Why Some Brands Vanish From AI Search While Their Google SEO Looks Fine
Joshua Etheridge
A brand can sit near the top of Google and still be missing from the answer an AI assistant gives.
If that sounds backwards, it is one of the most common things we run into. A company has spent years on SEO. The rankings look healthy. Then someone asks ChatGPT or Gemini for a recommendation in their category, and the brand is nowhere, while a competitor with a thinner looking site gets named instead. Those two outcomes used to move together. They are quietly coming apart.
Ranking and being quoted are now two different games
Traditional SEO is a contest for a position on a page of links. You want to be the result someone clicks. AI search is a contest for inclusion in a written answer. The assistant reads across sources, decides which ones it trusts enough to repeat, and produces a short reply that names only a few of them. Most brands never make that shortlist. Showing up once for one phrasing is not the same as being named reliably when the same question is asked ten different ways.
We watched this play out in our own work. When AIO audited a sample of UK law firms across consumer ChatGPT and Google Gemini, every firm had a real web presence and respectable traditional visibility, yet not one reached even a moderate level of AI presence, and the typical firm was named in only about 15 of 50 prompts. Strong on Google, quiet in the answer. You can read that one here.
So what is the AI actually weighing
This is the part people get wrong in both directions, so I want to be careful. No, the assistant is not reading your analytics. It cannot see your bounce rate or your time on page directly. There is a good instinct in the idea that behaviour and intent matter more than they used to, but the mechanism is not that the model is peeking at your dashboards.
What is closer to the truth is simpler. The things that make a human engage with a page are usually the same things that make a page easy for a model to quote. A page that answers a real question cleanly, in plain language, with the point stated up front, is both the page people stay on and the page an assistant can lift a line from. Clarity is the shared currency. Links still count as a credibility signal, but they are no longer the headline act.
From what we actually measure, the brands that get named consistently tend to share a profile. They are described the same way wherever they appear, so a model can tell they are one real entity rather than a fuzzy guess. They publish specific, quotable material instead of vague positioning. They turn up in the third party places a model already leans on, like reputable directories and genuine editorial coverage. And their pages carry structured data that spells out what they are. To be clear, that is a pattern we observe in the output, not a claim about the inner workings of any model.
The two playbooks, side by side
| Question | Optimising to rank on Google | Optimising to be quoted by AI |
|---|---|---|
| What you compete for | A high position in the list of links | Inclusion in the written answer |
| The signal that moves it | Backlinks and on-page keywords | Entity clarity and corroboration from trusted sources |
| What winning content is | Long, comprehensive pages | Specific, quotable claims a model can repeat |
| How you measure success | Rank and clicks | How often you are named across many real prompts |
One column is not replacing the other. A lot of the groundwork overlaps, and a strong organic presence still helps. The mistake is assuming that doing the first column well automatically earns you the second.
The question worth asking now
The most useful reframe I have heard recently is to stop asking whether you rank number one and start asking whether an AI would trust this page enough to use it in an answer. That single change rewrites the brief. You stop writing for a crawler that ranks and start writing for a reader, human or machine, that has to be able to repeat you accurately and without hedging. If your page cannot be quoted in one clean sentence, an assistant will quote someone whose page can.
What to actually do about it
- Make every page that matters answer one clear question, with the answer in the opening line rather than buried under preamble.
- Get described the same way everywhere you appear, so models read you as a single, real entity.
- Earn mentions in the sources AI tools already pull from, not just any backlink you can get.
- Add the structured data that tells a machine exactly what you are, so it does not have to guess.
- Measure the right thing. Track how often you are named across a battery of real questions, not where you sit in a ranking.
None of this means traditional SEO was wrong. It means the finish line moved. The brands pulling ahead in AI answers are not always the biggest or the oldest. They are the ones who noticed the goal changed and started building to be quoted rather than just to rank.
Frequently asked questions
Why does my site rank well on Google but never appear in AI answers?
Because ranking and being quoted are different outcomes. Google ranking rewards your position in a list of links. AI assistants build a short answer from sources they can read clearly and trust, which leans on clear entity signals and credible third party mentions as much as on rankings.
Does AI search look at my website analytics or user behaviour?
Not directly. An assistant does not read your bounce rate or time on page. What overlaps is that the same clarity and specificity that keep real users on a page also make that page easy for a model to quote, so good engagement and good citability tend to travel together.
Are backlinks still useful for AI visibility?
Yes, as one credibility signal, but no longer the main event. Being described consistently across the web and showing up in the sources a model already trusts now carry a lot of the weight. Structured data that makes your pages legible to a machine matters too.
How do I know if AI search can see my business at all?
Run the questions your customers would ask an AI assistant and record whether you are named, how often, and who gets named instead. That citation rate across a range of prompts is the honest measure. AIO's free audit does exactly this across the major assistants.
Want to see where you actually stand? AIO's free audit shows how often ChatGPT and Gemini name your business across the questions your customers are really asking, and where the gaps are. It takes about two minutes. Run your free AIO audit and find out whether AI search can see you at all.