When a model can compress a public record in seconds, the story that survives is the one a summary cannot honestly stand in for.
That is the bet news leaders described to the Reuters Institute for its 2026 predictions report. Faced with answer engines that swallow service journalism and evergreen utility, they said they would move toward original reporting, context, and human stories, the work they consider harder to commoditise. Nic Newman made a related point in the Institute’s conversation about the report: video and audio are not immune to compression, but a compressed clip is a poor substitute for a person the audience has a reason to trust.
This is a consumption shift as much as a production one. People still want to know what happened. A growing share of “what happened” will be answered before they reach you, inside a chat window or an overview. What they cannot get there, reliably, is the thing you witnessed, the document you obtained, the correction you made, and the judgement about why this fact outranks that one. Those are the reasons to open your site, and the reasons to pay.
The trap is to hear “more original reporting” as a slogan and then fill the site with more of the same rewritten wire. Distinctiveness is expensive. It is also the only inventory that gets more valuable as generic copy gets cheaper. A newsroom that spends its AI time savings on volume is competing with the model. A newsroom that spends them on reporting the model does not have is competing on something else.
There is a product consequence. The homepage, the newsletter, and the subscription pitch have to show that difference in the first screen. If the first thing a reader meets is a paragraph any tool could have written, you have trained them to stay in the tool. Lead with the reporting, and let the formats — audio, a short selection, a sourced answer — carry that reporting into the moment they actually have.
AI is helpful here as a production layer under work that stays identifiably human. It is a bad strategy as a substitute for the work.