How do you measure AI search visibility? Run the same 15–30 prompts a real customer would type, across ChatGPT, Perplexity, and Gemini, several times each, and log how often your business gets mentioned. **There's no stable "rank" to chase in AI search, but mention frequency across repeated runs is a real, trackable number — and you don't need a paid tool to start measuring it.**
Why Doesn't My Business Have a Rank in ChatGPT?
Because ChatGPT isn't pulling from a fixed, cached leaderboard — it generates a new answer each time. SparkToro ran 2,961 combined prompt runs across ChatGPT, Claude, and Google's AI, using 600 volunteers testing 12 different "recommend a brand" prompts over November and December 2025. The result: less than a 1-in-100 chance that ChatGPT or Google's AI returns the same list of brands twice for an identical prompt, and roughly 1 in 1,000 that two runs return that list in the same order (SparkToro, 2026). Asking "where do I rank in ChatGPT" is the wrong question — there isn't a fixed answer to find.
So What Can You Actually Measure?
Appearance frequency, not rank. SparkToro's own data shows this holds even with the randomness: a digital-marketing firm appeared in 85 of 95 responses, a cancer hospital appeared in 69 of 71 (a 97% visibility rate), and leading headphone brands showed up in 55–77% of responses regardless of how the prompt was phrased (SparkToro, 2026). That stability is what you're measuring — a mention rate, tracked across enough prompt runs to average out the noise in any single answer. LLM Pulse's formula for it is simple: mention rate equals the number of responses naming your business divided by total prompt runs, times 100; share of voice swaps the denominator for your mentions plus a named competitor's mentions, so you see relative standing, not just an absolute score (LLM Pulse, 2026).
How Do You Run a DIY Prompt Audit With No Paid Tool?
Write 15–25 prompts the way a real customer would actually type them, then run each one 3–5 times on ChatGPT and Perplexity, logging whether your business appears, where, which competitors show up alongside it, and which sources get cited (Passionfruit, 2026). A Kalispell roofer, for example, might run "best roofing company in Kalispell," "who do I call for hail damage near Whitefish," and "roofing contractor Flathead Valley" — each one three or four times per platform — and drop the results into a spreadsheet with a date column. Repeat monthly and the mention rate becomes a trend line, which is the only number that matters here: whether it's climbing or flat, not what it reads on any single day.
| Method | Cost | Platforms tracked | What it actually tells you |
|---|---|---|---|
| Manual prompt spreadsheet | $0 | Whatever you test by hand — typically ChatGPT + Perplexity | Your own mention rate on your own prompt list; no history, no automation |
| Ahrefs Brand Radar | Free with an Ahrefs subscription | 6 platforms, drawing on a 300M+ prompt database (Passionfruit, 2026) | Mentions and citations tracked automatically, no manual logging |
| Otterly.ai | $50–$200/mo (Passionfruit, 2026) | 6 platforms including Google AI Overviews and AI Mode | Mentions, citations, competitor presence, plus a GEO audit |
| Passionfruit Labs | $19/mo ChatGPT only, up to $99/mo with Gemini, Perplexity, and Claude added (Passionfruit, 2026) | Scales from ChatGPT to 4 platforms | Page-level citations and revenue attribution |
When Is a Paid Tool Worth It — and When Is DIY Enough?
Honestly: if you're a single-location business checking a handful of prompts once a month, a spreadsheet is enough — it's the exact same method the paid tools automate, just without the history and the daily refresh. A paid tool starts earning its cost once you're tracking several named competitors, more than a couple dozen prompts, or want weekly rather than monthly data — LLM Pulse's own cadence guidance is that daily tracking is "noisy," monthly "loses signal," and weekly is the sweet spot most businesses should run at, with at least 3–5 named competitors in the set so the number means "share," not just raw visibility (LLM Pulse, 2026). Below that scale, paying monthly for automation of a task that takes twenty minutes by hand isn't worth it yet.
What Should You Do With the Mention Numbers Once You Have Them?
Use them to find out why you're missing, not just that you are. Ahrefs' analysis of 1.4 million ChatGPT prompts and the citations behind them found that 88.46% of citations trace back to standard search results, versus 1.93% from Reddit, that the median cited page is roughly 500 days old, and that pages with a natural-language URL slug get cited 89.78% of the time versus 81.11% for pages without one (Ahrefs, 2026). None of that is fixed by tracking harder — it's fixed by making the pages behind your business more citable: current, clearly titled, and written to actually answer the question a prompt is asking, not just to rank for a keyword.