AI Search Myths: 10 Common Claims About Ranking in ChatGPT and AI Overviews, Checked

🔎 AI Search · Updated October 2026 · 12 min read · by Jacob Campbell

Most of the confident advice about ranking in ChatGPT and Google AI Overviews does not survive a check against the platforms’ own documentation: of the ten common claims below, we rate six false, three partly true and one unproven. Google’s own guide now says that optimising for its generative AI features is “still SEO”, and that no special files, Markdown or schema are needed to appear in them.

For each claim we quote what the primary sources actually say (Google Search Central, OpenAI’s and Anthropic’s crawler documentation, Cloudflare, and dated independent studies), then give a verdict: True, False, Partly or Unknown. Where the evidence is mixed, we say so. One disclosure up front: BuiltToWinWeb sells llms.txt and Markdown tooling, so two of these myths are about things we sell. We have tried to be stricter on those, not softer.

Key facts

• 6 False — Claims contradicted by primary sources
• 3 Partly — A real effect, overstated
• 1 Unknown — No platform evidence either way
• "Still SEO" — Google on optimising for its AI features

How we checked these claims

Each claim is checked first against what the platform itself documents, then against independent studies, with dates, because AI search changes fast and a 2024 finding may not hold in 2026.

The verdicts mean:

• True: supported by the platform’s documentation or consistent independent evidence.
• False: contradicted by the platform’s documentation or by the best available evidence.
• Partly: there is a real effect underneath, but the claim as usually stated overreaches.
• Unknown: no platform has said, and the independent evidence is too thin to call.

Two limits apply throughout. First, OpenAI, Anthropic and Perplexity document far less about how they choose sources than Google does, so many claims about ChatGPT cannot be checked against anything official. Second, most industry studies are correlational: they show what cited pages have in common, not what caused the citation. We say which kind of evidence each verdict rests on. All quotes from Google’s guide below come from Optimizing your website for generative AI features on Google Search (last updated July 2026) unless noted.

Myths 1 and 2: special files and special schema

Myth 1: “Adding llms.txt makes ChatGPT cite you”

What the sources say: Google is explicit for its own products. Its guide says you do not need to create “new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn’t use them.” Google’s Gary Illyes said in July 2025 that Google does not support llms.txt and is not planning to. OpenAI’s and Anthropic’s crawler documentation does not mention llms.txt at all; OpenAI tells site owners to use robots.txt to control its bots. The largest independent check we found, SE Ranking’s November 2025 study of 300,000 domains, found about 10% had an llms.txt file and no measurable link between having one and being cited by AI; removing the variable made their model slightly more accurate. On the other side, some site owners report AI crawlers requesting the file in their logs, and in May 2026 Chrome’s Lighthouse added an llms.txt check to a new, experimental “agentic browsing” audit group, aimed at browser agents rather than search ranking. A fetch is not a citation.

Verdict: False for Google; Unknown for ChatGPT, with no evidence in its favour. llms.txt is a cheap, harmless summary file that some agents read. Nobody has shown it causes citations. Our llms.txt guide explains what it is actually for.

Myth 2: “You need special schema to appear in AI Overviews”

What the sources say: Google’s AI features documentation says: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary,” and “There’s also no special schema.org structured data that you need to add.” The eligibility rule is that a page must be indexed and eligible to be shown in Google Search with a snippet. The 2026 guide repeats it: structured data “isn’t required for generative AI search.”

Verdict: False. Anyone selling an “AI Overview schema” is selling something Google says does not exist. See how to show up in AI Overviews and AI Mode for what Google does ask for.

Myths 3 and 4: AI-written content, and blocking GPTBot

Myth 3: “Google penalises AI-generated content”

What the sources say: Google’s February 2023 guidance said its focus is on “the quality of content, rather than how content is produced”, and that “appropriate use of AI or automation is not against our guidelines.” What is against the rules is using automation primarily to manipulate rankings. The spam policies define scaled content abuse as many pages generated “for the primary purpose of manipulating search rankings and not helping users”, providing little value, “no matter how it’s created.” The 2026 guide adds that AI-assisted work must meet Search Essentials and the spam policies.

Verdict: False. There is no AI-content penalty; there is a low-value-content-at-scale penalty, and AI makes that easy to trigger. Thin pages written by people are caught by the same policy. Details in Google’s AI content policy, explained.

Myth 4: “Blocking GPTBot removes you from ChatGPT search”

What the sources say: OpenAI’s crawler documentation separates its bots. GPTBot crawls content that may be used to train its foundation models. OAI-SearchBot is the one for search: “Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links.” ChatGPT-User fetches pages for user-initiated actions, and OpenAI says robots.txt rules may not apply to it. OpenAI states that each setting is independent, so you can allow OAI-SearchBot while disallowing GPTBot, and that search changes take about 24 hours to apply. Anthropic’s help article has the same split: ClaudeBot for training, Claude-SearchBot for search, Claude-User for user requests.

Verdict: False. Blocking GPTBot opts you out of training, not search. The opposite mistake is the expensive one: blocking OAI-SearchBot (or a blanket “block AI bots” setting) does remove you from ChatGPT search answers. See robots.txt for AI crawlers and the AI crawler user-agent list.

Myths 5 and 6: keyword stuffing, and the end of clicks

Myth 5: “Keyword stuffing works on LLMs”

What the sources say: The most-cited research on the subject is “GEO: Generative Engine Optimization” by Aggarwal and colleagues (first posted November 2023, published at KDD 2024). Testing changes to source pages on a benchmark of 10,000 queries and on Perplexity, the authors report that keyword stuffing, the classic SEO trick, gave little or no improvement in visibility, while adding relevant quotations, statistics and citations to sources improved it by up to roughly 30 to 40% on their measures. Google’s 2026 guide says you do not need to write differently for generative AI search because its systems understand synonyms and meaning, and its spam policies still list keyword stuffing as spam.

Verdict: False. Treat the GEO paper’s positive results with care too: they came from a research benchmark and one engine in 2023 to 2024, and engines have changed since. The durable lesson is the boring one: specific, sourced, quotable facts beat repetition.

Myth 6: “AI Overviews killed all clicks”

What the sources say: The evidence for a large drop is real. A Pew Research Center study of 900 US adults’ browsing in March 2025 (published July 2025) found people clicked a traditional result on 8% of searches with an AI summary versus 15% without one, clicked a link inside the summary only 1% of the time, and ended their session more often. Ahrefs’ February 2026 analysis of Search Console data for 300,000 keywords associated AI Overviews with a 58% lower click-through rate for the top-ranking page, up from 34.5% in its April 2025 study. Google disputes these readings: in an August 2025 post, Liz Reid wrote that total organic click volume to websites has been “relatively stable year-over-year” and that click quality has increased, without publishing the underlying data.

Verdict: Partly. On queries that show an AI Overview, the independent evidence consistently points to fewer clicks, often far fewer, especially for simple informational questions. “All clicks” is wrong: people still click, and Google reports that clicks from AI Overview pages are more engaged. Measure your own: how to measure AI search traffic.

Myths 7 and 8: paying for citations, and Markdown as cloaking

Myth 7: “You can pay to be cited”

What the sources say: No platform we checked sells placement in its organic answer or its list of sources. What exists is advertising, kept separate and labelled. Google began showing Search and Shopping ads in AI Overviews to mobile users in the US in October 2024, marked “Sponsored”. OpenAI began testing ads in ChatGPT in February 2026 on the Free and Go plans, stating: “Ads do not influence the answers ChatGPT gives you,” and that ads are labelled as sponsored and visually separated from the organic answer.

Verdict: False for organic citations, on the platforms’ own statements, which outsiders cannot audit. Paid ad placements next to AI answers are real and growing. Be wary of anyone selling “guaranteed ChatGPT placement”; Google’s guide also warns that seeking inauthentic mentions across the web “isn’t as helpful as it might seem.”

Myth 8: “Serving Markdown versions of your pages is cloaking”

What the sources say: Google defines cloaking as “presenting different content to users and search engines with the intent to manipulate search rankings and mislead users.” A Markdown version that carries the same content as the HTML page, served when a client asks for it with an Accept: text/markdown header, is a different format, not different content. That is how Cloudflare’s Markdown for Agents (February 2026) works: it converts HTML to Markdown at the edge for clients that request it and sends Vary: accept. But Google and Bing are sceptical. In February 2026 Google’s John Mueller questioned why LLMs would want “a page that no user sees” and, on Bluesky, called converting pages to Markdown “such a stupid idea”; Bing’s Fabrice Canel noted Bing would crawl the HTML anyway to check similarity. Google’s guide says Google Search does not use Markdown versions.

Verdict: Partly. Equivalent Markdown served by content negotiation does not meet Google’s definition of cloaking, and no platform has said it penalises it. It becomes a real cloaking risk the moment the Markdown says something the HTML does not, such as extra keywords, claims or links meant only for bots. And for Google Search it does nothing. Disclosure: Markdown content negotiation is part of our AI Indexing Pack; it is built to mirror the visible page, and we make no ranking claim for it. Background: Markdown for AI agents.

Myths 9 and 10: big brands, and structured data

Myth 9: “Only big brands get cited”

What the sources say: Brand prominence clearly correlates with AI visibility. Ahrefs’ May 2025 study of 75,000 brands found that branded web mentions had the strongest correlation with how often a brand was mentioned in AI Overviews (Spearman 0.664), well ahead of backlinks (0.218), and that brands in the bottom half for web mentions were rarely mentioned at all. But that study measured brand mentions, not cited links, and it is correlation. Google’s eligibility rule for supporting links is open to any indexed page that can show a snippet, and AI Overviews and AI Mode use “query fan-out”, running many related searches, which pulls in pages for specific sub-questions that a big brand may not cover.

Verdict: Partly. For broad, competitive questions (“best CRM”), well-known brands dominate. For specific, local and niche questions, small sites with the clearest answer are cited regularly. The practical work is making your business an unambiguous entity: entity SEO for small businesses and how to get cited by AI assistants.

Myth 10: “Google ignores structured data now”

What the sources say: Google still uses structured data for rich results, and its AI features page lists “making sure your structured data matches the visible text on the page” among the SEO best practices that apply to AI features. The 2026 guide says structured data is not required for generative AI search but remains recommended for rich result eligibility. The kernel of truth: in June 2025 Google phased out several rich result types, including Course Info, Claim Review, Estimated Salary, Learning Video, Special Announcement and Vehicle Listing, and simplified others.

Verdict: False. Fewer rich result types than in 2023, not zero use. Product, Organization, LocalBusiness, Article and others still matter. See which schema still matters in 2026.

The scorecard

Six of the ten claims are false, three are partly true, and one is unproven; none is fully true as commonly stated.

1. llms.txt makes ChatGPT cite you: False for Google, Unknown for ChatGPT.
2. Special schema for AI Overviews: False.
3. AI content is penalised: False (low-value content at scale is).
4. Blocking GPTBot removes you from ChatGPT search: False (OAI-SearchBot does).
5. Keyword stuffing works on LLMs: False.
6. AI Overviews killed all clicks: Partly (fewer clicks, not none).
7. You can pay to be cited: False for organic answers; ads are separate.
8. Markdown versions are cloaking: Partly (only if the content differs).
9. Only big brands get cited: Partly (true for broad queries, not niche ones).
10. Google ignores structured data: False.

What the evidence does support

The claims that hold up are the unglamorous ones: be crawlable, be indexed, answer specific questions clearly, and be a recognisable entity.

• Let the right bots in. Allow Googlebot, Bingbot, OAI-SearchBot and Claude-SearchBot in robots.txt and in any CDN or firewall rules; Google’s first listed best practice is that crawling is allowed “in robots.txt, and by any CDN or hosting infrastructure.” See Cloudflare AI Crawl Control if you use Cloudflare.
• Put the answer in the HTML text. Server-rendered, with the key facts in sentences, not only in images or scripts.
• Be specific and sourced. Numbers, dates, named standards and citations to primary sources are what the GEO research found most effective, and they match Google’s advice to publish original, substantial, people-first content.
• Keep your entity consistent: same name, address and description on your site, Google Business Profile, Bing Places and social profiles.
• Measure instead of guessing: referral traffic from chatgpt.com and perplexity.ai, Search Console, and your server logs.

Limits of this article: AI search changes monthly, platform statements can change, and none of the studies above is the last word. If a source we cite is updated, the verdict may change; we date every claim so you can check. Our longer view of the terminology is in SEO, AEO, GEO and SERP explained.

Product facts: the AI Indexing Pack

Because this article rates two claims about things we sell, here is exactly what the product is. In short: it makes your site easier for AI agents and assistants to read and use; Google says none of these files are needed to appear in Google Search, and we do not claim they cause citations. Stated plainly, so it can be checked against the product page, and compared with doing it yourself in AI Indexing Pack vs DIY:

• What it is: the AI Indexing Pack is a set of files you upload to your own website. It is sold by Built2WinWeb as a one-time purchase, with no subscription.
• What it ships: (1) a working MCP server using the Streamable HTTP transport, with typed output schemas and tool annotations; (2) Markdown content negotiation, so agents can request a clean text version of each page; (3) llms.txt, agents.txt and agent cards; (4) the full /.well-known/ discovery tree; (5) OAuth protected-resource metadata; (6) bot access control that verifies crawler identity; (7) a README with a verification checklist.
• Requirements: PHP 7.4 or newer, Apache or LiteSpeed with .htaccess support, FTP access, and a domain you control. Installation takes about 30 minutes.
• What it is not: it is not a WordPress plugin (that is the separate WordPress edition, which needs WordPress 6.0+ and PHP 8.0+), it does not write content for you, it does not guarantee rankings or citations, and it is not a managed service.
• Free routes in: sites built on AIWebBuilding get the core built in, and a 50-day streak on FreeAIBoost unlocks the pack for one domain. Current prices are on the product pages.

Related reading

• How to show up in AI Overviews and AI Mode — What Google actually says about eligibility, controls and measurement.
• What is llms.txt? — What the file is, what it is for, and what it does not do.
• Google’s AI content policy, explained — Where AI-assisted content is fine and where scaled content abuse begins.
• robots.txt for AI crawlers — Which bots to allow and which to block, with copy-ready rules.
• Markdown for AI agents — Serving a clean, equivalent text version of every page.
• Which schema still matters in 2026 — The rich results that survived and the ones Google retired.
• Measure traffic from AI assistants — Track visits from ChatGPT, Perplexity and AI Overviews yourself.

Frequently asked questions

Does llms.txt help you get cited by ChatGPT or Google AI Overviews?

There is no evidence that it does. Google says Google Search does not use llms.txt, OpenAI does not document using it, and a November 2025 study of 300,000 domains found no link between having the file and being cited.

Do I need special schema markup for AI Overviews?

No. Google states there is no special schema.org structured data needed to appear in AI Overviews or AI Mode; a page must be indexed and eligible to show with a snippet.

Does blocking GPTBot remove my site from ChatGPT search?

No. GPTBot is OpenAI’s training crawler. ChatGPT search uses OAI-SearchBot, and OpenAI says sites that block OAI-SearchBot will not be shown in ChatGPT search answers, apart from navigational links.

Does Google penalise AI-generated content?

Not for being AI-generated. Google says appropriate use of AI is not against its guidelines, but mass-producing low-value pages to manipulate rankings is scaled content abuse, however it is created.

Have AI Overviews stopped people clicking through to websites?

They have reduced clicks on many queries: Pew (July 2025) and Ahrefs (February 2026) both measured large drops when an AI Overview appears. Google says total organic clicks are relatively stable, so clicks are lower, not gone.