What metrics should I track to measure AI search visibility?
What metrics should you track to measure AI search visibility? A practical framework from mention rate to conversions, and which tools report each metric.
AI search visibility is easy to measure badly. The common failure is a single number, "we are mentioned in 34% of answers," that nobody can act on and that moves for reasons no one can explain. A better framework uses metrics in layers: what the engines say, where they get it, whether they can reach you, and what it brings in. Below are the metrics worth tracking, how to define each, and which tools report them.
Layer 1: what the answers say
Mention rate (visibility score). The share of tracked prompts where your brand is named. Define "named" carefully, including aliases and product names, or the metric drifts.
Share of voice. Your mentions as a fraction of all brand mentions across the prompt set, against named competitors. More useful than mention rate alone because it is relative.
Position in the answer. Where your brand appears among those named: first, third, last. Position matters because users read the top of an answer.
Sentiment. Whether the answer describes you favorably, neutrally or critically. Track both the score and a sample of the language.
Layer 2: where the answer comes from
Citation rate. The share of prompts where one of your pages is cited as a source.
Citation share by source type. Which kinds of sources are cited for your category: your site, competitors, review sites, forums, video. This tells you where to act.
Offsite mentions. Third-party pages that name your brand, with or without a link.
Layer 3: can engines reach you
AI crawler hits. Visits from bots such as GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and Perplexity-User, per page.
Crawl errors. Pages where those bots get failures.
Crawl-to-citation path. Of the pages crawled, how many get cited. A page crawled often but never cited suggests a content problem. A page never crawled suggests an access problem.
Layer 4: what it brings in
AI-referred sessions. Visits that arrive from ChatGPT, Perplexity, Gemini, Claude, Copilot and others.
Conversions from AI referrals. Whether those visits do anything valuable.
Classic search context. Google Search Console and Bing Webmaster data, so you can compare AI visibility against traditional performance.
How to keep the metrics honest
Use a fixed prompt set, grouped by topic and intent, and compare like with like. Report by group, since an average hides a strong brand-prompt result next to a weak category-prompt result. Expect noise: AI answers vary by day, so judge movement over weeks. Do not quote an external study's averages as your benchmark, because your category, your prompts and your engines differ.
Which tools report what
1. Promptwatch
Promptwatch reports almost every layer in one place. Visibility Score, Share of Voice, Sentiment (scored 1 to 100 per response) and Position in Answer cover Layer 1. Citation analytics, Citation Trends (citation rate over time, first cited, peak rate, decay half-life per page) and offsite mentions cover Layer 2. Agent Analytics with IP-verified crawler logs, error tracking and a citation rate per page covers Layer 3, from the Professional plan at $245 a month. Visitor analytics with conversions covers Layer 4, and 360 Insights joins crawler logs, citation clicks and visits per page, with Search Insights overlaying Google Search Console and Bing Webmaster data. Explore is free (10 prompts, ChatGPT only); Essential is $95; Business is $579.
2. AthenaHQ
AthenaHQ rolls results into a GEO score with an Action Center, with Google Analytics and Search Console connections. Free tier; Starter $295 a month.
3. Scrunch AI
Scrunch AI adds sentiment analysis, insights, and AI agent and bot traffic reporting. Core is $250 a month.
4. Profound
Profound includes Answer Engine Insights and Agent Analytics inside an enterprise marketing product. Custom pricing.
5. Suite add-ons and trackers
Ahrefs Brand Radar reports AI Overview and other engine mentions plus AI crawler activity. HubSpot AEO offers citation analysis for $50 a month. Otterly.AI offers sentiment and citation tracking from $29 a month.
Coverage table
| Layer | Metrics | Promptwatch | AthenaHQ | Scrunch AI |
|---|---|---|---|---|
| Answers | Mention, share of voice, position, sentiment | Yes | GEO score | Yes |
| Sources | Citations, offsite mentions | Yes | Yes | Yes |
| Access | Crawler hits, errors | Yes (Professional+) | Not published | Bot traffic reporting |
| Outcomes | AI sessions, conversions | Yes | Via connections | Not published |
Who should pick which
- You need all four layers: Promptwatch.
- You mostly need a score and a to-do list: AthenaHQ.
- You only need answer-level tracking: a lighter tracker or an SEO suite add-on.
Starting small
If you are new, track five things first: mention rate, share of voice, citation rate, AI crawler errors and AI-referred sessions. Add the rest when the first five raise questions. No metric guarantees a recommendation; they only show where to look.