Content intelligence for advertising
Every page, read the way an editor would.
Vanegraph tells ad platforms what a page or an AI conversation is about, whether it is safe for brands, and what the reader is trying to do. In the IAB categories your stack already speaks, with a reason for every label. Polish, English and more. No personal data.
Rząd przyjął plan dopłat do aut elektrycznych na 2027 rok
Program obejmie samochody elektryczne do 150 tys. zł. Eksperci przewidują wzrost sprzedaży w drugim kwartale, a dealerzy już przygotowują oferty dla klientów indywidualnych.
Try it on a real page or chat.
Paste a link, some text, or an AI conversation. This is the live engine in fast mode; the API's default mode is a little slower and more accurate.
Results appear here: topics, a brand-safety tier and intent, each with the reason behind it.
Three answers, in the language your stack speaks.
Every response maps straight into OpenRTB and curated deals. No new taxonomy to learn.
01 · Topic
What it is about
Up to three of the 704 IAB Content Taxonomy 3.1 categories, most relevant first, with confidence.
02 · Brand safety
Whether a brand belongs next to it
A tier from floor to none, and the risk categories behind it. News is judged on how it treats a subject, not on scary keywords, so reporting stops being blocked by default.
03 · Intent
What the reader wants now
Learning, comparing, buying or navigating, plus a 0 to 1 commercial score. For AI chats it reads the user's actual request.
We publish our accuracy. Per language.
Nobody else in contextual advertising does. These are measured against gold labels on real news pages, scored on the live API.
| Measure | English | Polski |
|---|---|---|
| Main topic correct | 81% | 82% |
| Brand-safety tier, exact | 82% | 80% |
| Brand-safety tier, within one level | 100% | 100% |
| Reader intent | 95% | 98% |
Accurate mode, engine v0.5, November 2026. Gold labels are the consensus of three independent AI models from different families on 86 English and 126 Polish news pages; a human audit of disagreements is in progress. We will publish larger sets and more languages as they are ready.
For AI-app ad networks
An independent rating for ads in AI chats.
Advertisers need someone other than the chat app to vouch for where their ads appear. Vanegraph rates each conversation for topic, brand safety and the user's intent, without storing a word of it.
- No personal data, by designWe classify content, never people. No cookies, device IDs or profiles, and nothing to consent to.
- Conversations are never storedChat text is read once and discarded. It is not cached or logged.
- A polite reader of the webVanegraphBot respects robots.txt and honours Content Signals such as ai-input=no.
- Built in the EU, for the EUDesigned around GDPR, the DSA and the political-advertising rules, which allow contextual signals and restrict profiling.
Simple prices, published.
No minimums, no sales call to see a number. Publishers: ask about a revenue-share pilot instead.
hello@vanegraph.com