How Viclaro Atlas Measures AI Recommendations
Atlas is a public record of recommendation behavior: the prompts asked, the assistants tested, the firms named, and the uncertainty around each rank.
Atlas measures recommendation behavior, not firm quality
Viclaro Atlas does not rate whether a firm is good, ethical, successful, or right for a particular client. It measures a narrower thing: which businesses AI assistants name when asked buyer-style questions in a city and category.
That distinction matters. A firm can be excellent and absent from the scan. A firm can rank highly and still be wrong for a specific person. Atlas is a map of AI recommendation behavior, not professional advice.
We start with a category and a prompt set
Each Atlas vertical is defined by a metro, industry, and category: for example, New York City, legal, divorce and family law. For that category, we maintain a prompt set across buyer states such as crisis, learning, decision, validation, constraints, and follow-up.
Those prompts are written to resemble how people actually ask for help. Some are broad recommendation questions. Others include situations, constraints, urgency, or uncertainty. That mix is what makes the ranking more useful than a simple keyword leaderboard.
We query multiple assistants and keep the receipts
Atlas scans ChatGPT, Claude, Gemini, and Perplexity. For each prompt, we record the assistant response and extract the firms named. The public ranking is an aggregate; the underlying research record ties each citation back to a specific prompt, provider, and response.
That receipt trail is the difference between a score and a study. If a firm asks why it ranked where it did, the answer should not be "because the model said so." It should be traceable to the sampled prompts and responses.
We publish uncertainty instead of pretending the number is exact
AI recommendation behavior is noisy. Assistants update. Search integrations change. The same broad question may produce a slightly different answer later. Atlas handles that by using repeated samples, methodology versions, and confidence intervals.
The goal is not false precision. The goal is directional truth strong enough to guide action: who is consistently surfaced, who is missing, which assistants disagree, and which gaps are worth investigating.
Key takeaways
- Atlas measures AI recommendation behavior, not professional quality.
- Every ranking depends on a disclosed prompt set and methodology version.
- The audit layer turns the public ranking into firm-specific fixes.
Next step
Atlas shows the public map. A Viclaro audit turns that map into the prompts your firm is losing and the page edits most likely to change the next scan.
Open the Atlas methodology.