← Blog
#ai-visibility#how-to#geo

How Can I See What AI Says About My Company?

September 10, 2026

How Can I See What AI Says About My Company?

Open ChatGPT, Claude, Gemini and Perplexity, ask each one the questions your buyers ask, and write down whether your name appears and which sources the answer cites. That takes about twenty minutes and it is the correct first step. The catch is that a single run of a single question is not a measurement, and understanding why is what stops you drawing the wrong conclusion.

Table of Contents

The twenty minute manual version

  1. Write five buyer questions, not brand questions. "Best tools for X", "which vendor should a company like mine pick", "how do I solve Y". Asking "what is [your company]" tells you almost nothing useful, because the model will find you if you name yourself.
  2. Ask each question in all four assistants. Use a fresh chat each time so previous context does not leak in.
  3. Record three things per answer: were you named, who else was named, and which sources were cited.
  4. Repeat each question at least three times. This is the step everyone skips, and it is the one that matters.
  5. Turn off personalisation and memory where you can, or use a logged-out or temporary chat. Otherwise you are measuring what the model knows about you personally.

Why one run is not an answer

Assistants are probabilistic. The same question asked three times can name three different sets of companies. If you ask once and you appear, you might be in a coin-flip position rather than a stable one. A rate, such as named in seven of nine responses, is a fact you can act on. A single yes is an anecdote.

This is also why screenshots in vendor marketing should be treated carefully, including ours.

Where manual checking breaks down

  • Repeat runs. Five questions across four assistants with three runs each is sixty prompts. Once a week, by hand, forever.
  • Drift. Answers change as models update. Without a stable prompt set and a record, you cannot tell your improvement from their update.
  • Sources. Recording cited domains by hand across sixty answers is where the exercise usually dies, and the sources are the most useful part.
  • Competitors. Tracking who is named instead of you, over time, is a spreadsheet nobody maintains.

The automated version

Automated tools do exactly the above on a schedule and keep the history. If you want a fast, honest baseline without setting anything up, the free AI visibility check runs a sample of buyer prompts and returns a score with no signup. If you want to see what the full picture looks like before deciding, the live demo uses a real scan.

For comparison shopping across the category, see the AI visibility tools comparison.

FAQ

Should I ask the assistant "what do you know about my company"? It is a useful sanity check for accuracy, but it does not measure visibility. Buyers do not start with your name; that is the entire problem.

Does asking repeatedly train the model to mention me? No. Your prompts do not change what other people see.

Which assistant matters most? The ones your buyers use. For most B2B categories that means ChatGPT and Perplexity carry the most research traffic, with Claude and Gemini worth tracking for coverage.