Ask ChatGPT to recommend a supplier in your category and it will name five or six companies without hesitation. Somebody is on that list. If it is not you, that is now a commercial problem, and until recently nobody could tell you how big.
The instinct is to type your category into a chat window, read the answer, and either panic or relax. That tells you almost nothing. Ask the same question twice and you get different names. Ask it from a different account and you get different names again.
So the first job is to stop treating it as a lookup and start treating it as measurement.
The method, plainly
Fix the questions. Write the questions a real buyer asks, not the ones you wish they asked. Who are the best suppliers of X in the UK. How much does X cost. X versus Y. Who should I call about Z. Then freeze the list. A panel that changes between runs measures nothing.
Cover more than one engine. ChatGPT, Gemini, Perplexity, Google's AI Mode and AI Overviews all answer differently and draw on different sources. Measuring one and calling it AI visibility is like measuring one newspaper and calling it coverage.
Run it repeatedly and record everything. Answers vary, so single runs are anecdotes. Capture every brand named and every source cited, per answer, per engine.
Weight by usage. These surfaces do not have equal audiences. A mention in the consumer ChatGPT app is worth more than one in a surface almost nobody uses. Pick weights, write them down, and publish them so the number can be argued with.
Score share of voice, then rank. Share of mentions across the panel, against every competitor that appears. Rank is the number that survives a board meeting.
What we found when we ran it properly
Resonance built this as a standing index for UK B2B technology PR. The first edition ran 86 prompts across six surfaces, which produced 584 scored answers, several thousand brand mentions and a few thousand citations.
Three findings were uncomfortable enough to be worth repeating.
Your own website barely features. Citations overwhelmingly went to third-party sources: directories, roundups, trade press, comparison articles. One single roundup URL earned more citations across the panel than most agencies' entire domains.
Being good is not the mechanism. The brands that came top were the ones with a category page, consistent third-party descriptions and a presence on the handful of roundups that engines actually cite.
Pricing questions are a blind spot. On the prompts about cost, most brands were never mentioned at all, because almost nobody publishes a straight answer about what things cost.
What moves the number
In order of impact, from what we have watched since.
A page that answers the category question in the category's words. Not your product page. The page that answers "who does X in the UK" with a real answer, including competitors.
Corroboration you do not own. Directory profiles that agree with each other. Roundups that include you. Trade coverage that describes you the same way twice. This is the PR job and it is the slow part.
One description of your business, used everywhere. If your LinkedIn, your directory profiles, your schema and your homepage describe you four different ways, you are asking a machine to guess which is right.
Answering the cost question. The cheapest win available to most B2B companies, and the one most of them refuse.
What does not move it
Stuffing keywords, adding FAQ schema to everything, publishing more posts about the same thing, or paying for an AI visibility dashboard that gives you a score with no method attached. If a tool cannot show you the prompts, the surfaces and the citations, it is giving you a feeling.
Resonance publishes the panel, the weights and the method rather than a proprietary number. The work is here, and the PR practice is where the corroboration half gets done.
