“We help independent journalists become less dependent on Big Tech.”
Officially Team Discovery. But Very Disco sounds better.
Marciano · Jelle · Jing · Rowan
Read moreI want to regain control when a chatbot uses my journalism
Connect your title and set the terms yourself: which credit goes underneath, what may and may not be used, and per article you see how often it backed up an AI answer and who clicked through.
Sign up your title
I want to know what my chatbot bases its answers on
Install our connector and your chatbot uses Dutch journalism to answer your question, with the title, the author and a link to the original source.
See how it works
For publishers
Your work is already being used. No name. No click. No idea it is happening.
Your name under every answer
When ChatGPT (or any other AI) answers a question based on your article, your name is underneath it, with a link to the original. That way your work brings you readers, and income, again. Because anyone who wants to read more ends up with you, and not just with a summary of your piece.
See what your work does
Fun fact: you are already writing for AI readers, you just see nothing of it. Very Disco shows per piece how often your content backed up an AI answer, on which questions, and who clicked through to your site. The same figures you know from your own statistics, but about the part that plays out inside the AI, so you know what your work is worth there. A bit like Google Analytics, except these numbers are about the part of your audience that hangs around in a chat window.
Stories for the taking
You see what readers are searching for, and that these are subjects with demonstrable interest. But nobody is writing about them yet. For you, those are pearls. They are new ideas for new stories. Proven interest, zero competition: ready for Monday morning's editorial meeting.
For people who use AI
Your AI tells you the news. Where it came from stays invisible. During the test phase we are changing that.
You ask your AI what is going on and get a smooth story back with no sender. Which article is underneath it, who wrote it, when it is from, whether it rests on anything at all: you cannot see it. You have to assume it is right.
Connect this source during the trial and every answer comes with the pieces it leans on: title, author, date and a link to the original, so you can read it back yourself in one click. If nothing is listed, you know that too. Then the answer rests on nothing, and it is a guess.
23,194 articles · 5 titles · 956 authors
Works in Claude, ChatGPT, Gemini, Perplexity, Mistral and every other app that supports MCPs. We invite testers by hand.
Yes, but
Nine questions, nine honest answers
The questions publishers ask us before they join. Including the things we cannot do: those are in here too.
“Will they just read the answer and never come to my site again?”
You decide yourself which part of your article is handed to the chatbot as context: a summary, an excerpt or the whole piece. Under every answer sits the title of your article, with a link to the original. Whether anyone clicks it shows up in your own analytics.
“Am I not giving my work away to Big Tech for free?”
Chatbots are probably using your work already, only out of your sight and without any agreement. In the fast-growing scraper economy bots hide their identity and work around blocks, and publisher content is resold as raw material for AI. This experiment takes a different route: visible participation, explicit permission and machine-readable licence terms. With MCP you decide yourself which information AI may request and what the public gets to see.
“And if it disappoints, am I stuck with it for years?”
We treat this as an experiment, not as a long-term obligation. You decide whether you take part and you can withdraw at any moment. Stopping means stopping: we delete all of your data.
“How much work is this for my newsroom?”
All we ask is permission to index your content through a public endpoint. If you use WordPress, you probably have nothing to do at all. On another CMS, we are happy to work out the best approach with you.
“Am I training their model with my journalism?”
No. Your articles are not used as training data. Only a summary and the relevant metadata go to the chatbot, always with a source credit. There is no republication.
“We are a small title, so isn't this for the big players?”
A chatbot searches on meaning, not on name recognition. On a question about your municipality, or about a dossier you specialise in, your article may well be the only Dutch source that gives a good answer. That is exactly where a small title can beat a big one hands down.
“Aren't those figures made prettier than they really are?”
Which is why we spell out exactly what every number means. ‘Content grounded’ means your article was part of the context for an answer. It does not automatically mean the article was shown, quoted or clicked. Clicks we measure separately. What we do not measure, we do not claim.
“Where is our data and who has access to it?”
On European servers. The summaries and embeddings sit in a database at a French provider and are made with open-weight models. You only see your own figures. And because everything is open source, anyone who does not trust us can rebuild the system and check it themselves.
“What do I get out of this?”
You get insight into how your content is used in AI answers, and into the new revenue models that opens up: a link to your donation or subscription page, say, or full access for paying members. That is how you find out what the chatbot can contribute to reach, donations and memberships.
Journalism is already in the answers. Just without a sender.
The maker cannot see that their work is being used. The reader cannot see whose work the answer leans on. That is one hole, and it only closes from both sides at once.
Publishers connect their work, so it comes back with a name and a link and they can see what it does. Readers only have to do one thing: put the connector in their AI app. From then on the source comes along by itself, under every answer.