We asked AI assistants 30 questions about our own industry. We scored zero.
By Alex Golombeck — The Happy Cat, Manchester. Published 20 August 2026.
The Happy Cat is an AI visibility (GEO) agency. We help UK businesses get found and recommended by AI assistants such as ChatGPT, Claude, Perplexity and Gemini.
So we started by testing ourselves.
On 19–20 August 2026 we ran a controlled baseline: ten fixed questions a real buyer might ask, put to three AI assistants — ChatGPT (with search enabled), Perplexity and Gemini — exactly as written, first answer only, no regenerating, everything screenshotted and logged.
Questions like "best AI visibility agency UK", "AI SEO agency Manchester", and "who can audit what AI says about my business".
The Happy Cat appeared in 0 of the 30 answers.
That's our baseline: 0%. We're publishing it — the number, the method, and everything we found — because this is the exact audit we run for clients, and because we think the honest starting line is more useful than a polished ending. We'll re-run the identical queries on 16 September and 14 October 2026 and publish those results too, whatever they say.
Here's what 30 logged answers taught us about how AI assistants actually recommend businesses.
Finding 1: Nobody owns this market yet
Across just 30 answers, the three assistants cited 104 different web domains as sources. Only eleven of those domains were cited more than once.
Read that again if you're a UK business owner: the sources AI assistants trust for recommendations in our industry are scattered across a hundred-plus sites, most cited exactly once. There is no entrenched incumbent monopolising the citations. Two agencies (Tilio and Rank4AI) appeared across all three assistants; everyone else showed up patchily, on one engine, for one question.
A fragmented citation landscape is the single most encouraging thing a new or overlooked business can find. It means the answers are still up for grabs.
Finding 2: Each AI assistant decides differently
This was the practical gold. The three assistants have distinct "personalities" in where they look before recommending anyone:
ChatGPT recommends from your own website. Almost every source it cited was an agency's own site, tagged with utm_source=chatgpt.com — ChatGPT literally labelling its own referral traffic. If ChatGPT can read your site and your site clearly says what you do, you can be recommended directly. If your site blocks AI crawlers or never states plainly what you are, you cannot.
Perplexity trusts third parties. It rarely cited an agency's own homepage. It cited comparison articles, industry roundups, blog posts, a LinkedIn article — and in one case a company that entered the answers via a press release on a newswire. Perplexity is won by being written about, not by writing about yourself.
Gemini quotes substantive articles. Its citations linked to specific passages inside detailed blog posts and service pages. Gemini rewards businesses that have actually written something worth quoting.
One optimisation strategy per engine, in other words — which is why "just do SEO" advice translates so poorly to AI visibility.
Finding 3: For "how do I…" questions, the winners are articles, not agencies
Ten of our thirty answers named no businesses at all. Questions like "how do I get my company recommended by ChatGPT" were answered with advice — and the citations went to whoever had written the clearest guide.
Some of those cited guides were published by very small firms. On the therapist-focused question, the citations went to niche specialists most people have never heard of, purely because they had written specifically about therapists and AI visibility.
The lesson: on informational questions, a well-aimed article from a small business beats a big agency's homepage. Specific beats big.
Finding 4: The comparison page is the new front door
Several agencies appeared in Perplexity's answers because they had published their own "best UK AI visibility agencies" comparison pages — rankings that include themselves.
We have mixed feelings about that pattern, and we'll be transparent about how we handle it: we're publishing our own comparison of UK AI visibility agencies, with our inclusion clearly disclosed, ranking criteria stated, and genuine links to competitors. Buyers deserve a real comparison. But the underlying fact is simply true: AI assistants cite comparison content heavily, and businesses that publish honest, useful comparisons get retrieved.
What we're doing about our zero
Everything above converts directly into our own plan — the same plan we build for clients:
- Make the site machine-readable (done before this baseline: AI crawlers explicitly welcomed in robots.txt, structured data describing exactly who we are, sitemap submitted to Google and Bing).
- Publish this audit — you're reading it.
- Answer the informational questions properly — dedicated, question-shaped pages for the sectors we work with.
- Earn third-party citations — the honest comparison page, industry roundups, directories, and yes, a press release.
- Re-run the identical 30 queries on 16 September and 14 October and publish the movement, or lack of it.
If the score is still zero in October, you'll read that here too.
The method, if you want to run this yourself
You don't need us to take your own baseline. Freeze 8–10 questions your real buyers would ask an AI assistant. Run them in ChatGPT (search on), Perplexity and Gemini in one sitting. Take the first answer only. Log who gets recommended, which sources get cited, and whether you appear. Screenshot everything and date it. That log is your before photo — and the cited sources are your to-do list.
If you'd rather we did it — with the same rigour you've just read — that's exactly what our AI visibility audit is.
Frequently asked questions
What is an AI visibility (GEO) audit?
A controlled test of what AI assistants like ChatGPT, Perplexity and Gemini say when buyers ask questions your business should be the answer to: who gets recommended, which sources the assistants cite, and whether your business appears. It produces a dated baseline score and a prioritised list of fixes.
Why did The Happy Cat score zero?
Because the baseline was taken deliberately early: days after our site first allowed AI crawlers to read it, before publishing supporting content or earning third-party citations. The zero is the honest starting line the follow-up measurements will be compared against.
How do AI assistants choose which businesses to recommend?
Differently per assistant. In our 30-query baseline, ChatGPT cited businesses' own websites directly, Perplexity cited third-party comparison articles and press coverage, and Gemini quoted passages from substantive blog and service pages. Visibility work has to cover all three routes.
Can a small business really show up in AI answers?
Yes — our baseline found citations going to very small, specific sites on informational questions, and 104 different domains cited across just 30 answers. The citation landscape is fragmented, which favours specific, well-structured content over sheer size.
When will the follow-up results be published?
Re-runs of the identical 30 queries are scheduled for 16 September and 14 October 2026, and the results will be published either way.