How Digest AI works
Digest AI is a news desk for artificial intelligence. Every twenty minutes it reads about eighty sources, from the research labs' own announcements to the major technology press and the communities where practitioners talk, and turns what is new into one page per story.
For the engineering side (the pipeline, the fact checks and how it runs on free tiers), read How Digest AI runs a round-the-clock AI news desk on free tiers.
What you get on a story page
- A headline, three key points and a short digest written by our editorial model from the sources listed on the page.
- The full text of the lead article, credited and linked, so you can read it without leaving. Publishers who do not permit reuse are shown as a digest with a link instead.
- Every other outlet covering the same story, so you can compare accounts.
How stories are ranked
Ordering combines the editorial importance our model assigns, how recent the story is, how many independent sources cover it, and what readers actually open and read. The last signal is learned continuously: we record page views, time on page, how far a story is read and clicks to the source, without cookies or personal identifiers, and use them to predict which new stories will matter to you.
Editorial standards
- Every claim has a source. A story page lists each article it draws on and links to it. Digests are written from those articles only, never from memory or speculation.
- Machine-written, clearly labelled. Headlines, key points, digests and thread summaries are generated by our editorial model and labelled as such on every page. An automatic check compares each headline's figures and names with the source articles and flags any it cannot find there for a person to review. It does not verify every sentence, so the sources remain the record.
- Primary sources first. When a lab, company, court or regulator has published something itself, that document is included and marked as the primary source.
- No paid placement in the news. Ranking is decided by editorial importance, freshness, breadth of coverage and reader behaviour, never by payment. Sponsorship, if ever introduced, will be labelled.
- Publisher rights. Full texts are reproduced with attribution and a prominent link; publishers who prefer a digest only are honoured on request.
- Privacy. No accounts, and no cookies of our own. Visits are recorded as anonymous events (the page, the site the visit came from, how far down a story you read, what you searched for, and a random visitor number that stays in your browser and is replaced every day), used only to count readers and rank stories, and deleted within 90 days. We do not store your IP address in those events, though the servers that deliver the site and the publishers whose pictures a story shows see it as any server does. Google Analytics runs only if you choose Accept on the notice, and you can change that choice with Cookie settings in the footer. The privacy page lists everything, including the other services a page loads.
Corrections
Automated summaries can be wrong. If a headline, key point or digest misstates what the sources say, email hello@digestai.news with the story link. Corrected stories are updated on the next publishing run and marked as updated. Publisher requests and the takedown process are on the contact page.
Follow Digest AI on LinkedIn, Bluesky and X for the day's biggest AI stories.
Feeds and data
Follow every story by RSS, listen through the podcast feed, or read Digest AI as data: each story is also published as JSON and Markdown, with the latest stories, today's briefing and each category as small JSON files. The API page documents them and shows how to cite a story.
Publishers
If you publish on AI and want to be added as a source, or would prefer we show only a digest of your articles, write to the same address. Digest AI sends readers to your site with a prominent link on every story and identifies itself as DigestAIBot when it fetches pages.