Home · Field notes

November 17, 2026 · Guide

AI search optimization
What Google and Bing document, and what is folklore.

AI search optimization is the work of making your pages easy for AI answers to find, understand and cite. Google says it is not a separate discipline: the SEO fundamentals still apply, and there are no extra technical requirements to appear in its AI Overviews or AI Mode. Most of what is sold on top of that is not documented by anyone.

What the terms cover

"AI search optimization", GEO and AEO are used almost interchangeably. We define them in GEO, AEO, AI SEO. This note is narrower: of everything you will be told to do, what do the two search engines that publish guidance actually say?

What Google documents

Google's own AI features and your website page says a page has to be indexed and eligible to be shown in Search with a snippet, and that there are no additional technical requirements. Its guide to optimizing for generative AI features adds a list of things you can ignore:

  • llms.txt and other "special" files. Google says you do not need new machine-readable files, AI text files, markup or Markdown, because Google Search does not use them.
  • Chunking. It says there is no requirement to break content into tiny pieces, and no ideal page length.
  • Writing for the machine. You do not need to write in a specific way just for generative AI search.
  • Manufactured mentions. Seeking inauthentic mentions across the web is not as helpful as it might seem.
  • Structured data as a lever. It is not required for generative AI search and there is no special schema.org markup. Google still suggests using it as part of normal SEO.

The same guide recommends the Generative AI performance report in Search Console to see how your content shows up, and warns that no third-party tool has access to Google's internal ranking or AI systems.

What Bing documents

Bing's Webmaster Guidelines say the same thing from the other side: Bing and Copilot rely on the same crawling, indexing and ranking foundation as traditional search, and the practices that support discovery, indexing accuracy and content clarity also support eligibility for grounding results and citations. Bing is the one that uses the term directly, describing GEO as focused on content eligibility for grounding and reference in AI responses. It also has an AI Performance report in Bing Webmaster Tools, and notes that changes in share can come from user demand, content changes across the web, model updates, freshness and partner refresh cycles.

What we found when we asked the models

On 17 November 2026 we asked ChatGPT (GPT-5.4, with web search) and Perplexity (Sonar Pro) the same question: what is AI search optimization, what should a small B2B company do, and which claims are documented by Google or Bing and which are unproven?

  • Both said it is not a replacement for SEO and that the foundations are crawlability, useful pages and corroboration.
  • ChatGPT built its answer almost entirely on Google's and Bing's own documentation, and separated what those pages say from what the market claims.
  • Perplexity labelled llms.txt, chunking and ideal page length as unproven and contradicted by Google, which matches the guide above. Its sources, though, were mostly agency and vendor blogs, and its third footnote pointed at "Bing/Microsoft guidance" without a link.

One run each is a sample, not a measurement. The point is the pattern: when the question is phrased as "what is documented", the models lean on the primary sources. When it is phrased as "how do I rank in AI", the results we saw in Google for this topic on the same day were mostly agency guides.

What is left to do

Strip out the undocumented tactics and the list is short, and mostly familiar:

  1. Be crawlable and indexed. Check robots.txt and any CDN or firewall rule that blocks crawlers. Google names the CDN explicitly.
  2. Put the answer in text. Important content should be available as text, with clear internal links.
  3. Write pages for the questions buyers ask, with a direct answer near the top, then the detail, the limits and the evidence. Google calls this unique, non-commodity content.
  4. Keep structured data honest. It must match the visible text. Do not expect it to earn citations.
  5. Get named elsewhere, legitimately. Customers, partners, directories and press that describe you the way you describe yourself.
  6. Measure with the free tools first, then with a fixed prompt set, as in AI visibility.

Where we are careful

Google and Bing document their own products. We found no equivalent official guidance from OpenAI or Perplexity on how their answers choose sources, so for ChatGPT and Perplexity the honest status of any tactic is "plausible, not documented". That includes some of what we do ourselves, such as writing a direct answer in the first two sentences and building corroboration. We think it is sound; we cannot show you a rule that says so.

The part that costs us

If your site already ranks, is crawlable and says specific things, most of this list is work you or your developer can do in an afternoon, and you do not need us. Where we add value is the slower part: finding which buyer questions you are absent from, fixing those pages, and measuring the trend month after month. That is the generative engine optimization service, and it starts from the boring list above, not from a secret one.

Related service: Generative engine optimization.

Want this run on your own site? We'll tell you where you stand in Google and in the AI answers — free, back in 72 hours.

Get your free audit Book a call