AI visibility

What does AI visibility measure?

Serpel measures where an AI answer can mention your website: in ChatGPT with web search and in Google AI Overviews. For each prompt it records whether the answer links a page of your domain, which Serpel calls cited, whether your name or domain only appears in the text, which it calls mentioned, and which sources the answer lists. The AI visibility feature page gives the overview, and the page on AI Overview tracking covers the Google side.

Serpel doesn’t query Perplexity, Gemini, Claude or any other assistant. An AI answer can change from one run to the next, so one check is one sample. Compare several checks in the history of a prompt before you act on a single result. Rank checks also record whether a Google AI Overview appears for each tracked keyword, which the rank tracking guide describes.

How does Serpel decide whether you are cited or mentioned?

Serpel evaluates every answer with fixed rules. No model judges the answer. The rules look at the sources the answer lists and at its text.

The domain counts as mentioned when its text appears anywhere in the answer, ignoring case. Your project name counts when it has at least 4 characters, differs from your domain and appears as a whole word. If ChatGPT writes your brand differently from your project name, rename the project under Settings → General. Serpel stores the same matching for Google AI Overviews, but the dashboard shows only three results for them. Cited means that a source of the AI Overview links your domain. Not cited means that Google shows an AI Overview and your website is not among its sources. Not mentioned means that Google shows no AI Overview for the prompt.

How do prompts work?

A prompt is the question a customer would ask an AI assistant. Write it the way people ask, with a location, an occasion or a price range. A prompt needs 3 to 500 characters. Adding a prompt costs nothing and doesn’t start a check. When you add prompts in the dashboard, they use the country and language of the project. With the CLI or the API you can set a different country and language per prompt. How many prompts a project can hold depends on your plan:

1. 1

   Open AI visibility.

   In the sidebar, open AI visibility under Monitoring.

2. 2

   Click Prompt.

   The button at the top right opens the dialog Prompts for ChatGPT. On an empty page, the button reads Add prompts.

3. 3

   Enter your prompts.

   Type one prompt per line in Prompts, one per line. The dialog shows how many slots are free and warns about prompts shorter than 3 or longer than 500 characters. Prompts that already exist are skipped.

4. 4

   Add and check.

   Click Add prompt. With several lines, the button reads Add and the number, for example Add 3 prompts. The answers arrive with the next check. To remove a prompt, use the bin icon Remove prompt. This also deletes its history.

How do you run checks and monitoring?

To check one prompt, open it and click Check prompt. To check all prompts, press Ctrl+K or Cmd+K on AI visibility and choose Check prompts. Both show the credit estimate. The check runs in the background, a banner shows its progress, and only one check runs per project at a time.

Monitoring checks all prompts on your schedule, together with your tracked keywords, and then writes a report. You set the frequency and the time under Settings → Monitoring. A schedule needs at least one tracked keyword or one prompt. The monitoring and reports guide explains schedules, reports and notifications.

How do you read the AI visibility page?

Cited is the strongest result: the answer used a page of yours as a source. Mentioned means the answer names you but links no page of yours, so open Sources of the answer to see which pages it used instead. With Not mentioned, the same list shows who the assistant cites for that prompt. The badges Newly cited and No longer cited mark changes between two checks. Use Cited sources to see which domains ChatGPT links most often for your topics, and compare several checks before you draw conclusions from one.

What does an AI visibility check cost?

A check costs 2 credits per prompt for the ChatGPT answer. The Google AI Overview adds 2 credits per prompt, but only when Also check Google AI Overview is on. It is on by default and sits under Settings → Ranking check. A check of 10 prompts costs up to 20 credits with ChatGPT alone and up to 40 credits with both surfaces.

Serpel reserves the maximum when a check starts and charges only for results it delivered, so a failed check is free. If Serpel serves a recent stored answer for the same prompt, the prompt page marks the answer as served from the cache, and it is charged like any other delivered result. Reading results in the dashboard, with the CLI or over MCP is free. Credits and billing lists every price and explains reservations.

How do you use AI visibility from the CLI, API or MCP?

The CLI and the REST API cover everything above. Over MCP an agent can only read the stored results. The CLI adds prompts, runs checks and prints the summary. The REST API returns the same data as JSON, and its reference also lists the endpoints for a prompt’s history and for removing prompts. The MCP tool get_ai_visibility returns up to 25 prompts per call and never starts a check. The Claude Code guide shows how to connect an agent.

serpel ai add --project <id> --prompt "best espresso beans for home"
serpel ai run --project <id> --wait
serpel ai status --project <id>

What can go wrong?

Where to go next