How to use AI Mention Monitor

AI Mention Monitor reads how ChatGPT and Google AI Overview answer the questions your customers ask. It works two ways: scan a topic to see the brands, sources and fan-out queries across the prompts about it, or check one exact prompt live to see whether the engine cites you — or a competitor instead.

Two modes: topic scan and check a prompt

The tool has two tabs, for two different questions:

Topic scan — across the questions people ask AI about a topic, how visible am I? Enter a short topic; it reads a cache of measured prompts and shows the brands named, the sources cited, and the queries the engine fans the seed out into. The volume picture.
Check a prompt — for this one exact question, does ChatGPT cite me? Paste a full question a customer might ask; it runs live and tells you whether you're cited, and who is instead.

Most of this guide covers the topic scan. Skip to Check a prompt for the exact-prompt mode.

Topic scan in 30 seconds

Pick an engine: ChatGPT (US English only at launch) or Google AI Overview (all 11 supported countries).
Type a short search-style query (e.g. best CRM, not What is the best CRM for small teams in 2026?).
Click Scan AI mentions and wait 10 to 20 seconds for results.
Read the three-arc spiral and the three detail tables below it.

How to phrase queries that return rich data

We match your seed against a cache of real prompt questions. Short, universal search-style phrases match the most prompts. Specific natural-language questions often return little or nothing.

Examples that work well:

best CRM
best noise-cancelling headphones
best SEO tool
how to start a podcast

Examples that match few or zero prompts:

What is the best CRM for SMB in 2026? — too specific
content marketing — too broad (matches every prompt that mentions the phrase)
best Italian restaurant Leeds UK — the city + country qualifier filters too aggressively

If a scan returns 0 results, your quota isn't charged. Try a shorter or more search-style phrasing.

Reading the spiral

The visualisation has three arcs radiating from the seed at the centre:

Top arc (teal): Brands — companies and products the engine named in its answers.
Right arc (yellow): Fan-outs — related sub-queries the engine expanded the seed into. Stem length encodes AI search volume on a log scale; zero-volume entries are model-internal noise.
Bottom-left arc (purple): Sources — real URLs the engine cited when composing answers.

Hover any node for the full label, mention count, and metadata. Each arc maps directly to one of the detail tables below the spiral.

ChatGPT vs Google AI Overview

The two engines have different strengths:

ChatGPT tends to surface more named brand entities (it answers conversationally). Locale: US English only at launch.
Google AI Overview tends to surface more cited sources (it's tied to real Google SERP results). Locale: all 11 DFS-supported countries.

Run the same query on both for a fuller picture. Each scan counts against your monthly quota.

AI overview

After a scan completes, click Generate AI overview for a 2 to 4 sentence summary of the patterns in the data. Useful for client reports or sharing with non-technical stakeholders.

Downloads

PNG exports the spiral with a DATA HIT logo + caption bar. Drops cleanly into client decks.
CSV exports all four data sections (brands, fan-outs, sources, demand) with scan metadata. Cited-source URLs are stripped of tracking parameters.

Check a prompt

A topic scan reads a cache of measured prompts, so a specific, long-tail question can come back empty. Check a prompt closes that gap: it sends your exact question to ChatGPT live and reads which sources the answer cites.

Open the Check a prompt tab.
Paste one exact question, up to 500 characters (e.g. What's the best tool for visualising GA4 seasonality?).
Click Check this prompt. It runs live against ChatGPT — usually a few seconds, occasionally up to a minute.

You get one of four reads, framed around who ChatGPT points to:

Cited — your domain is one of the sources the answer cited.
Not cited, a competitor is — ChatGPT cites someone else for this question, and names who. The headline signal.
No sources cited — ChatGPT answered without citing anyone.
Mentioned, not cited — your brand appears in the answer text but not as a linked source. A weaker signal than a citation.

Each result shows the answer excerpt and the full list of cited sources. It compares against your connected Search Console domain, so connect one to get the “are you cited?” read.

A check is a live snapshot — ChatGPT's answer varies between runs. Re-check the same prompt to see how stable a citation is. Empty or failed checks don't count against your quota.

Your AI visibility trend

The topic scan and prompt check are snapshots. The visibility trend is the time axis: on the Topic scan tab, Pro users get a Your AI visibility trend panel that charts how ChatGPT's mentions of your own domain have grown or shrunk, month by month, back to August 2025.

Connect Google Search Console and pick your property — the trend tracks that domain.
On the Topic scan tab, click Show trend.
Read the line: rising means you're gaining ground (more mentions gained than lost over time); falling means you're losing it. The headline calls it Gaining ground, Losing ground, or Holding steady.

Viewing your trend is free — it never uses a monthly AI Monitor action, so check it as often as you like. It refreshes once a day.

Quota and pricing

Topic scans and prompt checks draw from one shared monthly pool of AI Monitor actions:

Free: sample-data preview only (no live scans or checks).
Pro Solo (£19/mo): 50 AI Monitor actions per month.
Pro Agency (£49/mo): 250 AI Monitor actions per month.

Quota resets on the 1st of each UTC month. A cached topic scan on a previously-scanned (keyword, country, engine) combination doesn't count; every prompt check is a fresh live call, so each one counts (empty or failed checks don't). Viewing your AI visibility trend is always free and never counts against the pool.

What leaves your browser

For a topic scan, just the seed query, country, and engine — we don't send anything that identifies you. Results are cached server-side for 7 days under the (keyword, country, engine) key, so repeat scans return instantly.

For a prompt check, your exact prompt is sent live to ChatGPT to generate a fresh answer, so keep personal or confidential details out of it. Prompt checks aren't cached — each is a fresh snapshot.

Limitations

ChatGPT scans are US English only because the chat_gpt data source is US-locked at launch.
The underlying index updates on its own cadence; a very fresh trending query may not yet have coverage.
Mention counts reflect what appeared in the sampled answers, not what the engine would surface for every visitor. Use scans as directional signal, not absolute truth.
A prompt check is a single live snapshot. Because AI answers vary between runs, treat one check as directional and re-check to gauge how stable a citation is.