Spoke
What Google actually documents
Google published a guide to showing up in its AI features, and a big chunk of it is a list of popular tactics it says you don't need. That makes it the most useful source in this field, and the most uncomfortable one.
Every quotation here is from Google's own documentation on optimizing for generative AI features and on AI features in Search, retrieved 15 November 2026 and linked below.
For most of this field's short life there was no primary source. People inferred. Agencies published checklists. Nobody could check any of it against something a platform had actually said.
That changed. Google published a guide to showing up in its AI features, and a big part of it is Google telling you which popular tactics you don't need.
Does Google publish how to optimize for AI Overviews and AI Mode?
Yes. Google publishes a guide to it, and the main point is that this isn't a separate job: "From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO."
Much of the guide is a list of things you don't need to do. Google says there's no special markup required, no need to chop content into small pieces, no need to write a particular way for generative search, and no need for new machine-readable files, AI text files or Markdown — and that such files neither help nor hurt.
What's actually required
Short and boring:
- The page has to be indexed.
- It has to be allowed to show with a snippet — which is why killing your own snippet kills your AI eligibility.
- The site must not be opted out of Search's generative AI features.
Google adds that clearing those doesn't guarantee it'll crawl, index or serve you. It's a floor, not a promise.
The mythbusting, which is the good part
This is the bit worth printing out, because every line has a thriving market behind it.
| What gets sold | What Google says |
|---|---|
| Publish an llms.txt or similar AI file | "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search" — and such files neither help nor hurt |
| Chop your content into small chunks for AI | "There's no requirement to break your content into tiny pieces for AI to better understand it" |
| Hit the optimal page length | "There's no ideal page length" |
| Rewrite everything in an AI-friendly style | "You don't need to write in a specific way just for generative AI search" |
| Add special AI schema | "There's no special schema.org markup you need to add" |
| Seed mentions of your brand to get cited | "Seeking inauthentic 'mentions' across the web isn't as helpful as it might seem" |
Query fan-out
Worth understanding, because it changes what "ranking for a search" even means.
Google documents that its AI features may use query fan-out, which it describes as "a set of concurrent, related queries generated by the model to request more information." The system then goes and fetches more results to answer the question. Google says this lets it show a wider, more varied set of links than classic search would.
What that means for you: the question somebody typed isn't necessarily the question your page got pulled in for. Covering the neighboring subtopics creates chances that exact-match optimization never will.
Why this page exists
Because the gap between what's documented and what's sold is the widest it's ever been, and the documentation is finally good enough to close it.
Three things follow.
Most GEO checklists are guesswork. That's not automatically wrong — practice always runs ahead of documentation. But a checklist that doesn't separate what's documented from what's guessed isn't giving you enough to decide with.
Several popular tactics are now documented as unnecessary by the platform they're aimed at. That's a lot stronger than "unproven." If somebody is billing you for llms.txt implementation as an AI-visibility measure, Google has published that it doesn't affect visibility in its features.
This is all about Google. It says nothing about how ChatGPT, Claude or Perplexity pick sources, and none of them publishes an equivalent.
What it doesn't settle
How any engine chooses among the pages it already retrieved. Google documents who's eligible and debunks the myths. It doesn't publish a ranking spec for AI answers, and nobody else does either.
So here's the honest position. The entry requirements are documented and dull. The popular shortcuts are documented as useless. And how anything actually picks a source is still unpublished. A field sitting on those three facts ought to be a lot more modest than it is.
More in GEO & entities
How AI engines retrieve
Retrieval is what lets a model quote something you published this morning. Understanding it tells you which levers exist — and which widely repeated tactics are aimed at nothing at all.
Training vs retrieval
One of these you can change this afternoon. The other changes when somebody else trains a model. Knowing which is which tells you where effort is worth spending and where it is wasted.
Platform matrix
One row per engine, each with its own last-verified date and a link to the platform's own documentation. Where a row could not be verified, it says so instead of being filled in from memory.
Entity consistency
The highest-value work in getting quoted by AI is also the dullest. Describe your company the same way everywhere, give it one stable identifier, and stop handing machines reasons to hedge about who you are.
Comparative content
Comparison questions are among the most valuable a buyer asks and among the most common an AI engine is given. The format decides whether you can be quoted — a table in HTML can be, the same table as an image cannot.