SerpkaiSERPKAI

METHODOLOGY

How we measure your visibility in AI

AI visibility tools have a bad reputation for giving scores nobody can verify. Here we explain, with no small print, how Serpkai’s AI visibility works: what we ask, how we count and what its limits are.

In short

  • You choose the prompts your customer would ask; we run them every week on each AI.
  • We count whether your brand or your site appears in the answer or in its sources, and who appears instead.
  • We store and show you the real text of every answer.
  • We compare what the AI says with your brand facts to warn you about errors.
  • There is no opaque score: everything can be checked.

1. The prompts

Each project has its own prompts, written the way a person would ask them (“which rank tracker do you recommend for an agency?”). You write them in the dashboard. The same prompts are repeated every week, so the changes you see are changes in the AI, not in the measurement.

2. How we ask

  • With no history or signed-in session: every prompt starts from zero.
  • With the same prompts, written in your audience’s language.
  • On the same day of the week and at the same time.
  • On Pro plans and above, each prompt is asked 3 times on ChatGPT, Gemini and AI Mode, so you get a real appearance rate (for example, 2 of 3 answers).
  • Through each platform’s official interfaces, via our data provider.

3. What we count

  • Mention: your brand or your domain appears in the text.
  • Citation: your site appears among the linked sources.
  • Competitors: which other brands appear in the same answer.
  • Trend: how many of your prompts mention you week after week, per platform.

4. Errors about your brand

If you fill in your brand sheet, when an answer names you a second AI model compares it with your real facts and flags what does not match (an old price, a service you do not offer, the wrong area). You always see the original sentence so you can decide.

5. Limits worth knowing

  • AIs give different answers to each person: we measure a fixed, comparable sample, not what every user sees.
  • A single answer can be chance; what is reliable is the trend over several weeks.
  • Platforms change models without notice; you can annotate those changes on the chart to take them into account.
  • The error review is done by an AI: it is an aid, not a verdict.

Frequently asked questions

Why does ChatGPT give me different answers from the ones I see in Serpkai?

AI assistants generate each answer from scratch, and it changes with the moment, the model, the country and the history of whoever asks. Serpkai always asks with no history and with the same prompts, so the data is comparable week after week. That is why the trend matters more than a single answer.

What does it mean that my brand “appears”?

That your brand name or your domain shows up in the text of the answer or among the cited sources. We count it per prompt and per platform, and we show you the passage where it appears.

How often are the AIs measured?

Once a week, with the prompts you choose. AI answers change slowly, and measuring them every day would only add noise and cost.

How do you know if the AI says something false about my brand?

You fill in a sheet with your real facts (services, prices, areas). When an answer names your brand, a second model compares it with that sheet and flags what does not match. It is an aid for review, not an absolute truth: we always show you the original text.

Do you use an “AI visibility score”?

We do not give you a magic number. We show how many of your prompts mention you, on which platforms, which competitors appear instead and the real text of each answer, so you can check it yourself.

Which platforms are measured?

ChatGPT, Gemini, Perplexity, Claude, AI Mode and Google’s AI Overviews.

Want to see what AI says about your brand? Try it free, no card.

ES · Español