How AI describes you today.
Where you're cited, where you're absent, and what the models get wrong — for ChatGPT, Gemini, Perplexity and Claude, in the words a buyer in your category would actually use.
What Measure does
Measure is the first step of the loop and the one every plan includes. It asks each AI model the questions your buyers ask, records what comes back, and turns the answers into five things you can act on: how each model describes your business, how often you're recommended, where you're missing, who is recommended instead, and what has changed since last week.
What you see
| Metric | What it tells you |
|---|---|
| How AI describes you | The model's description of your business, by model, graded right / incomplete / wrong — e.g. a service you don't offer, a service area it doesn't know |
| AI visibility | The share of your category's prompts where you appear at all |
| Recommendation rate | The share where you're recommended, not merely mentioned |
| Share of voice | Your recommendations against named competitors |
| Citations | Which of your pages the models cite, and how often, over a rolling window |
| Average position | Where you sit in the list when you're named |
| Description errors | Count of factual errors in the models' descriptions, and whether it's falling |
| Alerts | What changed this week — a new error, a lost recommendation, a new citation |
Illustrative dashboard values: AI visibility 36%; ChatGPT 34 · Claude 40 · Gemini 33; recommendation rate 12% (▲4 pts); citations 14 in 30 days (▲12); average position 2.5; description errors 2 (▼5).
The prompt set is the instrument
Rank tracking assumed the same query returns the same list. AI doesn't: ask the same question twice and the odds of an identical list are under one in a hundred.
So Measure doesn't track a keyword; it tracks a prompt set — the questions your category's buyers ask, written per audience segment, run repeatedly, and scored as probabilities rather than positions. You add your domain; the first scan builds the category prompt set and starts on the spot. Prompts are unlimited on every plan.
Engines, cadence, history
ChatGPT, Gemini and Claude on every plan; Perplexity and Grok on Agency and the managed program. Weekly runs; history kept, so a change in a model's answer is dated to the week it happened. Your prompts and history carry into the managed program if you move — nothing resets.
What it doesn't do
It doesn't change the answer. Measure tells you what AI said; the rest of the loop changes what AI knows. If you already have a monitoring tool, this step will look familiar — the difference is what's attached to it.
Questions
Start with the score.
Enter your website. In 24 hours: your GEO Score, how each model describes you, where you're missing, and the first three fixes.