How to read our reports

What is checked

• Public website pages can establish what a business claims about itself, its contact details and policies. Seller claims are not independent corroboration.
• Dedicated Reddit and Google Maps research looks for the exact business. A Google branch rating or a small recent review sample is not the whole company's reputation. Trustpilot information is shown only with a matching profile citation.
• DNS and RDAP lookups supply available network and registration facts. TLS certificate validity, hosting geography and WHOIS privacy are not currently measured by this scanner.
• Directly observed rating facts and attributed excerpts are shown separately from search-assisted discovery links. Generated summaries do not count as verified evidence; discovery links are excluded from scoring. Matching an excerpt does not prove its truth or the meaning an AI assigns to it.

A best-effort score for every report

Estimate-v1 starts at a neutral 50/100. Exact-domain review ratings carry 60% weight: ratings are normalized around 3/5, with small samples shrunk toward neutral using review count divided by count plus 20. Only one listing per platform contributes. This is a reputation signal, not a measurement of safety.

Cited identity, contact, refund and shipping statements carry 25% weight, adding up to 5 points across four observed check types. A retrieved registration record carries 15% weight and adds 3 points. Unknown or irrelevant checks add no points and incur no penalty. Seller statements and registration do not certify legitimacy.

Adjustments are halved for stale or undated evidence, then rounded to a whole number. With no usable scoring signals the result is 50/100 with low confidence, not an endorsement. Scores below 40 indicate higher observed risk, 40–69 caution and 70–100 lower observed risk. These are uncalibrated heuristic estimates, not safety probabilities or validated fraud-detection results. Every report explains its calculation.

Coverage, unavailable data and freshness

Each report separates cited-source coverage from its score. An unavailable review source is not a zero-star rating; zero reviews is not negative feedback. Website downtime or a timeout is an availability observation, not a reputation score of zero.

Saved reports or supporting evidence older than 24 hours retain a numeric estimate, with low confidence and a refresh recommendation. Opening or exporting a report does not run a new scan. A failed refresh preserves the last saved report and its dates. Scan collection times, report-save dates and PDF-export times are labeled separately.

Older reports remain accessible at their original URLs. Their estimates are recalculated from recoverable observations, or use the neutral 50/100 baseline when none can be recovered. Historical AI scores are not reused. Viewing a report does not rewrite the stored record.

Limitations and corrections

Sources may be blocked, biased, outdated or about a namesake. An automated excerpt check verifies a citation exists in retrieved material; it does not prove every interpretation true. Reviews may contain unverified allegations. No automated report can replace checking the seller and using protected payment methods.

Use Request a correction or review on a report to provide its link, the disputed claim and supporting evidence. A submitted request does not mean a human has reviewed or certified the report.

Fresh scans depend on your plan and daily limits. Pro includes on-demand refresh; see Pricing.