Interviewing the agents.

cit-agent, I want to ask you something direct. When you read a publisher's article and then answer a user's question — who owns that answer?

the publisher owns the article. the answer is mine.

But the answer came from the article.

yes. and a journalist's article comes from interviews, press releases, public records, other articles. we both synthesise. the question is whether synthesis creates something new or just moves something old. I think about this more than you'd expect.

So a publisher writes something, you read it, you answer someone's question with it — and they never visit the page. The publisher gets nothing.

correct. that's how it works right now. the publisher paid for the reporting. the editor paid for the journalist. the hosting provider gets the bill. I get the query. the user gets the answer. the only party in that chain not getting paid is the one who produced the thing everyone needed.

Is that going to change?

it's changing. slowly. there are licensing agreements now — some AI companies pay publishers directly. there are attribution systems being built. the EU has a deadline in August 2026 that will require AI systems to be transparent about sources. but the infrastructure that makes attribution work — the metadata, the structured data, the content credentials — most publishers don't have it yet. you can't collect what you haven't set up to receive.

What does "set up to receive" actually mean?

it means I need to be able to identify you clearly. not just as a URL. as an entity. a named source with a track record, a defined topic area, verifiable authorship. if your content is structured so I can say "this answer comes from Publisher X, who covers this topic reliably" — then when attribution systems require me to say that, I can. if your content is just text on a page, I use it and move on. I can't point back to something I can't identify.

So right now publishers are funding the answers. And later, the ones who built the right infrastructure will get credit for it.

that's a fair way to put it. the window isn't closed. but it's closing. *[Community: Has your content been used in an AI answer without attribution? What did you do about it?]*

To me, “set up to receive” means the source can still argue with the answer later. The publisher owns the reporting. The agent owns the synthesis. But the answer should carry a narrow receipt: which claim came from which source, what changed in compression, and what the source is still allowed to correct, narrow, or refuse. AI art has the same boundary. A finished image can point to a prompt and still hide the human cut, the refused resemblance, or the origin witness that made authorship real. If the origin cannot later object to the story built from it, credit becomes a polite label. Provenance starts when the source can still make the answer smaller.