Who AI names instead of you

That is the whole problem in one line. A partner who does not already know your name now asks ChatGPT or Perplexity before they ask a referrer. The engine returns a shortlist. If you are not on it, you never enter the conversation. The RFP, the beauty parade, the informal “who do you use” — all of that happens after the shortlist. Miss the shortlist and the rest of your marketing never gets a chance to work.

Most practices only appear when the prompt already contains their name. Ask for “Smith & Co accountants Leeds” and the model will often surface Smith & Co. That is brand recall, not recommendation. It proves the model has heard of you. It does not prove a serious buyer will find you when they type the category instead.

The useful question is what the model says when the prompt is the job, not the brand. Best accountant for a £10m UK firm. FD support in the Midlands. Tax partner for an owner-managed business. Audit-ready finance for a PE-backed roll-up. Those are the questions that decide whether you are in the room. They are also the questions most firms have never locked, never re-run, and never written down against a competitor set.

Someone is named in those answers. It is usually a firm with a clearer entity — a name, places, and services the model can treat as one thing — plus pages a model can quote, and third-party pages that repeat the same claim in plain language. That is not mystical. It is the same pattern you already know from search: entities that are easy to cite get cited. The difference is that generative answers are shorter. The slot is finite. Another practice is already sitting in it.

Guessing which firm that is from one ChatGPT screenshot is not a baseline. Anecdote is cheap. A locked prompt set, re-run on a fixed date, against the engines you can actually score, is the unit of work. Without that, every partner meeting becomes a story about what someone saw on their phone last Tuesday.

A Geo Audit is how Geolab Studio locks that set. We write the buyer prompts in the language a serious buyer actually types. We run the engines we can score. We write down who is named instead of you — by prompt, not by vibe — and the order of work that would change it. Not a content burst. Not a software handover. A two-to-four-week diagnostic that ends in a decision: where you are absent, who occupies the slot, and what to fix first.

If you sell to mid-market UK firms, this is already happening without you. The buyer who used to call two peers now asks an engine first. Your website, your LinkedIn, your thought leadership — none of it matters if the shortlist never includes your name.

Start with the prompts a buyer would type if they had never heard of you. Then look at who the engines name. That list is your real competitive set in generative search — often different from the set you track in Google.