Do AI assistants recommend your firm?
For accountancy and finance firms, and the companies that sell to them.
Your buyers now ask ChatGPT, Claude, Gemini and Perplexity who to use. The answers name two or three firms. If yours isn't one of them, you lose the deal without ever knowing it existed. Check your domain below. Free, about a minute.
Prefer to read first? How the work runs
The run
A UK accountancy practice, top 30 by size · August 2026
- ChatGPT
- 0/6
- Claude
- 0/6
- Gemini
- 0/6
- Google AI Overviews
- 1/6
- Perplexity
- 1/6
- All engines
- 2/30
Grant Thornton was named in 10 of 30.
Buyers used to shortlist from Google results. Now ChatGPT narrows the list for them.
When someone needs a new accountant, a payments provider or an audit tool, more of them now ask an AI assistant instead of searching. They describe their situation and their budget, then ask who can help. Back comes a handful of firms. That's the new shortlist. If someone clicks through, it lands in your analytics as a referral. Most of the time nobody clicks: your name is read out, the reader has what they needed, and there is nothing to see at your end at all. If you are not on the list, you never find out.
Named in 2 of 30 answers
Across the questions its buyers actually ask, how many answers from each engine mention it at all?
Answers naming the firm, per engine. Figures unedited.
Who gets named instead
When the engines answer those same questions, which firms do they reach for?
Answers naming each firm, out of 30. The subject is the row in full ink.
ChatGPT0/6Claude0/6Gemini0/6Google AI Overviews1/6Perplexity1/6
The biggest brands always get named. The problem is losing to your real competitors.
AI engines lean toward big, famous companies, so household names in these answers prove little: they're rarely competing for the same clients. The example below shows the real problem. When this firm's buyers asked their questions, the engines kept recommending its direct rivals, and almost never the firm itself. Each of those answers was a potential client it never got in front of.
What the run measured
A UK accountancy practice, top 30 by size. August 2026. Every figure came out of it.
2 of 30
answers named this firm
1
cited it as a source
8 of 9
of its own named rivals the engines surfaced
5
engines tested, on the questions its buyers ask
How an answer is built
When buyers ask who to use, ChatGPT searches the web before it answers.
- 01The questionSomeone types a whole question, the way they would ask a person: what they need, their budget, their situation. Not the two or three words they would put into Google.Outside your control
- 02A live searchFor questions like these, the AI searches the web at that moment rather than relying on what it already knows. It goes out and reads real pages.Outside your control
- 03Sources retrievedThe AI picks a handful of pages to read. Whether yours makes the cut depends on how easily it can find and read your pages.Your pages decide this
- 04The answer writtenThe AI answers the question using those pages. Clear, specific writing gets quoted; vague writing gets passed over, even when it was read.Your pages decide this
- 05Citations attachedThe finished answer recommends some companies by name, and links a few pages as its sources. Those are two different wins, and you can get one without the other.Outside your control
Notice steps 03 and 04: they depend entirely on your website, and they are marked that way above. On buyer questions, the AI reads your pages at the moment someone asks rather than from anything it learned months ago. That is why you can fix this, rather than wait it out.
Check your firm in about a minute.
Free, no sign-up. You get a simple count for each chat engine: how many answers named you.
- Enter your domain
- That's all it needs. It reads only your public pages: no logins, no analytics access, nothing about your customers.
- It works out what you sell
- And who you compete with, from your own pages. You can correct anything it gets wrong before the run.
- It finds your buyers' real questions
- Questions people already put to the chat engines, on topics they genuinely search for. If there aren't enough real ones, we add likely questions ourselves, clearly marked.
- It asks the five engines, live
- ChatGPT, Claude, Gemini, Perplexity and Google's AI Overviews, the same way a real buyer would.
- You see who wins each answer
- Which answers named you, who they named instead, and whose websites they linked to as their sources.
If you're missing from the answers, the audit tells you why.
There are a few different reasons a firm goes unnamed, and each needs a different fix. Maybe the AI can't read your pages at all. Maybe it reads them and finds nothing worth repeating. The audit tells you which reason applies to you and what would fix it. Do that work yourself, hand it to your agency, or hire me.
This work has a name
Getting named in AI answers is called generative engine optimisation, or GEO. The answer side of it is answer engine optimisation, AEO. Some people call the whole thing LLM SEO. They mean much the same: making your pages easy for an assistant to read, and worth quoting when it does. We do that work for accountancy and finance firms.
Your run tells you which one you need
- Absent from the answers
- Four different things can keep you out, and each needs a different fix. The first job is finding which one applies to you. Fixed fee.
- Named, but never cited
- The engine answers from what it already believes and never reads your site. You need pages it has a reason to go and fetch. Monthly.
- Cited, but not named
- Your pages build the answer and a competitor gets the introduction. The fix is saying plainly who you are for, not just the facts a model can lift. Monthly.
- Named and cited
- You hold the answer. The work is keeping it: the same questions re-run every week on the same engines, so displacement shows up while it's still small.
See where you stand.
The same check the accountancy firm got, on your domain. About a minute. Costs nothing.
James Finlayson, AI visibility engineer