PubTrust vs GeoEdge

Several things, and a publisher should weigh them seriously. Deployed footprint and operating history: GeoEdge has been doing this for well over a decade and is running on a very large amount of live inventory, which means their verdict data is drawn from a correspondingly large population of real creatives — and in this category the size of the observed creative population is a genuine technical advantage, not just a sales one. Ad-ops familiarity: many ad-ops teams already know the interface and the workflow, and the cost of retraining a team is real. A mature support organisation with people who have seen a lot of incidents. The verdict-cache architecture: their tag carries precomputed verdicts so that most decisions require no network round trip at all — which is the correct design for real-time blocking, and one we adopted deliberately rather than invented around. And the inert-park technique — parking creative markup in nested inert parsing contexts so nothing executes before a verdict — is, in our assessment, the right way to block malvertising, and we implement it because they demonstrated it, not in spite of that.

1. Every block carries an estimated revenue impact.

   We know the winning bid's CPM at the moment of the decision for header-bidding demand, so every blocked impression in PubTrust carries a price and every policy carries a running total. As of September 2026 we are not aware of a vendor in this category that prices blocks for the publisher. The publisher grievance in this market, as far as we can document it, has never been about misses — it has been about not being able to measure what blocking cost.

2. We publish measured latency.

   Decision time is measured in our tag, reported with every beacon, and displayed in the dashboard per site and per layer at median and p95. As of September 2026 we could not find a published latency figure from any vendor in this category — ourselves included, until now. This is a category norm we are deliberately breaking, and the most useful thing a publisher can do with it is hold us to it.

3. Multilingual corpora, authored natively.

   16 packs × 14 languages = 12,034 rules, each language written against its own market's regulation and vocabulary, with false-friend hazards documented per term. German GlüStV evasion vocabulary, Nordic self-exclusion-register evasion, the Turkish grey-market lexicon, Brazilian Portuguese versus European Portuguese. We make no claim about anyone else's language coverage — ask them directly, and ask specifically whether the non-English rules were authored or translated, because the answer determines whether the pack blocks a Swedish sports desk's own match reports.

4. An agent-native control surface.

   An MCP server with OAuth 2.1, where an agent can investigate and draft policy but cannot switch enforcement on. We are not aware of another vendor in this category shipping one.

Read or screenshot a cross-origin creative from the page. When a creative renders in a SafeFrame or any cross-origin iframe, its document is unreadable from JavaScript — contentDocument returns null — and screenshotting it is impossible by four independent mechanisms: an iframe is not a valid canvas image source, canvas tainting blocks pixel readback, the SVG foreignObject technique fails because there is no readable DOM to serialise and because SVG-in-<img> runs in secure static mode, and getDisplayMedia requires a user gesture and a picker. Libraries like html2canvas do not solve it — they do not screenshot at all, they re-render the DOM they can read, and against a cross-origin iframe they silently emit a blank rectangle with no error, which is worse than failing. Anyone claiming client-side capture of cross-origin creative content is describing something browsers do not permit. We capture creatives server-side, in a headless browser, where the boundary does not apply. So, we would expect, does everyone else.

Confiant, which ships a Prebid Real-Time Data module and has deep programmatic relationships. HUMAN, whose cleanAD product came from its 2022 acquisition of clean.io — note that cleanAD belongs to HUMAN, not The Media Trust, a confusion that appears in a lot of comparison content. The Media Trust, whose competing product is Media Filter. Boltive, formerly Ad Lightning, which rebranded rather than being acquired. Only four vendors ship a Prebid RTD module at all — GeoEdge, Confiant, HUMAN and The Media Trust — which is why we ship one.