Capabilities are easier to establish than outcomes
Claims verified
Located evidence supports the material meaning of the recorded claim.
Claims partially verified
Only part is supported or a material limitation changes the claim's scope.
Claims unable to verify
Available evidence does not permit a reliable conclusion.
Claims classified misleading
Evidence materially conflicts with, or changes, the impression created by the claim.
What did the claims review show?
The current dataset contains 168 claims across 54 products. 109 are Partially Verified, Unverified or Unable to Verify. Those statuses do not mean a claim is fraudulent; they describe what the reviewed evidence could establish.
Verification outcomes
Claim categories
Interpretation
Feature claims
A public workflow or documentation can support that a feature exists. It does not show the feature improves returns or suits a particular strategy.
Performance and AI claims
Backtests, testimonials, marketplace histories and AI labels are not independent live-return records. Verification needs a defined period, method, fees, drawdowns and evidence resistant to selection bias.
Scale and security claims
User counts and volume figures need definitions and independent support. Security controls can be documented while implementation effectiveness remains unaudited.
Method and the Misleading threshold
Each structured record stores the original claim, category, source URL, evidence found, verification status and analyst note. Counts are calculated during the build. They are claim-level statistics and must not be confused with the 54 product denominator or 328 source-record denominator.
“Misleading” is used only when available evidence materially conflicts with, or changes, the impression created by a claim. Lack of independent verification alone is not enough. Multiple company-controlled pages can provide useful primary evidence, but repetition does not convert them into independent corroboration.