There is a small industry selling tactics that Google has publicly told us do not affect AI Overviews. We checked each one against Google's own documentation. Eight of the most-repeated claims are wrong, and a few of them can quietly hurt you. This is the list, with the exact quote and link for every one.
What Google says
“There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.”
Why this matters for AI Overviews
Every "AI SEO" agency has the same incentive: invent a thing, sell the thing, then sell the maintenance. The thing rarely needs to be invented from scratch. Google has been unusually direct about what does and does not affect AI Overviews, and a careful reading of the AI optimization guide kills most of the popular tactics on contact.
We built the AI Overview Checker by reading that guide line by line and only implementing checks Google explicitly endorses. Along the way we kept finding "best practices" that don't actually map to anything Google said. Here are the eight we hear most often, with Google's exact words and a verdict on each.
How to fix it
Myth 1: You need an llms.txt file for Google AI Overviews
"You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in generative AI search." Source: Google AI optimization guide
llms.txt is a community proposal, not a Google standard. Google does not crawl it, does not consume it, and has stated they have no plans to. If your agency told you "we need to ship llms.txt for AI Overviews," that work is going to a file Google ignores.
We still build llms.txt files at SiteSpeakAI (our generator is here) because some third-party tools and AI assistants use them, but the value is outside Google Search. Be honest about which.
Myth 2: There is "AI-specific schema" you need to add
"There's no special schema.org structured data that you need to add." Source: Google AI features in Search
Schema is still worth implementing for rich-result eligibility (Product, Article, LocalBusiness, VideoObject). It is not an AI Overview lever. There is no AIPage, no LLMOptimized type, no special schema unique to AI features. If a vendor is charging you for "AI schema markup," ask which @type they mean, then check it against the schema.org gallery. It will be a normal schema type with a new name attached for the invoice.
Myth 3: You need to chunk content for AI
"There's no requirement to break your content into tiny pieces for AI to better understand it. Google systems are able to understand the nuance of multiple topics on a page and show the relevant piece to users." Source: Google AI optimization guide
Chunking is a Retrieval Augmented Generation (RAG) concept. It applies to systems you build yourself, where you control the chunk size and embed each chunk. It does not apply to Google. Google's models read the page, decide what is relevant, and select. Splitting one good page into five mediocre ones to "help the AI" usually demotes you in both classic Search and AI features.
Myth 4: You need a separate page for every keyword variation
"A high quantity of pages doesn't make a website higher quality or more relevant to users." Source: Google AI optimization guide
"While it might be tempting to create separate content for every possible variation of how people might search... doing so primarily to manipulate rankings... violates Google's scaled content abuse spam policy." Source: same as above
This one is worse than ineffective. The 2024 spam policy update made scaled content abuse an explicit penalty surface. Generating one page per "[best X for Y in Z]" template, with light wording changes, is exactly what the policy targets.
Myth 5: Blocking Google-Extended hides you from AI Overviews
"Google-Extended does not impact a site's inclusion in Google Search nor is it used as a ranking signal in Google Search." Source: Google common crawlers
This one trips up almost every privacy-focused team. Google-Extended controls whether your content is used to train Gemini and to ground Gemini Apps and Vertex AI. It does not touch AI Overviews or AI Mode, which are Search features that use Googlebot's index. If you want to opt out of Gemini training and stay in AI Overviews, block Google-Extended and keep Googlebot. If you wanted to opt out of AI Overviews and you blocked Google-Extended, you blocked the wrong thing.
Myth 6: Inauthentic "mentions" help you get cited
"Seeking inauthentic 'mentions' across the web isn't as helpful as it might seem." Source: Google AI optimization guide
The "get mentioned on 50 niche blogs and AI will cite you" pitch is the new link-building. Google's position is consistent: their ranking systems try to find high-quality content; spam systems try to demote inauthentic signal. Mass-paid mentions are the latter.
Myth 7: Long-tail keyword density still matters
"AI systems can understand synonyms and general meanings, so you don't have to worry that you don't have enough 'long-tail' keywords." Source: Google AI optimization guide
Writing the same idea seventeen ways to "cover variations" wastes words and dilutes the page. Write the idea well, once.
Myth 8: AI Overviews kill clicks, so SEO is dead
"When people click from search results pages with AI Overviews, these clicks are higher quality (meaning, users are more likely to spend more time on the site)." Source: Google AI features in Search
"With AI Overviews, people have been visiting a greater diversity of websites for help with more complex questions." Source: same as above
Two things can both be true: total clicks per result may fall, and the clicks you do get may be more qualified. The strategic move is to chase the kinds of queries AI Overviews trigger on (deep, complex, comparison-heavy) with content that earns the cite. That is regular SEO, done with the AI Overview surface in mind. It is not dead.
What we changed at SiteSpeakAI because of this
After we audited our own site against our own AI Overview Checker, we deleted a llms.txt content campaign we had planned, stopped chasing low-difficulty keyword variants, and replaced both with one priority: build the kind of content the model could not have produced from common knowledge. That is the underlying instruction from Google's "non-commodity content" framing. Everything else, including this page, is secondary.
If you want a copy-paste audit of your own site against the real Google rules (not the agency mythology), run it through the checker. The audit cites Google for every check it makes.
