A whole industry of promises has already grown around AEO and GEO: special markup, the supposedly mandatory llms.txt file, copy chopped into perfect little chunks and another dozen ways to “get into AI answers”.
I went back to the primary source — Google’s official guidance. It is much less dramatic. For a page to appear in Google’s AI features, it must be crawlable, indexed and eligible to appear with a normal search snippet. Google does not require separate AI markup or a special file.
What AI search features actually do
Google AI Overviews build a concise answer above the usual results. AI Mode lets people continue the search as a conversation, ask follow-up questions and receive a longer answer with source links. Google may run several related searches behind the scenes to find the right pages.
The practical conclusion: you do not need to publish “how to grow carrots”, “how to grow carrots on a windowsill” and “how exactly to grow carrots specifically on a windowsill” as three separate articles. You need a real question and a specific, useful answer. Surprise: content now has to be useful.
SEO is not dead.It is becoming less about ceremonial keyword placement and more about the quality of the answer itself.
What is actually worth doing
- Let search engines see the site. Pages must open, be indexable and connect through clear internal links. After launch, we add the site to Google Search Console.
- Write something only you can write. Concrete experience, numbers, limitations and conclusions are more useful than the eleventh polished summary of the same ten articles.
- Make the structure obvious. One clear title, logical subheadings, a direct answer near the top, lists where they help and normal language.
- Protect the reading experience. Mobile layout, speed, readable type, sensible navigation and no irritating overlays still matter.
- Check facts and keep the personal detail. In the age of AI slop, fresh information is the scarce part. A slightly crooked piece of original reporting can be worth more than 150,000 perfectly interchangeable articles.
- Measure the result. Google data lives in Search Console. It becomes our main observation point after launch.
What you do not need
You do not need a separate AI version of the site, a mysterious “AI schema” or one question repeated in ten grammatical outfits. Google does not name llms.txt as a requirement for its AI features. Its guidance still prioritises helpful, reliable, people-first content.
Structured data is useful when it accurately describes content people can see — the author, article, organisation or breadcrumbs. It is not a secret boarding pass for AI answers.
THE MINIMUM FOR AN EXPERT SITE
My working checklist
- One page answers one clear question or task.
- A short direct answer appears near the beginning.
- Then come the explanation, example and next step.
- The author, date and contact route are visible.
- Title, description and heading structure are correct.
- The page is in the sitemap and checked in Search Console.
What this means for this website
We do not need a factory of fifty interchangeable notes “about AI”. It is better to publish steadily around real client questions: how to build an agent for a process, when automation will not pay back, how to diagnose user acquisition or how to make a presentation that does not look like a template escaped from 2018.
Those articles demonstrate expertise, answer a real query and give search systems material worth citing. That is the sensible foundation for SEO, AEO and GEO — no incense required.