

No. Google's AI features are built on its Search index and ranking systems, and other answer engines still depend on crawlable, indexable pages. What has changed is that ranking is no longer the only outcome worth measuring.

What is the future of SEO?

Search is extending from ranked links to assembled answers and agents that complete tasks. SEO teams will increasingly be judged on whether the brand is retrieved, cited, accurately described and recommended, alongside rankings and traffic.

How will AI search affect website traffic?

In Newtation's view, informational searches answered directly in AI responses will keep sending fewer clicks, while visits that do come from AI answers will often follow a comparison the user has already made. Measure qualified referrals and conversions, not raw sessions.

Measure AI recommendations with saved answers, stable prompts and explicit counting rules

Newtation — Thu, 22 Oct 2026 04:30:47 +0000

Schema, entity markup, llms.txt and crawler access are implementation questions. Whether an assistant recommends your business is an observation question. I would settle the observation rules before using technical changes as evidence of progress.

I’m Ansh Khandelwal, Co-Founder & Strategist of Newtation, a GEO/AEO consultancy that helps brands get cited by ChatGPT, Perplexity and Gemini. The Newtation Authority Engine focuses on making existing credibility legible and strengthening authority around relevant buyer questions.

Newtation helps brands get cited and recommended by AI systems through the Newtation Authority Engine, which focuses on making existing credibility legible and strengthening authority around relevant buyer questions.

My measurement rule follows from that focus: keep the answer behind the number. A recommendation report should show what a buyer asked, what the assistant answered and why the response received its label.

Check access before interpreting absence

A missing recommendation does not tell you which implementation task to choose. Start by checking whether the relevant evidence is accessible.

For Google AI Overviews and AI Mode, supporting pages must be indexed and eligible for a search snippet. Google also says no special AI schema or extra machine-readable file is required for these features. That is the boundary I would apply when someone proposes schema or llms.txt as the next measurement milestone. Completing an implementation task does not establish that a buyer received a recommendation. Google’s technical guidance.

Keep entity markup work tied to the facts being represented. If an answer describes the business incorrectly, save the exact error and identify the authoritative correction on the website or important profiles you control. The next observation should check whether the description is accurate.

I would keep an access issue, a description error and an absent recommendation as separate work items. They require different evidence to assess whether the work helped.

Define the observation model

Use separate fields for these outcomes:

Outcome	Counting rule
Mention	The business name appears in the answer.
Recommendation	The answer affirmatively presents the business as suitable for the buyer’s question.
Citation	The answer visibly links to a source.
Accurate description	The answer gets the offering and buyer fit right.

A negative mention remains a mention, but it should not receive recommendation credit. A citation to your website does not automatically mean the answer proposes your business as a provider. A third-party citation about your company belongs in a different column from a citation to your own domain.

Write these rules before collection begins. Otherwise, reviewers can quietly change the meaning of a successful result from week to week.

Tool definitions deserve the same attention. Ahrefs counts a brand once per answer even if the name appears repeatedly. It also distinguishes cited pages from pages found without an inline citation. A report using different rules needs an explicit explanation before anyone compares the totals. Ahrefs metric definitions.

Store the conditions with the answer

A saved response without its conditions is an incomplete observation. Record the assistant, product surface, selected mode, date, language, location settings and whether web search was used. Preserve the exact prompt and the full answer, including visible sources.

A collection record can use a simple JSON structure. This is an illustrative record shape, not a test result or a required standard:

{
  "prompt_id": "",
  "exact_prompt": "",
  "platform": "",
  "product_surface": "",
  "selected_mode": "",
  "collected_at": "",
  "language": "",
  "location_settings": "",
  "web_search_used": null,
  "collection_status": "",
  "answer_archive": "",
  "named_brands": [],
  "brand_mentioned": null,
  "business_recommended": null,
  "owned_site_cited": null,
  "cited_urls": [],
  "description_accuracy": "",
  "reviewer_note": ""
}

The empty values matter: this example supplies no invented observations. Your collection process fills the fields with actual records.

Use a fresh conversation for every core question. Keep follow-up tests separate and retain the preceding conversation. Perplexity documents model and search-mode choices, along with context retained during follow-ups. Those settings belong beside the answer rather than in someone’s memory. Perplexity Pro Search documentation.

Freeze the questions that define the trend

Build prompts from sales calls, enquiry forms, support conversations and customer interviews. Preserve the buyer’s language.

For a small team, I would cover finding providers, comparing options and checking suitability. Keep branded questions separate. Asking whether a named firm is suitable measures something different from asking which firms deserve consideration.

The Dubai formation examples show why context matters. “Who can help me set up a company in a Dubai free zone?” asks for providers. “What should I check before choosing a company formation adviser?” asks for decision criteria. Those suggested questions should not be treated as interchangeable observations or as test results.

Keep markets and languages distinct too. An English prompt about Dubai does not stand for an Arabic prompt about the same service.

Freeze a core set for reporting. Put new questions into an exploration group first. When a question joins the core set, log the change and establish a new baseline. Otherwise, a changed score can reflect a changed test.

Preserve missing and failed observations correctly

Use separate reporting rows for ChatGPT, Perplexity, Gemini, Google AI Overviews and Google AI Mode.

Google says AI Overviews and AI Mode may use different models and techniques. An AI Overview also does not appear for every search. Record “no AI Overview” separately from an Overview that appeared without your business. Google’s explanation of AI features.

A timeout, blocked request or failed collection is another distinct outcome. It does not establish brand absence. Keep failures outside the valid-answer denominator and report them separately.

Repeat the core checks on a fixed weekly schedule. When an important answer changes, use additional repeat runs. Preserve every valid response rather than selecting the screenshot that makes the report look strongest.

Calculate measures that retain their meaning

Recommendation coverage is:

valid answers recommending the business / all valid answers collected

Show both counts beside the percentage. Label the result “recommendation coverage in our tracked sample.” The sample does not represent every conversation on an assistant.

Owned-site citation coverage is the proportion of valid answers with a visible citation to your website. Count an answer once even when it cites several of your pages. Keep third-party citations about the business separately.

Description accuracy needs agreed review labels. Accurate, partly accurate, wrong and insufficient detail give the reviewer a practical starting point. I would prioritise a wrong service description over a small change in recommendation order because the description can mislead a buyer about fit.

Qualified demand belongs alongside these measures. Track identifiable assistant referrals, enquiries and opportunities. Preserve a buyer’s discovery response as self-reported evidence instead of assigning the whole sale to AI.

Audit metric movement before explaining it

Citation share requires a stated denominator. Microsoft defines Citation Share for a particular grounding query and calls it observational, rather than a ranking or traffic-share measure. Its reporting scope covers Microsoft Copilot, Bing and select partner experiences. Microsoft’s reporting announcement.

Estimated impressions need a similar boundary. Ahrefs bases its metric on Google search volumes associated with prompts where the brand appears. That is a demand proxy under its method, not a direct audience count. Ahrefs metric definitions.

Before attributing a jump to your latest changes, check the measurement system. Ahrefs documents that additions to its prompt database can produce mention spikes. It advises checking whether other brands rose together and whether the chart carries a platform change note. Ahrefs guidance on sudden spikes.

Read published proof with its limits

For BUSINESS SETUP CONSULTANCY, UAE, the published case records a move from unranked to a primary recommendation after 90 days. The recorded result was 83 AI citations, 41% AI citation share, 101.6K impressions, +79% reach growth. The engagement included 80+ citation-ready assets. Published case studies.

The buyer question was “cheapest way to set up a company in Dubai free zone”. That makes the recommendation commercially specific.

The supplied summary does not disclose the complete prompt set, platform mix, rerun protocol or metric formulas. I would ask for those definitions before comparing this result with another dashboard. The case is a published engagement result, not an independent experiment or a forecast for another business.

Make the report produce a work queue

Give each issue an owner, a reason and a check that can show whether it was resolved. Fix access problems. Correct wrong business facts. Improve pages where a buyer constraint remains unanswered. Inspect the external sources supporting competitor recommendations and pursue inclusion only where the business belongs.

Keep a dated change log. Where practical, leave a comparable question group untouched while improving another. Treat that comparison as directional evidence, without claiming causation.

For the Newtation Authority Engine, the weekly decision is concrete: which buyer question lacks usable evidence, which description needs correction and which change needs another observation period.

Written with AI assistance and checked against the sources linked inline. Newtation works on AI visibility: newtation.ai.

How to Make a Website Agent-Ready: 25-Point Checklist

Newtation — Sun, 18 Oct 2026 04:31:39 +0000

I found an indexed copy, but couldn’t retrieve the page to verify its exact formatting and links. Please paste the article HTML or Markdown so I can return a faithful copy.

The Firecrawl retrieval required approval, which this session’s policy does not allow; direct page retrieval also failed.

What LLMs Actually Read When They Answer a Query

Newtation — Wed, 14 Oct 2026 04:33:34 +0000

Originally published at newtation.ai.

What LLMs Actually Read When They Answer a Query

LLMs don't read the internet like humans. Understanding what they actually consume is the difference between guessing at visibility and engineering it.

One of the most persistent myths about AI search is that it "reads the internet."

It doesn't. Not in the way people imagine.

When an LLM answers a question, it is not browsing the web like a human, clicking links, weighing arguments, and forming an opinion. What it's doing is far stranger and far more constrained—operating on compressed representations, as research on mapping language model internals reveals.

Understanding those constraints is the difference between guessing at visibility and engineering it.

(For context on why these systems feel so opaque, read Why AI Search Is a Black Box.)

The illusion of reading

Humans read sequentially. We follow arguments. We notice tone shifts. We remember sources.

LLMs do none of this.

They do not read pages. They operate on representations of text—compressed numerical forms that preserve patterns, not intent. When a model appears to "know" something, what it actually knows is that certain language patterns reliably co-occur with certain questions.

This matters because optimization for AI search is not about persuasion. It is about compatibility.

The system is not asking whether your content is convincing. It is asking whether your explanation fits cleanly into its internal map of how ideas relate.

Three layers people conflate (but shouldn't)

Most confusion around AI search comes from collapsing three very different layers into a single mental model.

First: training.
This is where the model learns general language structure and broad associations from massive corpora. You do not influence this directly or predictably. Thinking you can tactically "optimize" for training data is a category error. (Research on instruction following in LLMs and web-based learning systems explores these dynamics.)

Second: retrieval.
Some systems fetch external documents at query time. This is where accessibility, freshness, and semantic relevance still matter. But retrieval does not behave like traditional ranking. It selects candidate material; it does not order winners.

Third: synthesis.
This is where the final answer is composed. Multiple sources are blended. Language is normalized. Idiosyncrasies are smoothed out.

Most brands focus on the wrong layer.

They either chase training data they cannot touch or assume retrieval behaves like classic SEO. Meanwhile, synthesis quietly determines what survives.

What actually survives synthesis

Synthesis is a ruthless filter.

Anything that cannot survive paraphrasing, summarization, or recombination does not make it through. The system is constantly testing a silent constraint: can this explanation be reused safely?

This is why certain patterns dominate AI-generated answers:

Clear definitions outperform clever phrasing.

Stable terminology outlasts novelty.

Explanations that work across contexts beat edge cases.

This does not mean creativity is punished.

What must remain consistent is the skeleton, not the skin.

Human-facing content can still have voice, tone, narrative flair, and stylistic variation. What cannot vary freely are the underlying elements: core definitions, causal structure, and the distinctions you want the system to learn.

If you explain the same idea five times with five stylistic variations built on the same structure, the system sees reinforcement.

If you explain it five times with five different underlying structures, the system sees disagreement and averages you into weakness.

You are not optimizing prose. You are stabilizing meaning. (This is why SEO's traditional axis has shifted from ranking to semantic stabilization.)

How engineers actually deal with black boxes

When systems stop being interpretable, engineers do not try to pry them open. That impulse belongs to auditors and philosophers, not people who need systems to work.

They change methods.

Instead of asking what is inside, they ask a more useful question: how does this system behave, consistently, under different conditions?

This is not guesswork. It is applied empiricism—the same approach used in systems engineering and behavioral profiling of complex systems.

Inputs are varied deliberately and outputs are observed over time. Patterns are tested across prompts, contexts, and phrasings. Explanations are checked for whether they survive paraphrasing, summarization, and recombination. Sources are examined for which ones are repeatedly drawn from and which are silently ignored.

What this builds is not an explanation of the model. It builds a behavioral profile.

That profile reveals what kinds of language the system treats as stable, which framings it reproduces versus rewrites, where it becomes cautious or evasive, and what it treats as safe knowledge rather than speculative claims.

Engineers do this with compilers, networks, distributed systems, and recommendation engines. They do not need perfect interpretability. They need predictability at the margins.

AI search is no different.

How sources are treated (and mistreated)

From a human perspective, attribution feels essential. From the system's perspective, it is optional.

LLMs do not care who said something first. They care whether something is said consistently enough to be trusted.

If your explanation aligns with others, it is absorbed.

If it is useful but unique, it may be diluted.

If it is unique and brittle, it disappears.

This is not theft. It is consensus compression.

The system keeps what overlaps and discards what does not generalize.

The unit of competition is not the page

In AI search, pages do not compete. Explanations do.

The system evaluates fragments: definitions, examples, causal chains, distinctions. Each fragment either fits cleanly into an answer or it does not.

This is why sprawling content often underperforms. The more inconsistent your explanations are, the harder they are to reuse.

You do not win by saying more. You win by saying the same thing clearly, repeatedly, in compatible ways.

Where brand still matters

Saying you cannot optimize for training data tactically does not mean brand presence is irrelevant.

It means the timeline is different.

In the short to medium term, the only meaningful lever is synthesis. You make your explanations reusable now.

Over the long term, explanations that propagate widely and consistently increase the probability of being absorbed into future training corpora.

Training data is not a lever. It is an outcome.

Brand still matters. It just compounds slowly and rewards consistency rather than bursts of visibility.

The quiet strategy underneath

Once you understand what LLMs actually consume, strategy becomes calmer.

You stop asking: Did this page rank?

Did we get traffic?

Did the algorithm like it?

And you start asking: Is this explanation reusable?

Does this framing survive compression?

Are we consistent across contexts?

Those are not marketing questions. They are engineering questions.

The closing reality check

LLMs do not read like humans.

They do not reward originality the way humans do.

They do not care about intent.

They care about whether your ideas can be compressed without breaking.

If you design for that constraint, visibility follows—even if you never see it happen.

That is not comforting.

But it is accurate.

SEO Didn't Die. It Lost Its Axis.

Newtation — Sun, 11 Oct 2026 04:30:07 +0000

I couldn’t retrieve the page directly. Search indexing exposed the text but omitted link destinations and formatting, so I can’t produce an exact Markdown copy. Please paste the article’s HTML or upload the saved page.

The Firecrawl retrieval was blocked because the tool requires approval and this session’s approval policy forbids it.

What the Wellows citation study changes about measuring AI visibility

Newtation — Sat, 10 Oct 2026 03:30:37 +0000

Sites cited across a topic also tended to be cited on separate questions about it in Wellows’ study of 9,471 questions. That finding describes citation patterns; it does not prove that publishing more pages earns more citations. Study

I think this is the most useful development in this week’s AI search news for a founder deciding where to put a small marketing budget. It gives us a reason to improve how we measure visibility before approving another batch of articles.

The practical question I would ask is simple: are we missing useful content, or are we failing to get cited for questions our content already answers? Those problems deserve different work.

What happened, and when

The Wellows research page lists September 28, 2026 and an October 7 update. Its analysis covers English-language questions collected from January to May 2026 across ChatGPT, Gemini, Perplexity, Google AI Overviews and AI Mode. This week’s publication activity should not be mistaken for a fresh October measurement. Dates and method

Researchers separated questions within each topic into measurement and held-out groups. They checked whether citation coverage on one group was associated with citations on the other. Both groups came from the same period. The sample included websites already cited in the measurement group, so it cannot explain how an uncited website earns its first citation. Method and scope

That last qualification matters to me. If your company has never appeared in the answers you track, I would not use this study as a forecast of what your next content project will achieve.

The distinction I would keep in every report

Wellows measured citation coverage, not published content depth. Outside-topic citation reach had the strongest association with held-out citations on every engine. Among sites grouped by reach, ChatGPT had the weakest coverage relationship. These are observational results, not established ranking factors. Findings

My interpretation is that a content plan and a visibility report need separate jobs.

A content plan should tell me which buyer questions we can answer well. A visibility report should tell me what an assistant actually returned when we asked those questions. I would never accept the existence of a page as proof that the company is visible.

I would also keep being named separate from being cited. For a consultancy helping brands get named by assistants, I want to record whether the answer mentions the company and whether it links to the company’s website. Those are different observations. I would not silently substitute a citation metric for the business outcome a founder asked me to track.

What I think coverage misses

The tempting response to a study about topical authority is to commission a bigger topic cluster. I would first ask what evidence supports that decision for this particular company.

Suppose, as a hypothetical example, a specialist supplier already has a detailed page answering a buyer’s compatibility question. The assistant names another supplier. My first move would be to inspect that answer and its sources. Writing another version of the same explanation would be an untested response to the gap.

The reverse situation calls for different work. If the company cannot point to a clear, accurate answer to the question, I would improve the information before treating the absence as a distribution problem.

The second-order effect I care about is how teams allocate work. If we combine content completeness and observed visibility into one score, I think we make it harder to decide who should act. The content owner needs a specific missing answer. The person investigating citations needs the saved assistant response and the sources it used.

I also want a later test. I would not treat a relationship observed earlier in the year as a current operating rule without checking it against the questions our buyers ask now.

Four steps I would take before expanding the content plan

1. Choose buyer questions and keep them stable

I would start with questions that affect a real buying decision. For example, a founder could ask sales staff which questions repeatedly delay a purchase, then select the ones the company can answer from direct knowledge.

I would save the exact wording and keep the location and collection settings consistent. If I change the question, I would label it as a new test.

The reason is comparability. I want to know whether the observed answer changed. I do not want a rewritten prompt to become an apparent marketing win. I would also retain disappointing answers, rather than replacing them with prompts that make the company look better.

2. Map each question to the best existing answer

For every selected question, I would identify the page that should help a buyer answer it. Then I would read the relevant passage without relying on the page title.

Does it answer the question directly? Can the company support what it says? Where there is a limitation, does the passage explain it?

I would mark a missing or incomplete answer as a content task. If the answer is already useful, I would record that separately from whether an assistant cited it.

The reason is to avoid prescribing more writing before diagnosing the gap. This is also where I would resist making a page merely to satisfy a spreadsheet row. The proposed page should have a clear use for a buyer.

3. Keep an engine-specific record of mentions and citations

I would save each answer with its collection date and the engine used. Beside it, I would record whether the company was named and which source pages supported the response.

I would review each engine separately before producing any combined summary. If the company appears in Perplexity but is absent from ChatGPT, I want that difference visible to the person choosing the next task.

The reason is decision quality. A blended score can make a specific weakness look like an average result. I would rather give a founder a small set of inspectable answers than a dashboard whose headline cannot be traced back to the evidence.

4. Make a bounded change and leave a comparison untouched

I would choose a content change with an explicit reason. Perhaps the existing page leaves a buyer’s question unanswered. I would record the edit and continue collecting the same prompts. I would leave comparable pages unchanged where practical.

Before making the edit, I would define what I hope to observe: a more accurate description of the company, or a citation to the improved answer. I would keep those outcomes separate.

The reason is to make the result reviewable. If visibility improves, I want to examine whether untouched pages changed too. If it does not, I want a clear record of what we tried. Either outcome should help decide the next piece of work without turning a single answer into a promise.

Sources

Wellows: Does Topical Authority Show Up in AI Citations? What 9,471 Questions Reveal, including its methodology and scope.

I founded Newtation, where we work on AI visibility.