Google Names the AI SEO Hacks That Don’t Work

Google’s new guide on optimizing for generative AI features does something Google does not do very often: it names specific tactics and tells site owners to stop bothering with them. Tucked into the guide is a section called “Mythbusting generative AI search: what you don’t need to do,” and it reads almost like a list of what the AEO and GEO hype has been selling for the past two years, from special files to chunked content to copy rewritten for machines. Google’s position on all of it is that none of these tactics are necessary to show up in AI Overviews or AI Mode, and that the effort is better spent on fundamentals.

For a marketplace built on the idea that most of this work has no shortcuts, the list is satisfying to read. Here is what Google says you can stop doing.

llms.txt and the myth of special files

One of the most popular ideas in the AEO space has been the llms.txt file, a proposed text file that site owners add to tell AI systems how to find and read their content. Plenty of tools and guides have pitched it as a way to get an edge in AI search. Google’s guide addresses it directly: you do not need to create new machine-readable files, AI text files, markup, or Markdown to appear in generative AI search. Google may discover, crawl, and index many kinds of files beyond HTML, but as the guide puts it, that does not mean the file is treated in any special way.

Adding an llms.txt file does not give a page special standing in AI answers, because Google pulls content into AI features from its regular index, built by crawling the web the way it always has. A separate file aimed at AI systems is not part of that process, and the time spent creating and maintaining one does nothing for visibility in AI Overviews or AI Mode.

Chunking content into pieces nobody needs

Another common piece of advice has been to “chunk” content, breaking pages into small, self-contained blocks on the theory that AI systems parse tiny pieces more easily. Google’s guide pushes back on this too. There is no requirement to break content into tiny pieces for AI to understand it, because Google’s systems can already handle the nuance of multiple topics on a page and how they relate to each other.

Google does add a sensible caveat: shorter content can work well sometimes, depending on the audience and the subject. But the guide is clear that there is no ideal page length for generative AI search, which takes the air out of the idea that there is an optimal chunk size to engineer toward. Writing for a reader, at whatever length the topic needs, does more than slicing a page into fragments for a machine that does not need them.

Rewriting your copy for the machines

A third hack has been rewriting content in a special way for AI systems, stuffing in synonyms and long-tail variations so a page matches every possible phrasing. Google’s guide says this is unnecessary. AI systems can understand synonyms and the general meaning of what someone is looking for, and they can connect a searcher to content that does not use the exact words in the query. The guide spells out the implication: you do not have to worry that you are short on long-tail keywords or that you have not captured every variation of how someone might phrase a search.

For anyone who has been told to rewrite human-friendly copy into something tuned for machines, this is permission to stop. The content that reads well for people is the content Google’s systems are built to understand, which means the editing effort is better spent making a page clearer and more useful than making it legible to an algorithm that already reads plain language fine.

The structured data myth resurfaces

The last of the technical hacks is overfocusing on structured data. The belief that adding schema markup boosts AI visibility has been one of the stickiest ideas in the space, and Google addresses it plainly: structured data is not required for generative AI search, and there is no special schema.org markup that needs adding. Google does recommend continuing to use structured data as part of an overall SEO strategy, since it still helps a page become eligible for rich results, but it draws a firm line against treating schema as an AI ranking factor.

This matches what the data has already shown. The Ahrefs study we covered recently tested whether adding schema actually causes more AI citations and found that it does not, even though schema is roughly three times more common on AI-cited pages. The correlation was real, but the cause-and-effect story fell apart under testing. Google saying the same thing in its own guide should settle the question: schema earns its place for rich results and clarity rather than as a lever for AI citations.

The fundamentals Google keeps pointing to

Strip out the hacks and the guide keeps returning to the same short list: valuable, non-commodity content built on real experience, a clean technical foundation that lets pages get crawled and indexed, and the authority signals that core ranking has always rewarded. The myth-busting section is really an argument for spending time on those things instead of on markup tricks that do nothing.

None of it is as flashy as a new file format or a clever schema trick, but it is the work that actually moves AI visibility. Link building and digital PR build the third-party authority that makes a page a strong candidate for AI features, the kind of signal Google’s systems actually weigh, unlike an llms.txt file. Thorough, well-structured content gives Google something substantial to retrieve and cite, and neither needs a special format, a chunk size, or a schema trick to work.

Google publishing this list is useful beyond the specifics, because it hands site owners a way to tell real advice from hype. Anyone selling an AI-optimization tactic that Google explicitly says is unnecessary is selling something the guide already tells you to skip. The fundamentals are unglamorous and they take longer, but they are the only part of the AI search picture that has survived scrutiny.