AI search September 18, 2026
What AI search is, and why classic SEO is no longer enough
AI search answers in the results page and cites sources. Google’s own 2026 guide says it is still SEO, but some expectations have changed.
AI search means a search engine or assistant writes the answer itself and shows sources next to it. Users no longer open ten pages. Your site has to be what the answer is built on.
What changed in 2026
On 15 May 2026 Google published its first official guide to optimising for generative AI. The core message is simple: AEO and GEO are not separate fields, this is still SEO. AI answers rest on Google’s ordinary quality and ranking systems.
The guide also names practices you do not need for Google Search:
- llms.txt and other AI-specific files: Google says it does not need or use them (June 2026 clarification)
- chopping content into small chunks for AI: Google can find the relevant part on a multi-topic page
- rewriting for every long-tail variant: the system understands synonyms
- fake mentions on blogs and forums: quality and spam systems work against it
What it means for traffic
Example: in a Pew Research Center study (900 US adults, March 2025), users clicked a regular result in 8% of visits when the page showed an AI summary and in 15% when it did not. A link inside the summary was clicked in about 1% of visits. An AI summary appeared in 18% of the searches studied.
That is US data from a single month, not a law. The direction is still clear: some questions get answered without a click. Google itself says clicks from AI pages tend to be higher quality, with users spending more time on the site.
How sources are chosen
Per Google’s documentation, AI Overviews and AI Mode use “query fan-out”: several related searches run to find a wider set of pages. That is why a page outside the classic top ten can show up in an answer.
To qualify, a page must be indexed and eligible to show with a snippet. Google states it plainly: there are no additional requirements to appear in AI Overviews or AI Mode. AI traffic is counted in the ordinary “Web” report in Search Console, there is no separate AI report.
What to do: five steps
- Check that Googlebot, OAI-SearchBot and PerplexityBot can reach the page (robots.txt and CDN)
- Keep important content as HTML text, not in images or late JavaScript
- Answer the question at the start of the paragraph, then add reasoning and examples
- Add real experience and data nobody else has: Google calls this non-commodity content
- Measure Search Console queries, citations in AI answers and qualified traffic, not just positions
A practical example
On this site, article text is rendered on the server, so it is present in the very first HTML response. That is why a bot that does not run JavaScript sees the same text as the reader. It is not an AI trick, just ordinary technical hygiene, and Google recommends it too.
Three questions clients ask
Do we need separate pages for AI?
No. Google’s documentation says there are no additional requirements for AI features and no separate AI files or markup are needed. Make sure the ordinary page is good.
Is llms.txt worth making?
Not for Google Search: per the June 2026 clarification it is not used and does not affect rankings. You may keep the file for other services, but do not expect a Google result from it.
How do you measure AI visibility?
Google’s AI traffic sits in the ordinary “Web” report in Search Console. For other assistants, ask questions on your topic and note which sources they show. Repeat monthly with the same questions to see change.
Summary
AI search does not need a new discipline, just good old work: an indexable page, a clear answer, real experience and trustworthy sources. What is new is that some questions get answered without a click, so you need to measure visibility, not only visits.
