Google Discover and News Surface Programme for Publishers

Discover does not have a keyword you can target, and the single largest technical cause of a site being absent from it is still one line: no max-image-preview:large, or a hero image under 1200 pixels wide. We check that in the first hour. After that the work stops being technical, because Discover picks articles by topic and entity association rather than by query, and the question becomes which subjects Google already believes your masthead is a legitimate source on. The reading we do before we advise: Sixteen months of the Search Console Discover report, joined to your CMS export so every row carries author, section, publish hour, and image aspect. Then we find the shape of the successes. It is rarely what the newsroom thinks. On gaming titles it is usually one writer and two franchises carrying the whole surface; on pet publishers it is service journalism about symptoms and cost, and the news desk is contributing nothing. That report is the only honest record of what Google associates you with, and most publishers have never had it read properly. Entity and authorship work: Author pages that exist as real entities with an sameAs trail, bylines that resolve to a person rather than a desk, Article markup with datePublished and dateModified that reflect reality rather than a nightly rebuild touching every timestamp. Sitewide, we sort out which sections you are actually a source for and which you should stop publishing into. Publishers lose Discover more often by diluting what they are known for than by any technical fault. The desk-level changes: A house rule on images: originals, wide, no logo bars, no text overlay, and never the same stock frame two publishers are already using. Headlines written to be true rather than withheld, since Discover suppresses the withholding pattern and your audience punishes it on the second click. Publish timing set against when your own successes historically landed, not against a generic best time. Update policy for running stories, so a substantive rewrite gets a fresh URL when it deserves one and a modified date when it does not. What we will not do: We do not write headlines for you or staff your desk. We do not run a Google News Publisher Center application as a standalone product. We will not build an aggregation or rewrite operation, and if that is the plan we will decline the work at the call rather than at the invoice. We do not promise Discover volume, ever. It is uncontracted traffic that can halve on a core update through no fault of yours, and any agency putting a Discover number in a proposal is selling you a coin toss. Who this suits and who it does not: It suits publishers already doing 200,000 sessions a month with an editorial team that can change its own workflow, and it does not suit a content site hoping Discover will substitute for a search programme it never built. If more than half your traffic already comes from Discover, the honest engagement is the reverse of this one: reduce the dependency before the next update does it for you, and we will say so on the first call. Deliverables are a written decision document, the section-level call on where to double down and where to stop, a technical fix list scoped to what your platform team can ship in a sprint, and a working session with the desk. Ongoing engagements add a monthly read against the same joined dataset so drift shows up while it is still a trend rather than a quarter.