AI search optimisation exists because something has quietly changed about how people research a purchase. They used to type a question into Google, open four tabs, and form an opinion. Increasingly they ask an AI assistant and act on the paragraph it gives back.
That paragraph names two or three businesses. If you are not one of them, the enquiry never existed as far as you are concerned — there is no impression, no click, nothing in your analytics to mourn. This is the part that makes it dangerous: you cannot see the traffic you are losing.
Why AI answers are not just rankings with extra steps
Ranking first on Google means you won a page. Being cited in an AI answer means the model decided you were the most reliable way to answer a specific question. Those are different games.
Google ranks a page. A model assembles an answer from many sources — your site, a directory, a forum thread, a news piece, a comparison article on someone else's blog. Your own website is one input among many, and often not the loudest one.
What actually seems to influence it
1. Being described consistently, everywhere
Models build a picture of your business from every mention of it. If your site says performance marketing, your LinkedIn says growth consulting and a directory lists you as a web design studio, that picture is blurry — and a blurry entity does not get cited confidently.
Pick one description of what you do and repeat it, word for word, across your site, your profiles and every listing. This is unglamorous and it matters more than most technical work.
2. Content shaped like an answer
Pages that get quoted tend to state the answer plainly, early, and then justify it. A page that opens with three paragraphs of throat-clearing before reaching the point gives a model nothing clean to lift.
- Put a direct, one-sentence answer immediately under each heading.
- Use headings that are the actual question someone would ask.
- Give specific numbers, ranges and conditions — vague claims are unquotable.
3. Being talked about somewhere other than your own site
This is the uncomfortable one. A model weighs what independent sources say about you more heavily than what you say about yourself. Reviews, forum answers, industry roundups, a podcast transcript, a comparison post — these are what turn a claim into a fact from the model's point of view. It is the same shift that makes brand mentions matter more than raw backlinks.
4. Technical access
Some assistants fetch live pages. If yours is slow, blocks their crawler, or hides the substance behind JavaScript that only renders after interaction, you are invisible to that path. Check your robots rules and make sure the important text exists in the HTML.
What is largely a waste of time in AI search optimisation
- Keyword stuffing for AI. There is no keyword density trick here. The model is reading meaning, not counting words.
- Publishing volume for its own sake. Forty thin pages give a model forty weak signals. One genuinely definitive page beats them.
- Chasing every new acronym. AEO, GEO, LLMO — the labels keep changing. The underlying work has not: be clear, be specific, be corroborated.
The businesses getting cited are mostly the ones that were already easy to understand. AI just raised the penalty for being vague.
How to check where you stand
There is no rank tracker for this yet that we would trust, so do it manually and honestly:
- Write down the ten questions a customer asks right before buying from you.
- Ask each one in ChatGPT, Gemini and Google's AI Overview. Use a fresh session with no memory of you.
- Note who gets named. Usually it will not be you — note who it is and where the model got them from.
- Go and get mentioned in those places.
Repeat monthly. It is crude, it is manual, and it is still more useful than any dashboard currently on the market.
The honest caveat
Nobody can guarantee a citation in an AI answer, and anyone selling you that certainty is guessing. What you can do is make your business the easiest, clearest, best-corroborated answer to a specific question — which happens to be the same thing that has always worked in search, just with less room to fake it.
Questions people ask
What is AI search optimisation?
Making a site easy for an answer engine to extract and cite, rather than only easy for a search engine to rank. In practice that means self-contained answers, clear structure, structured data, consistent entity details and being mentioned in sources the model already trusts.
Is it different from SEO?
It overlaps heavily but the unit is different. Search optimises a page for a ranking; answer engines lift a sentence or two out of a page to use in an answer. Content written so a paragraph still makes sense on its own gets cited more.
How do I check whether I am being cited?
Ask the assistants the questions your buyers ask and see who is named. Then look for referral traffic from chatgpt.com, perplexity.ai, gemini.google.com and claude.ai in your analytics. Both are rough, and both are better than guessing.
What does not work?
Keyword stuffing for robots, mass thin pages, and anything that treats an answer engine as a ranking algorithm with different inputs. The models are trained on what is already credible, so manufactured signals age badly.
If you want a look at how your business currently gets described across the web, tell us what you sell and we will go and check.