The pitch arrives in my inbox most weeks: forget SEO, you now need a completely different discipline for AI search, and here is a new acronym for it. I read these with interest and then go back to what I was doing, because when I examine how an AI overview or an answer engine actually assembles a response, the path to being included looks very familiar.
The system has to find you first
Before a language model can summarise or cite a page, something has to fetch that page. That fetch is a crawl. The page has to return a 200, not be blocked in robots.txt, not be hidden behind a login or a JavaScript render the crawler does not wait for. If the content is injected client-side after an API call, there is a reasonable chance the system sees an empty shell. These are the same crawlability and indexability problems I have been fixing for years. They just have a new consequence: instead of not ranking, you are not in the answer at all.
Then the page has to be understood. A system choosing what to cite needs to work out what the page is about, which parts answer which question, and whether it is trustworthy enough to attach a source link to. Clear headings, one topic per page, direct statements of fact, an identifiable author and a site that is coherent around a subject all help. That is semantic structure and topical authority, described in different words.
An answer engine can only cite what it managed to crawl, parse and trust. That list of requirements is the SEO fundamentals.
What does change
I do not want to pretend nothing is different. A few things are. The click is no longer guaranteed, because the answer may be given on the results page, so informational pages that relied on volume alone will feel it. Being cited matters more than being position one, and citation seems to favour pages that make a specific claim clearly rather than pages that cover everything vaguely. Search is also spreading across more surfaces: Google's AI features, standalone assistants, chat interfaces inside apps. Visibility is no longer a single list on one site.
Those are real shifts, and they change what I prioritise. They do not change the foundation. If anything, a site with clean architecture, fast rendering, well-structured content and a strong topical footprint is in a better position in this environment than it was before, because the systems doing the choosing are less forgiving of mess than ten blue links were.
What I do about it
I run the fundamentals harder. Technical health first, because invisibility to crawlers is fatal. Then content that answers specific questions in specific sentences, so there is something quotable. Then a site that is clearly about something, so the citation is plausible. Then I measure: which pages get cited, for which queries, on which surfaces, and what changes when I alter the page. I have started documenting that in an AI search citation experiment, because I would rather have evidence than a new acronym.
For teams in India this is especially relevant. Mobile-first users, a lot of queries in Hindi and regional languages, and heavy reliance on Google mean that AI features in search will reach a very large audience quickly. The sites that will be cited there are the ones that got the basics right. I go into the detail of what that looks like in the AI SEO guide.