AI Visibility Monitor

AI Visibility Monitor

The AI Visibility Monitor asks ChatGPT, Claude and Gemini the questions your customers ask, on a schedule, and records whether each answer cites your site, how early it mentions you, and which competitors appear instead. Site monitoring checks whether your pages can be cited; this monitor checks whether they are.

Set up a monitor

1. 01Open your project and click the AI Visibility card, then New monitor (or Create your first monitor).
2. 02Assistants: name the monitor, tick the assistants to query (ChatGPT, Claude, Gemini) and pick a Cadence: Weekly, Bi-weekly (the default) or Monthly. The form estimates the scans and assistant calls per month. An assistant marked Unavailable is not connected yet; contact us to enable it.
3. 03Prompts: one per line, written the way your customers ask. Suggestions from your project are filled in for you to edit; adding an industry and target audience in project settings gives better ones.
4. 04Brand: check the pre-filled Brand domain, Brand aliases / names, Competitor domains and Competitor names (taken from your project and its competitors), add optional Alert recipients, then click Save monitor.
5. 05Open the monitor and click Run first scan to get your baseline score.

Plan. Creating a monitor and running a scan need a paid plan or an active trial, because every scan queries the assistants. Without one, both return 403 with code: "subscription_limit". Viewing, editing, pausing and deleting existing monitors is not plan-gated.

What the results mean

Each scan sends every prompt to every selected assistant, so a scan holds one answer per prompt per assistant. An answer the assistant failed to return is recorded as an error and left out of every number below; VectraSEO never fills the gap with an invented answer.

Counts as cited. On the monitor page you choose what a citation is. Domain match only (recommended), the default, counts an answer as cited only when one of its source links is on your brand domain (subdomains included) or your domain is written in the answer. Domain or name mention also counts your brand name in the answer text. It suits a unique brand name; a name that is also a common phrase matches answers that never meant you, and the page warns you when one of your names looks like that.

Citation heatmap. One row per prompt, one column per assistant. Click a cell to read the full answer and its sources.

• Improvement over time charts the score from your second completed scan onward. Its verdict, Improving, Declining or Flat, compares the first and latest scan in the chart; a change of 3 points or less reads as Flat.
• Lost citations · repair queue lists prompts from the latest completed scan where a competitor was cited and you were not, ranked by impact. Each one suggests a page to rewrite or create, why the competitor won, and a fix order. Export downloads the rewrite brief or copies it as Markdown.
• Alerts. When a scan's score is 10 or more points below the previous completed scan, each address in Alert recipients gets one email. A first scan never alerts, because there is nothing to compare it with.

Refresh cadence

A monitor scans every 7 days (Weekly), 14 days (Bi-weekly) or 30 days (Monthly). The first scheduled scan is one interval after you create the monitor, so use Scan now for an immediate baseline. A scheduled scan starts within the hour after it falls due. Scan now works at any time, one scan per monitor at a time: while one is running a second request returns 409. Pause stops scheduled scans until you Resume; Delete removes the monitor with its prompts, scans and results.

Over the API, every route is in the AI Visibility Monitor reference. Creating and editing take projects:write, starting a scan takes jobs:write, and a scan answers 202 with a job_id to poll. Read a scan's answers from GET …/scans/{scan_id}/results. GET /capabilities/visibility-assistants reports which assistants are available.