If you have had a pitch promising to “win at AI search,” you are not alone. Add an llms.txt file. Serve Markdown copies of your pages to bots. Chunk your content into bite-sized pieces. It all sounds urgent, and most of it costs a real budget.
So we went to the source. We took what John Mueller, Search Advocate and Search Relations team lead at Google, said at the Search Central Live Deep Dive 2026 and used it to test the myths most marketers still believe. Google’s own published guidance on generative AI features also confirms most of his points, and we flag the few that come only from the event. Below are 11 myths, each with a verdict on the claim, a recommended action, what it means for your marketing, and what to do instead.
A note on scope: Everything here is about Google Search, including AI Overviews and AI Mode. Other AI platforms, such as ChatGPT, Perplexity and Gemini itself, may behave differently, and we flag where that matters.

#	Claim	Verdict	Action
1	AI search has its own bot and playbook	Myth (Gemini is the exception)	Keep doing SEO fundamentals
2	You need an llms.txt file	Myth for Google Search	Skip as a Google tactic
3	Serve Markdown copies to bots	Myth, and risky	Skip
4	You must chunk content for AI	Myth	Skip
5	Google penalises AI-generated content	Mostly myth, with a catch	Review before publishing
6	More crawling means better rankings	Myth	Ignore crawl-count vanity metrics
7	Top-10 lists and templates scale safely	Myth	Avoid
8	Google matches exact keywords	Myth	Write naturally
9	Quality Rater Guidelines are ranking factors	Myth	Use as a reference only
10	Hreflang decides a page’s language	Myth	Use hreflang only to link translations
11	If the browser shows it, crawlers see it	Myth	Fix with server-side rendering

Myth 1: AI search uses its own bot and its own playbook

The reality: AI Overviews and AI Mode use Googlebot and classic Search. There is no separate crawler or ranking system behind them. Google’s guide describes its generative features as rooted in core Search ranking and quality systems, with each fan-out query running through normal Search. Gemini is the exception: it is not Search and may use a different crawler.

Why it matters for your marketing: Your existing SEO work already feeds Google’s AI features. When a vendor sells “AI visibility,” ask which system they mean, because Search and Gemini are not the same thing.

What to do: Make sure key pages are indexed and eligible to show with a snippet. Track AI visibility in Search Console’s Generative AI performance report.

Myth 2: You need an llms.txt file

The reality: Google Search doesn’t use llms.txt or similar files such as cats.txt. Google’s guide says you don’t need machine-readable files, AI text files, markup or Markdown to appear in Google Search, because Search doesn’t use them. Gary Illyes confirmed at Search Central Live that Google isn’t pursuing llms.txt.

Why it matters for your marketing: Building and maintaining these files costs developer time with no upside in Google. If a vendor pitches llms.txt as a route to AI Overviews or AI Mode visibility, that claim doesn’t hold for Google Search.

What to do: Spend the time on content and technical health instead. If you want the file for other services that read it, Google says that is fine, and it won’t help or harm Google visibility.

Myth 3: Serve Markdown copies of your pages to bots

The reality: John Mueller flagged this as a skip at Search Central Live 2026. A separate bot-only Markdown version creates duplicate content. If implemented incorrectly, it also becomes a cloaking risk, because crawlers would see something different from what users see. Google’s guide lists Markdown alongside llms.txt among the things you don’t need to create to appear in Google Search.

Why it matters for your marketing: A bot-only version adds development and maintenance cost and puts organic visibility at risk, with no benefit in Google Search. Treat any vendor proposing it as a red flag.

What to do: Serve one version of each page to everyone, and put the effort into improving that page. Ask your developers to avoid rules that show different content to bots.

Myth 4: You must “chunk” your content for AI

The reality: At Search Central Live 2026, Google’s John Mueller made clear that publishers don’t need to break content into bite-sized pieces for LLMs. That lines up with Google’s guide, which says there is no requirement to split content into small pieces, because its systems can understand multiple topics on one page and surface the relevant part for users. The guide also says there is no ideal page length.

A caution on the reasoning. The event discussion pointed to Gemini’s large context window (around a million tokens) as one reason chunking isn’t needed. We wouldn’t lean on that argument, because search retrieval and Gemini’s context window are different things. The stronger evidence is Google’s own guidance.

Why it matters for your marketing: Splitting existing pages to suit AI adds production cost without a Google benefit, and fragmented content can read worse for the people you want to convert.

What to do: Organise each page for the reader, with clear headings and well-built paragraphs. Make pages as long or as short as the audience and topic need.

Myth 5: Google penalises AI-generated content

The reality: Google doesn’t penalise content simply for being AI-generated. John Mueller said at Search Central Live 2026 that the question is quality, not how the content was made, and that the real target is low-effort material, such as LLM-written “top 10” reviews of products the author never used.

The catch: Using generative AI to produce many pages without adding value for users may violate Google’s scaled content abuse policy. Google advises fact-checking and reviewing AI-generated content before publishing. This is where agencies are most exposed, because AI makes high-volume publishing easy.

Why it matters for your marketing: AI in your workflow is not a ranking risk in itself. Publishing at scale without anything original is, and that is where quality problems and policy issues start.

What to do: Use AI to support research and structure. Have a human add the experience, the fact-checking and the point of view before anything is published.

Myth 6: The more Google crawls you, the better you rank

The reality: John Mueller told attendees at Search Central Live 2026 to stop chasing crawl frequency. This matches Google’s documentation: a page has to be crawled to appear in results, but crawling is not a ranking signal.

The event added two points that go beyond the documentation:

• A spammy past slows crawling, because historically spammy sites are deprioritised in the crawl scheduler.
• Recovery can take time. John Mueller said SpamBrain, Google’s spam-detection system, now catches about five times more spam, and that manual actions likely help train it. That may explain why a site recovers slowly even after a penalty is lifted. 

Why it matters for your marketing: Crawl numbers are easy to report, but they don’t show whether rankings are improving. A report that presents “more crawling” as a win says nothing about visibility.

What to do: Improve content quality, remove thin pages and keep the site clean.

Myth 7: Top-10 lists and templated pages are a safe way to scale

The reality: John Mueller said at Search Central Live 2026 that commodity content loses. Commodity content is generic content anyone could write, such as a “Top 10” list repeating what every other site says. A templated page is built from the same layout with a few details swapped, for example the same page repeated for 500 cities.

Google defines quality as effort, originality, skill and accuracy: how much of a real person’s knowledge went into the page. He added that low-quality pages may not even be retrieved, and that quality is reused in ranking. It is one signal among hundreds, but one of the most important.

Google’s guide supports this. It says unique, non-commodity content is likely to matter more than any other tip for visibility in generative AI search over time, and uses “7 Tips for First-Time Homebuyers” as an example of generic content, set against a first-hand, experience-led piece. It also warns that creating separate pages for every query variation breaks the scaled content abuse policy, and that more pages don’t make a site higher quality.

He also warned about scale: a template that adds value across 100 pages may add far less across 10,000.

Why it matters for your marketing: Publishing more pages is not a growth strategy on its own. High volumes of similar content cost budget and can put the quality of the whole site at risk.

What to do: Before publishing, ask what the page offers that a reader can’t find elsewhere. If the answer is “nothing,” rework it with first-hand experience, original data or expert opinion, or don’t publish it.

Myth 8: Google matches your exact keywords

The reality: John Mueller explained at Search Central Live 2026 that every query is silently rewritten. Google doesn’t only look for the exact words typed: it also uses synonyms and corrects typos and plurals, and he described the synonym system as one of the most important parts of ranking.

Google’s ranking systems guide agrees. It can return relevant content even when a page lacks the exact words used in a search, because it understands how words and concepts relate. Google’s generative AI guide says the same applies to AI search, so you don’t need to capture every long-tail variation.

Why it matters for your marketing: Repeating a keyword phrase, or writing a separate page for each small wording change, gives no advantage in classic search or AI features. It adds workload and makes pages read poorly.

What to do: Use one well-built page instead of many near-identical ones, and write for the person searching.

Myth 9: Quality Rater Guidelines are ranking factors

The reality: John Mueller explained at Search Central Live 2026 that the Quality Rater Guidelines are instructions for human reviewers who test proposed changes to Search before launch. He cited about 800,000 quality tests in 2023 and about 5,000 launches a year. Google’s documentation says the same: the guidelines are not a guide to ranking first, and ratings don’t directly influence ranking.

They are still a useful reference. They describe how raters judge scaled content abuse (section 4.6.5) and content made with little effort, originality or added value (section 4.6.6).

Why it matters for your marketing: A “rater guidelines audit” that scores pages as if they were ranking factors is built on a wrong assumption and can send budget towards box-ticking instead of better content.

What to do: Read the guidelines to understand what Google means by quality, and use them to shape content standards.

Myth 10: Hreflang decides a page’s language

The reality: Google works out a page’s language from the content on the page. It doesn’t use hreflang, the HTML lang attribute or language codes in the URL, and it assigns each page one language. Hreflang does a different job: it tells Google which pages are translations of each other, so the right one is shown to each searcher.

Why it matters for your marketing: Adding hreflang or a language code to the URL won’t change the language Google sees. Only the words on the page do. Machine translation is not a problem in itself. It is a business choice, and it works if the translation is good.

What to do: Write each page fully in one language. Make sure each version links back to all the others.

Myth 11: If the browser shows it, crawlers see it

The reality: At Search Central Live, Google’s guidance was that a page’s HTML is parsed into a DOM (Document Object Model), a tree-like map of the page. Google then separates the main content from the header, sidebar and footer. Therefore, the DOM structure is more important than ever for JavaScript-heavy sites.

Why it matters for your marketing: Many AI crawlers don’t render JavaScript at all, according to the event recap. Content that only appears after scripts run may never reach them.

What to do: If you want your content to be cited, put it in the raw HTML, or render it server-side.

Conclusion

Across all 11 myths, Google’s message was the same: there is no secret playbook for AI search. AI Overviews and AI Mode run on the Search systems you already work with, so the fundamentals still do most of the work. Keep pages crawlable, content original and the site clean.

Next time a pitch promises AI visibility through a new file, a bot-only version of your pages or content cut into fragments, ask three questions: what does Google say about it, what will it cost to maintain, and what happens to organic visibility if it goes wrong?

If you’d like a second opinion on your SEO and AI visibility plan, get in touch with Focus Tags. We will tell you what’s worth doing, what to skip, and where the real gaps are.