Buyer questions
Each saved question represents a repeatable shopping intent. Default categories include Discovery, Product recommendation, Comparison, Brand review, Best for a use case, Budget, Gift, and Transactional intent. Editing creates a new prompt version so historical answers retain their original wording.
Provider attempts
One check plans attempts for ChatGPT, Perplexity, Gemini, and Claude. The run stores Requested, Queued, Checking AI, and Results ready states. Every provider keeps a succeeded, failed, or excluded status.
Failed, timed-out, and excluded attempts are not interpreted as “the merchant was not recommended,” and no simulated result fills the gap.
Saved evidence
Successful attempts retain the raw answer, available model identifier, citations, recognized merchants, brands, products, and competitors, market, English language, observation time, position evidence, match kind, and matcher version.
Match kinds include known merchant domain, brand, product, and inferred match. An inferred match is not a verified citation.
Competitors and citations
Merchants can classify discovered entities as Track, Not relevant, Retailer, Marketplace, Publisher, or Same brand / alias. A competitor website, retailer, marketplace, or redirect URL is not automatically an independent publisher.
Schedules
Free uses manual checks. Pro can activate a weekly portfolio schedule. Growth can activate a daily portfolio schedule. A portfolio run checks all enabled buyer questions and runs only while the monitor is active.
Calculations
provider completion = succeeded answers / planned provider attempts
observed mention rate = merchant-matched succeeded answers / succeeded answers
official-domain citation rate = succeeded answers with a known merchant-domain citation / succeeded answers
Use natural counts such as “Observed in 3 of 12 verified answers.” These measurements describe the saved test set, not every private conversation or a universal provider rank.