About PubTrust

Because the documented complaint in this market is not about detection. We went looking for what publishers actually say about their ad-quality vendors, and the grievances are almost entirely about opacity and collateral damage: you cannot find out why an advertiser was blocked, you cannot measure what blocking cost you, and you cannot get a straight answer about what the tag is doing to your page. Nobody complains that their vendor lets malware through. So the opening was never "detect better" — it was "be measurable". Every design decision in PubTrust follows from that: blocks carry a price, latency is published, verdicts name the exact rule that fired, and the limits are stated on the homepage rather than discovered in a trial.

1. Over-blocking is the real failure mode.

   A missed ad is recoverable. A blocked legitimate advertiser is lost revenue and a phone call from a sales director. Every weight in our corpus is calibrated around which of those is worse, which is why bare gun is 0.1 and buy ammo online is 1.2, and why Weapons, Political, Alcohol and Shock sit lower than a competitor's marketing department would like.

2. A claim we cannot evidence is worse than no claim.

   No detection rates, no customer counts, no fabricated logos, no case studies we did not run, no conformance to a standards body that dissolved in 2024. Ad-ops teams test claims, and being the vendor that is accurate about its own limits is worth more than any claim we would be giving up.

3. A vendor should be easy to remove.

   Kill switch, escape hatch, readable source of the exact build running on your site, and a one-toggle off. If leaving is hard, staying was never a choice.

4. Enforcement is a human decision.

   Sites start in Monitor. An agent can investigate, recommend and draft policy; it cannot switch blocking on. Somebody with a name and an account has to click, and it is written down who.

ClearTrust builds governed, auditable infrastructure for teams that have to explain their systems to somebody — a regulator, a board, a client, a reader. The same commitments run through every product in the line: append-only, hash-chained audit trails; explicit, documented data retention; no silent impersonation; enforcement actions gated behind a human; and no claim that cannot be evidenced from the product itself. PubTrust is the publisher-facing member of that family, and shares its design language and its standards.

→cleartrust.cc

Every rule carries a weight, a language tag and a written note in English explaining why it is there — glossing non-English terms, naming false-friend hazards, and justifying surprising weights. Those notes are not internal documentation; they surface in the dashboard when a violation is explained to you. New domains arrive mostly from the server-side deep scan, which follows landing pages and proposes confirmed bad domains back into the packs — and every one of them still gets a human weight and a human note before it is published to anybody. Packs ship monitor-first, always, for exactly this reason.