Google Discover Strategy
Discover is the largest traffic source most publishers feel they cannot control, because there is no query to optimise for and no ranking to check. I grew Discover traffic 30 to 40% at Sportskeeda and have run it for two sports newsrooms since, and most of that work is less mysterious than it looks.
The same story, three surfaces
Search, Discover and Top Stories reward different things, and a newsroom writing one headline for all three is compromising on all three. Here is the same match report framed for each.
Search
A query exists and someone typed it. Clarity wins: the headline should make it obvious this page answers the thing that was asked. Ambiguity costs you the click and the ranking.
England vs India third Test result and scorecard
Discover
Nobody asked anything. The feed decides this might interest you, so the headline has to earn attention from a standing start. Curiosity and a specific angle beat completeness.
England's batting collapse has been coming for three years
Top Stories
Freshness and fact. This is the surface where being first and being plainly accurate matters most, and where anything that reads like a tease gets skipped over.
England lose third Test by 108 runs after middle order collapse
Why Discover traffic disappears
Every publisher has had the week where Discover falls off a cliff and nobody can say why. These are the eight causes, in roughly the order they turn out to be responsible. The first one is the answer more often than the other seven combined.
Nothing broke
Discover has no queries, so there is no ranking to lose. Google simply stops surfacing a story. Most people go looking for a technical cause and there is not one, which is the single most useful thing to understand about this surface.
The interest moved on
Feed lifespan is short. A piece runs for hours or a couple of days and then stops, and that is the normal shape rather than a problem. Judge Discover over weeks of output, never a single article.
A core or spam update landed
Discover eligibility is judged at publication level as much as page level. When an update changes who Google trusts, Discover often moves before Search does and moves further, which is why publishers feel these updates first here.
The image stopped qualifying
Large images need max-image-preview:large and an image big enough to fill the card. Change a template, lose the meta tag or drop below the size threshold, and you fall out of the format that was earning the clicks.
Trust got spent
Headlines that overpromise work once. Across a whole site, a pattern of the article not delivering what the headline suggested is the fastest way to stop being surfaced, and it is very hard to undo.
It was never yours
A story big enough lifts every publisher covering it. When the traffic drops afterwards it looks like a penalty and is actually just the story ending. Check whether the peak matches an event before diagnosing anything.
A policy line got crossed
Discover applies content policies that are stricter than Search, and they cover things a publisher might consider normal editorial judgement. Worth reading properly if a section of the site went quiet on its own.
The reporting changed, not the traffic
Discover only appears in Search Console when a site passes a threshold, and the report has its own delays and gaps. Confirm the drop in analytics before spending a week on a cause.
What I work on
Headline and packaging for a feed
The headline and the image are the entire unit in Discover. There is no snippet, no URL, and no query to match against, which makes packaging the whole job rather than a finishing touch.
Image and template requirements
Large image eligibility is a technical setting most publishers get wrong once and then never check again. It is also the cheapest Discover fix available, because it applies to everything you have already published.
Entity and interest coverage
Discover matches content to interests rather than queries, and interests are built out of entities. Covering the people, teams and topics your audience follows consistently is what makes you a candidate. Related to how those entities are linked together.
Timing around events
Demand for a fixture or an event arrives on a schedule and leaves fast. Publishing into the window rather than after it is a planning problem, which for sports sites is most of the work.
Diagnosing a Discover drop
Working out which of the eight usual causes it actually is, before anyone rewrites the site. Most Discover panics are a story ending, and telling that apart from an update is the difference between doing nothing and doing damage.
Not breaking Search to chase Discover
Curiosity headlines that win in the feed can quietly cost you the query the page was written for. The answer is usually surface aware framing built into the editorial workflow, not choosing one surface.
How I approach Discover
Five rules, learned mostly by getting the first one wrong.
Establish what normal looks like first
Discover is volatile by design, and without a baseline every dip looks like a crisis. Before changing anything I want several weeks of output charted against events, so we can tell a real decline from the usual shape.
Fix eligibility before framing
There is no point workshopping headlines if the template is not serving large images. Check the boring technical requirements first, because they apply retroactively to the entire archive.
Write for the surface, not for all of them
One headline doing duty across Search, Discover and Top Stories is a compromise that underperforms on all three. Where the CMS allows a variant per surface, that is usually the single biggest available gain.
Judge it in aggregate
Individual Discover performance is close to noise. Patterns across a few dozen pieces are real. Any process that reacts to a single article's Discover numbers will teach the desk the wrong lesson.
Put the framing in the brief
Discover knowledge that lives in one person's head stops working the day they are on leave. It belongs in the brief and the template so the desk applies it without thinking about it.
"Discover has no queries and no rankings, so there is nothing to lose in the way people expect. A story stops being surfaced, and that is the whole event."
Discover work I can point at
Three examples with the company named.
Growing Discover traffic 30 to 40%
Inside a newsroom already doing over 100M pageviews a month, where the framing had to change without slowing anyone down.
Discover and Top Stories for a sports media site
Optimising coverage for both surfaces around major events, where demand arrives on a schedule and leaves just as fast.
Editorial systems behind the framing
Discover work only holds if the desk can apply it without a review step, so the framing went into the brief.
Frequently Asked Questions
Your Discover traffic dropped and nobody can tell you why
Send me the site and when it happened. I will tell you which of the eight it probably is, whether or not we end up working together.