How do I track if my brand is being mentioned by ChatGPT or Perplexity? Our method
How we track brand mentions in ChatGPT and Perplexity for clients: a fixed prompt panel, daily runs, share of voice against rivals, and traffic attribution on top.
The question sounds simple, and the first answer is: ask the engines. We tell new clients to do exactly that for an afternoon, because it builds intuition. Then we show them why it does not scale, and what we do instead.
Why manual checks mislead
An AI answer is generated each time. Ask ChatGPT the same question three times and the brand list can change. Perplexity differs again. Location, wording and context all play a part. So a single "we were not mentioned" screenshot is weak evidence, and a single "we were mentioned" is too. What carries information is a rate over many prompts and many runs.
How we set up tracking
We start with the prompt set, not the tool. With each client we write prompts in four groups: category questions, comparison questions, problem questions and brand questions. We tag them by topic and intent, and where it matters we add personas and countries, because the buyer's role and market change the answer.
Then we load them into Promptwatch, which is where we run our client programs. Promptwatch runs the prompt panel daily per model, country and language. That daily rhythm is what turns noisy single answers into a usable trend. Prompt Explorer helps fill gaps in the set by suggesting prompts from Search Console, ranking keywords or People Also Ask, and Prompt Volumes shows search volumes and difficulty so we do not guess which prompts matter.
We also set up the Brand Book. Its analysis instructions decide what counts as a mention, which matters when a brand name is also an ordinary word or a competitor's product shares a name.
What we report
Four numbers go in every monthly report: Visibility Score (how often the brand shows up across the panel), Share of Voice against named competitors, Position in Answer and Sentiment, which Promptwatch scores per response from 1 to 100 with a weekly AI summary. We add an A/B period comparison so the client sees movement between two time windows, and a heatmap that compares any two dimensions, for instance brands against models, to show where ChatGPT and Perplexity disagree.
Promptwatch monitors ChatGPT Search, Google AI Overviews, Google AI Mode, Copilot, Gemini, Perplexity, Amazon Alexa and Meta AI through their real UIs. Claude, DeepSeek, Mistral and Grok are API models. We tell clients which is which, because it affects how closely a result matches what a person sees.
Mentions are only half the picture
A mention tells you the brand was named. It does not tell you why, or whether it brought anyone to the site. So we add three layers.
Citations show which pages the engines used as sources, including third-party pages. Agent Analytics shows whether AI crawlers such as OAI-SearchBot and PerplexityBot fetched the relevant pages and whether the fetch succeeded. Visitor analytics shows AI-referred visits and conversions through a lightweight script, the GTM template or the WordPress plugin.
With all four in view, we can explain a drop. If mentions fall and crawler errors rose the same week, we know where to look. If mentions are steady but visits dropped, the cause may be elsewhere.
A simple version you can run yourself
If you do not need an agency, start small. Write 20 prompts, run them weekly by hand in a spreadsheet, and record the brands named. When that gets tedious, Promptwatch's free Explore plan covers 10 prompts on ChatGPT only, which is enough to see where you stand. Paid plans start at $95 a month for Essential, with all active models on every paid plan.
What tracking cannot do
It samples. It cannot cover every question a person might ask, and it cannot make an engine mention you. We judge trends over weeks and say so in the report. That honesty is part of the job.