How to track brand mentions in ChatGPT, Perplexity, Gemini, Claude, Grok and DeepSeek

By Get AI Vizible · Updated

Short answer

To track brand mentions in AI assistants, pick a fixed set of buyer questions, ask each one on every engine you care about on a regular schedule, and record whether your brand and each competitor is mentioned, in what order, which sites are cited and what is said. You can do it by hand in a spreadsheet for a handful of questions; for more questions, engines and competitors, an AI visibility tool such as Get AI Vizible runs the questions through the engines' APIs and calculates the metrics for you.

What to record for every answer

  • Engine and model
  • Question
  • Your brand: mentioned (yes/no), position among brands named
  • Each competitor: mentioned, position
  • URLs cited, and whether your domain is one of them
  • Anything inaccurate about you
  • Tone: positive, neutral or negative

The manual method

  1. Choose 10 to 20 questions.
  2. Open each engine in a private window, logged out where possible, so personal history does not skew results.
  3. Ask each question and paste the answer into a sheet with the fields above.
  4. Repeat monthly on the same date.

This works for a first look. It gets slow quickly: 20 questions on 6 engines is 120 answers per round.

The automated method

An AI visibility tool sends the questions through each engine's API on a schedule and parses the answers. With Get AI Vizible:

  • six engines: ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek
  • automatic scans weekly (Starter), every 4 days (Growth) or on a schedule you choose per brand (Agency)
  • visibility, position, share of voice, citation rate and sentiment, per engine and per question
  • FactCheck that flags wrong claims about your brand
  • the sources each engine cites, and which ones to target
  • trend charts, CSV exports and PDF reports

How often to track

Weekly is enough for most brands: AI answers move with content changes, new reviews and model updates, not hour by hour. Track more often around launches, pricing changes or PR events.

Reading the results

  • Look at trends over several scans, not single answers.
  • Compare engines: a gap on one engine points to that engine's sources.
  • Read the actual answers for questions you lose; they show who is named instead and why.

See AI visibility metrics explained for the formulas.

Frequently asked questions

Do API answers match what users see in the ChatGPT app?
They are close but not identical: the consumer apps may add personalisation, memory and different search settings. API tracking gives a consistent, repeatable baseline, which is what you need to measure change over time.
How many questions should I track?
Start with 20 to 50 questions that map to revenue: category, comparison, location and problem questions. Add more once you act on the first set.

Related