Policy
Corrections and right of reply
We publish evidence-based statements about named commercial products, some of them unflattering. Anyone affected must be able to challenge them, and manufacturers are explicitly welcome to do so.
Anyone may submit
Every robot profile carries a Report an issue or submit evidence link. Researchers, journalists, members of the public and manufacturers all use the same form and the same queue. There is no separate manufacturer channel, because a separate channel would be a preferential one.
What we commit to
- Acknowledgement within two business days of a submission entering the queue.
- Factual and evidence corrections are triaged first. A claim that misstates a source is a defect and is treated as one.
- A visible trail. Where a correction changes a profile, the score history records what changed and why. Submitter identity is never published.
How submissions are classified
- Factual correction
- We recorded something that does not match the cited source. Highest priority. Fixed and logged regardless of who reported it.
- New evidence
- A source we have not seen. It enters the normal workflow: durable capture, source classification, claim or deployment drafting, second review, then recalculation. It may raise a score, lower one, or change nothing.
- Disagreement or opinion
- A view that our methodology weights something wrongly, or that a score feels unfair. We read these and they inform methodology versions, but they do not change a score on their own. Disagreeing with a scoring rule is not the same as showing the record is wrong.
- Legal or privacy report
- Routed separately and promptly. Includes claims of rights infringement in imagery or quoted material.
What we will not do
- No paid score changes. Scores cannot be bought, sponsored or influenced by a commercial relationship of any kind.
- No direct editing. No manufacturer, advertiser or subscriber can edit a profile. Accepted evidence goes through the identical review path as evidence we found ourselves.
- No preferential treatment. A submission from a manufacturer, a sponsor or a paying subscriber is triaged on the same criteria as one from an anonymous reader.
- No quiet deletion. Evidence is superseded and logged, never removed to make a profile look better.
Why a score may not move even when we agree with you
A robot can be more capable than its score suggests and the score still be correct, because the score measures established evidence, not our estimate of the machine. If the supporting source is first-party, or not durably captured, or does not establish autonomy, the ceiling applies no matter how accurate we privately believe the claim to be. The remedy is better evidence, and we will say so plainly rather than adjusting a number.