What AI Tools Say About Your Firm, Measured as Three Numbers, Not One
An AI visibility audit checks whether your firm is one of the handful of names ChatGPT gives when a trader asks which prop firm to trade with. Most firms have never checked whether they are in that handful. Most of the ones who have checked are reading the wrong number.
Real Search Console positions for this page, 16 months to September 2026. Low volume so far, which is expected for a service this new.
volume still building
AI visibility is three numbers, not one
Citation volume, mention rate and AI visits count different events. A firm can be cited often and never named. A firm can be named constantly and get almost no visits. Treating the three as the same thing is the most common mistake in this area, and it is why two vendors can quote you wildly different figures without either of them lying.
| Number | What it counts | Where it comes from | What it cannot tell you |
|---|---|---|---|
| Citation volume | Times an AI answer used one of your pages as a source | Search Console AI report, Bing AI Performance | Whether your firm was named, which question triggered it, or whether anyone clicked |
| Mention rate | Share of a fixed question set where the answer names your firm | The questions run against each tool, repeated on a schedule | How many real buyers ask those questions, or in what words |
| AI visits | Visits that arrive from an AI surface | Your analytics and server logs | The true total, because a lot of AI traffic arrives with no referrer attached |
I run this on my own site before selling it to anyone. Readings current as of 26 August 2026.
| Source | Reading | Window |
|---|---|---|
| Search Console, AI features | 16,700 impressions across 57 pages | Since 18 May 2026, when the report starts |
| Search Console, web search | 270,938 impressions, 1,283 clicks, 0.5% CTR, avg position 14.6 | Six months to 24 Aug 2026 |
| Bing AI Performance | 14,000 citations, 5 cited pages per day on average | Six months to 26 Aug 2026 |
| Bing Webmaster Tools, web search | 16,400 impressions, 343 clicks, 2.1% CTR | Six months to 26 Aug 2026 |
| The questions run across tools | 106 answers citing 11 pages: AI Mode 26, Perplexity 24, AI Overviews 23, ChatGPT 19, Copilot 9, Gemini 5 | Latest run |
Read those rows next to each other and the problem is obvious. Google reports 16,700 AI impressions across 57 pages. Running the questions by hand finds 23 AI Overview answers across 5 pages. Both are correct. They count different events, and no dashboard reconciles them for you.
The Bing pair is the more useful comparison for a prop firm or broker. Copilot and the Bing AI answers cited my pages 14,000 times over the same six months in which Bing web search showed them 16,400 times. Citation volume is running at roughly 85 percent of normal impression volume on that engine. If you still treat Bing as a rounding error because of its search share, that ratio says otherwise.
What an AI visibility audit checks
Being clearly defined
Consistent entity signals across your site, About page and third-party profiles, so models can pin down what your firm actually is.
entity SEO →Answers near the top
A self-contained answer high on the page, so a model can lift it cleanly instead of skipping past it.
Specifics
Numbers, rules and named facts. Vague claims do not get quoted; specific ones do.
Markup that matches
Schema that reflects what the page actually says, not aspirational markup layered on top.
Being described the same way elsewhere
Wording that stays consistent across the sites and directories a model is likely to pull from.
Fresh facts
Stale facts on widely cited third-party pages get repeated by models for months after you fix your own site.
A Google search quality rater, checking the systems that replaced the results page
I worked as a Google search quality rater for close to ten years before moving into consulting, and I now do SEO and content for prop firms, brokers, and the vendors selling into them. The rater background is why I care about the quality of the sources in these answers, not just the mention count.
See the record and two rankings you can check yourself →How the audit runs
We agree the questions
Ten to fifteen, the ones a buyer types right before choosing, not vanity questions about your brand name. You approve the list before anything runs. It stays fixed after that, because a question set that changes between runs gives you a trend line that means nothing.
I pull the citation data, then run the questions by hand
The Search Console AI report and Bing AI Performance report, exported by page, read against clicks rather than on its own. Then the same wording run across the tools, recorded per tool, with the cited sources captured. This part is done by hand, because the automated tools sample inconsistently.
You get the report and the call
What came back, what is measured against what is worked out, and the fixes ranked by likely effect. Some of those are on your site. The most useful ones usually are not, because the pages shaping your AI answers usually are not yours.
What people ask before booking
A written report with three measured numbers. How often AI answers cite your pages, how often your firm gets named in answers to a fixed set of buyer questions, and how many visits arrive from AI surfaces. It also lists the competitors named in your place, the sources the models lean on, and every wrong fact found in those sources. A walkthrough call is included.
I'll run three of your buyer questions live and show you what comes back
Who gets named instead of you, and we scope the audit from there. If your firm already looks strong in the answers, I will say so and we can stop.