# How do I track whether AI assistants mention my business? > AI visibility has no rank tracker. Here is a repeatable method for measuring whether ChatGPT and Perplexity name your business, using a fixed prompt set. **Source:** https://parvaly.com/answers/how-do-i-track-whether-ai-assistants-mention-my-business/ **Author:** Sergey Kalashnikov, Founder, Parvaly **Published:** 2026-08-23 **Topics:** ChatGPT, Perplexity, Google AI Overviews, Generative Engine Optimization ## Short answer Measuring AI visibility means building your own instrument. Write 25 to 50 prompts a real customer would type, covering category, city, neighborhoods, price questions and comparisons. Run them monthly across ChatGPT, Perplexity and Google AI Overviews. For each, record whether you appeared, in what position within the answer, which competitors appeared, and which sources were cited. The single number worth reporting is presence rate: the share of prompts where your business was named. ## Key takeaways - There are no rankings in AI answers. The honest metric is presence rate across a fixed prompt set. - Answers are non-deterministic, so a single run proves nothing; the trend across a repeated set is the signal. - Log the sources each assistant cites. That list is a ready-made worklist of where to get mentioned. - Run the same prompts against competitors' categories too, so you know who you are actually losing to. - Monthly is the right cadence for a small business. Weekly produces noise, quarterly misses the movement. ## Why this needs a method rather than a tool Rank tracking works because a search result page is stable enough to photograph. An AI answer is not. Ask twice, get two different responses; ask from a different account, get a third. Any single observation is noise. The fix is the same one used everywhere else measurement is noisy: fix the instrument, repeat it, and read the trend rather than the reading. ## Building the prompt set Aim for 25 prompts for a single-location business, 50 for a competitive market or multiple services. Write them the way a customer speaks, not the way a keyword tool does. **Category and city.** "Best emergency plumber in Sacramento." "Who should I call for a burst pipe in Sacramento?" **Neighborhood level.** "Plumber near Land Park." Assistants handle sub-city geography better than most people expect, and competition thins out there. **Problem-first.** "My water heater is leaking, who do I call in Sacramento?" This is how people actually ask, and it is where generic category pages fail. **Price and process.** "How much does water heater replacement cost in Sacramento?" These pull in businesses that publish real answers about pricing. **Comparison and qualification.** "Which Sacramento plumbers offer same-day service?" "Best-reviewed plumber in Sacramento." Freeze the list. Changing the prompts between runs destroys comparability, which is the only thing you are buying. ## Running it Once a month, in a fresh session, logged out where the platform allows it. Cover ChatGPT, Perplexity and Google AI Overviews at minimum; add Claude and Copilot if your market warrants it. For each prompt, record five things: whether your business was named, where in the answer it appeared, which competitors were named, which sources the assistant cited, and the date. A spreadsheet with one row per prompt per platform per month is entirely sufficient. ## Reading the results **Presence rate** is the headline: named in how many of the prompts, as a percentage. This is the number to report and to watch. **Competitor frequency** tells you who your real rivals are in this channel, which is often not who you assumed. **Cited sources** is the most actionable column. Every publication, directory and roundup that gets cited is somewhere you could be present and are not. Over a few months this column writes your outreach plan for you. **Prompt-type gaps** show where the work is missing. Strong on category prompts and absent on price prompts means you have not published anything about pricing. ## What good looks like A business starting from near-zero and doing the underlying work, meaning clean listings, extractable content, steady reviews and a few real third-party mentions, typically starts seeing movement in presence rate somewhere in month two to four. Not a jump to dominance; a shift from being named in two prompts out of twenty-five to being named in eight. That is a real result, it is measurable, and it does not require anyone to promise you something they cannot control. ## Frequently asked questions ### Are the paid AI visibility tracking tools worth it? For a single-location business, usually not at first. A spreadsheet and an hour a month produce the same decision-quality information for the size of prompt set you need. Tools start earning their cost at multi-location scale, or when you need historical data across dozens of markets. ### Why do I get different answers each time I ask? Assistants are non-deterministic and the live search results behind them shift constantly. Personalization and memory of your past chats can also skew what you see, which is why measurement should be done in a fresh session, logged out where possible. ### What counts as appearing in an answer? Being named in the body of the response is what counts. A citation link at the bottom without a mention in the text is worth logging separately. It means the model read you but did not recommend you, which is a different problem with a different fix. ## Sources 1. [Local Consumer Review Survey 2026](https://www.brightlocal.com/research/local-consumer-review-survey/) · BrightLocal, 2026 --- © 2026 Parvaly LLC. https://parvaly.com