How Atlas AI Index Measures AI Search Visibility
Atlas measures AI visibility with repeatable buyer-intent prompts, provider-specific answers, competitor appearances and cited source evidence so you can track changes instead of relying on anecdotes.
1. Build a relevant prompt set
Atlas starts with the business category, location and known competitors to create buyer-intent prompts that reflect how a prospective customer might ask for recommendations. The aim is commercial relevance rather than random brand mentions.
A stable prompt set creates a baseline that can be repeated later, which is essential if you want to tell the difference between genuine movement and a one-off answer.
2. Query the active provider
Each scan is tied to a named provider and timestamp. The launch provider is Perplexity Sonar accessed through Vercel AI Gateway.
Atlas does not merge that result into a vague universal AI score. Provider labelling keeps the methodology honest and makes future multi-provider comparisons possible.
3. Record brands, competitors and sources
Atlas checks whether your business is mentioned, identifies competing brands in the response and captures the web sources used to ground the answer. Prompt-level evidence is retained so the summary score can be inspected rather than treated as a black box.
Source evidence is particularly useful for GEO work because it shows where competitors may have stronger third-party coverage.
4. Calculate recommendation share
Recommendation share summarises how often your business appeared across the tracked prompt set. Competitors receive comparable appearance scores so you can see relative visibility.
The score is a practical benchmark for the measured provider and prompt set, not a guarantee of what every individual AI user will see.
5. Repeat the methodology
Monitoring repeats the same measurement approach so you can compare snapshots over time. That makes it possible to evaluate whether changes to content, citations or brand presence coincide with stronger AI visibility.
Atlas monitoring is designed around approximately weekly recurring scans, with on-demand scans available to active monitoring accounts.
How Atlas works FAQ
Is the Atlas score a universal AI ranking?
No. It is a provider-specific recommendation-share benchmark based on the prompts used in that scan.
Why use repeatable prompts?
Repeatable prompts make snapshots more comparable over time and reduce the temptation to judge visibility from one favourable or unfavourable answer.
Does Atlas keep the underlying evidence?
Yes. Scan results include prompt-level responses, competitor appearances and source domains so the score can be understood in context.
How often are monitoring scans run?
Monitoring is designed to run approximately weekly, subject to provider availability and service limits.
See the methodology on your own business
Run a free scan and inspect the prompts, recommendations and sources behind your score.