Google Says Ignore llms.txt and “GEO Hacks”. Do This Instead

A lot of advice about AI search is sold under new names: AEO (answer engine optimisation) and GEO (generative engine optimisation). Google now has an official page that answers most of the questions people ask about it. The guide, Optimizing your website for generative AI features on Google Search, is written for site owners and covers AI Overviews and AI Mode. Its short answer is that AI search optimisation in Google is still SEO, and several popular tactics do nothing for Google at all.

What Google says AEO and GEO really are

The guide addresses the new terms directly. It says that, from Google Search’s point of view, “optimizing for generative AI search is optimizing for the search experience, and thus still SEO.” It then points anyone considering third-party AEO or GEO services to its guidance on evaluating outside SEO advice, which the earlier post on Google’s guidance on hiring an SEO walks through.

The reason is in how the features work. Google explains two techniques in plain terms:

• Retrieval-augmented generation (grounding): the AI feature uses Google’s normal ranking systems to fetch relevant, up-to-date pages from the Search index, then builds its answer from them and links to them.
• Query fan-out: the model runs several related searches at once to gather more material. Google’s example is a search for fixing a weedy lawn that fans out into searches about herbicides, chemical-free weed removal and prevention.

Both depend on the ordinary Search index and ordinary ranking. So a page that is not indexed, or not good enough to rank, has little chance of being used.

The tactics Google says you can skip

The guide has a section called “Mythbusting generative AI search”, which lists things you can ignore for Google Search. These are the main ones:

1. llms.txt files and other “special” markup. Google says you don’t need new machine-readable files, AI text files, markup or Markdown to appear in Google Search, “as Google Search itself doesn’t use them.” It adds that creating an llms.txt file for other services is fine, but it “will neither harm nor help” your visibility in Google.
2. Chunking content. There is no need to break pages into tiny pieces for AI. Google says its systems understand pages that cover several topics, and that there is no ideal page length.
3. Rewriting content just for AI. AI systems understand synonyms and general meaning, so you don’t need to chase every long-tail variation of a query.
4. Seeking inauthentic mentions. Paying for or planting mentions across blogs and forums is less useful than it looks, because the same spam systems that protect normal results also protect the AI features.
5. Overfocusing on structured data. Structured data is not required for AI search, and there is no special schema to add. It is still worth using for rich results.

On llms.txt, this matches what the earlier post on WebMCP and llms.txt found: the file may be read by some other AI services, but Google has now said in its own documentation that Search ignores it. If a service sells llms.txt as a way into AI Overviews, that claim is now contradicted by Google directly.

There is also a warning against a tempting shortcut. Creating separate pages for every way a question can be phrased, including fan-out queries, “primarily to manipulate rankings or generative AI responses” breaks Google’s scaled content abuse policy, the guide says.

What Google says to do instead

The guide’s positive advice is short. In order of importance, as Google presents it:

• Make non-commodity content. Google calls this more influential in the long run “than any of the other suggestions in this guide”. Commodity content is common knowledge anyone could write. Non-commodity content brings expert or first-hand insight. A later post in this series looks at how to apply that to a normal business page.
• Keep the technical basics right. To appear in AI features, a page must be indexed and eligible to show with a snippet. The site must also be included in the Search generative AI features setting in Search Console, which is on by default.
• Use good images and video where they genuinely help, following Google’s normal image and video guidance.
• Keep business details current in Merchant Center and Google Business Profile, because AI answers can show products and local business information.
• Measure with Search Console. The guide points to the Generative AI performance report, covered in the earlier post on what the Search Console AI report shows.

It also warns about tools: “No third-party tool has access to our internal ranking or AI systems.” Tools can help your workflow, but any tool claiming to know Google’s internal AI metrics is overselling.

A practical checklist for a WordPress site

If you want to act on the guide this week, these steps cover most of it:

1. In Search Console, open Settings → Search generative AI and confirm your site is included. If you excluded it on purpose, the earlier post on whether to exclude your site from Google’s AI features covers the trade-off.
2. Use the URL Inspection tool on your five most important pages. Each one should be indexed, and none should carry a nosnippet tag by accident.
3. Check that robots.txt, your firewall and your CDN all let Googlebot in. Google lists this as a basic.
4. Pick one thin page and rewrite it with something only you can say: a real process, a real example, a real number from your own work.
5. If you already have an llms.txt file, you can leave it. If you don’t, Google gives you no reason to add one for Google.

In WordPress, most of this happens outside any plugin. Rank Math can manage your titles, robots meta and schema, but the guide’s main message is about content and access, not settings. That is good news for small sites: the work Google asks for is the same work that has always helped people find you.