Google Discover editorial strategy: what a newsroom does before, at and after publishing
A Discover story earns most of its clicks in two days. Win it before publishing with the topic, the entities and the image, at publishing with an honest headline, and after publishing with follow-ups, not with more URLs.
Key takeaways
- A Discover story lives about two days: editorial articles stay about 48 hours in the feed.
- More URLs do not buy more reach: more articles share the same feed, so each hit gets smaller.
- Google now judges expertise topic by topic and prefers local, original, in-depth stories. Depth beats breadth.
- Name the event in the headline and keep a reason to click. Shock words and mystery templates lost.
- Change a headline only for a real gain, never the URL, and republish only with real news.
This article is only about the editorial side: what to publish, when, with which headline and image, and what to do with a story after it is live. The technical side is in my Google Discover SEO article.
What changed in 2026
On 5 February 2026 Google announced a Discover core update with three changes:
- More locally relevant content from sites based in the reader’s own country.
- Less sensational content and clickbait.
- More in-depth, original and timely content from sites with expertise in a topic.
It rolled out for English in the US from 5 to 27 February 2026, and Google announced that all countries and languages would follow in the months after.
What happened next, as measured by others:
- The US feed got more local and more descriptive. In a NewzDash panel, US sites rose from 88.86 to 89.94 percent of the feed, the average title grew from 72.9 to 84.2 characters, and California-local articles in the top 100 rose from 10 to 16.
- Templates dropped out. One publisher had 5 near-identical mystery headlines in the US top 1,000 before the update and 0 after.
- Big feeds got smaller. Reach, publisher of the Mirror and the Express, said its Discover traffic was down 46 percent in the second half of 2025.
- More publishers share the feed. Reach’s Nicola Agius reported at NESS 2026 that UK publishers with Discover visibility grew 46 percent and article volume 57 percent since January 2025.
- Not only publishers are in the feed. Marfeel counted AI summaries as 51 percent of Discover entries in the US, Brazil and Mexico.
Google itself calls Discover traffic less predictable than search traffic. Plan for that.
Four facts the editorial strategy rests on
A story lives about two days
Editorial articles stay in the feed for about two days: NewzDash measured 48.7 hours on average and a median of 44 hours in the US feed. So the editorial work that counts happens before publishing and in the first 48 hours. The timeline below is built on that.
More URLs do not buy more reach
More publishers and more articles now share the same feed, so each hit is smaller. Publishing more spreads the same readers over more cards. Google says itself that Discover traffic can change for reasons that have nothing to do with how often you publish. Frequency is not the lever.
The country decides who sees you
Discover now prefers sites based in the reader’s country. In Germany, sites from outside Germany lost much of their Discover reach with the update. No editorial trick wins those readers back; more output only spreads the remaining clicks over more stories.
The click rate is the editorial number
The feed card is a headline and an image, and the reader decides on those two alone. Shock-word and mystery headlines lose: in BILD’s own analysis, neutral titles beat clickbait, and templated mystery headlines dropped out of the US feed after the update.
The timeline: before, at and after publishing

Before publishing: topic, angle, entities, image
- Pick the topic from your own Discover data. Keep a list of stories that worked and did not, by desk and topic. BILD joined its Discover data with crawl data and moved effort to under-performing categories with growth potential.
- Copy yourself, not the competition. El Español’s audience team looks at what works inside its own paper and checks whether it carries over to other sections and regional offices.
- Give each story one purpose. At El Español every story is published with a clear intention, and stories built only on press releases or wire copy are rejected.
- Find your own angle. Google asks whether a page adds original information, reporting or analysis. An agency rewrite needs at least one thing of your own: a call, a local consequence, a calculation, a quote.
- Name the entities. Agius recommends original reporting and two to three entities per headline. Südwest Presse tagged articles with precise entities, for example “UEFA Champions League” instead of “Champions League”, and 820 tagged articles brought more than 1 million Discover pageviews. That case was written by the vendor that did the tagging.
- Plan the image with the story. The image is half of the card. My rules for the photo desk are in image language.
At publishing: the card
- The H1 is the feed headline. Write it for a reader who sees nothing else. See headlines and images.
- One large image, at least 1200 px wide. Google’s spec and code are in my technical article.
- A named author. In BILD’s data, articles with a named author earned nearly twice the clicks of those without.
- Shallow and linked. BILD also found that well-linked pages won more Discover traffic. Put the story on the home page and its topic page, and link it from related stories.
- In the news sitemap at once, with the real time. Google wants only articles from the last two days in a news sitemap.
Hours 1 to 6: read, do not fiddle
- Read impressions and clicks in your real-time analytics. Search Console shows Discover with a delay.
- Change the headline or the image only when the new version is clearly better: more specific, more current, or correcting an error.
- Never change the URL. At NESS 2026, NewzDash’s John Shehata reported that swapping headlines and URLs too often led to Discover penalties in Germany. That is a practitioner report, not a Google rule. I treat it as a reason to be careful.
Day 1 and 2: give the story a second life
The window is two days, so the second piece has to come inside it.
- The follow-up. A new development, a reaction, a consequence for the reader. New facts, new URL, linked to the first story and back.
- The takeaways piece. When a story works, I recommend a second article with its seven or ten most important points as a list, linked to the original. The Independent does this at scale with “Bulletin”: an existing story becomes five to ten bullet points, a journalist checks them, the original byline stays and the summary links to the full article. Its UK editor said those stories surface on Discover.
- The connection. Aftonbladet reporters already watched their stories in real time in 2016 and then adjusted related-article links, added video and pushed them on social. That case is older than Discover, but the routine still fits the two-day window.
Day 3 and later: out of the feed, into the archive
- The story has left most feeds. Link it from the new stories on the topic and from the topic page.
- Google says older content can still appear when it is helpful and relevant to a reader’s interests. That is the job of evergreen pieces, not of yesterday’s news.
Day 3: the aggressive reset for a story that should have worked

Sometimes a story should have worked and does not: three days in, no meaningful pickup in Discover. Against common SEO logic, and very aggressive, there is one more move that is worth trying:
- Unpublish the story. Delete it, so the URL answers HTTP 404 (or 410).
- Remove it in Search Console. The Removals tool hides the URL from Google Search within about a day, for about six months. The 404 makes it permanent: Google drops indexed URLs that answer with a 4xx code from its index.
- Update the content. New facts, a new angle, a new headline, a new image. Not the same text with a new date.
- Publish it under a new URL.
- Watch the first 48 hours. If it gets picked up, write down what changed. If not, stop.
There is no guarantee that this works. It is not a contradiction of “never change the URL” above: that rule protects a live story that is working, the reset is for a story that is already dead. Use it rarely, for stories you still believe in, and record every try with its result.
Every season: evergreen updates, same URL
- Update evergreen articles at least once a season or once a year: new facts, new image, new date, same URL. The full routine is in my evergreen section.
- Republishing is not a trick. Google explicitly asks whether you change dates to make pages look fresh without real changes. Republish when the story really moved, never only to get a second ticket into the feed.
Headlines and images
Google sets the frame: headlines should capture what the story is about, must not withhold the information needed to understand it, and must not play on morbid curiosity, titillation or outrage.
Inside that frame, these are the formulas I give desks. Each one discloses the event and keeps one honest reason to click.
| Story type | Formula | What the reader gets |
|---|---|---|
| Breaking news | Entity + what happened + consequence | The news itself, the reason to read on |
| Service | Rule or price + who is affected + date | Whether it concerns them |
| Sport | Club or player + confirmed step + what it changes | The fact, not the rumour |
| Weather | Place + phenomenon + time + impact | When and where to prepare |
| Explainer | Named event + the reader’s question | The question the article answers |
| Lost | Shock word + hidden subject | Nothing: the click is a gamble |
Three rules I add:
- Two or three named entities. A person, a place, an organisation. That is what the feed matches to interests.
- Length follows content. Agius cited about 90 to 105 characters; the US feed average rose to 84 characters after the update. Write the full thought, then cut.
- One mystery at most, and the article answers it in the first paragraph. A curiosity gap is fine. A hidden subject is not.
For the image, Google asks for large, relevant, high-quality pictures, not logos and not text-heavy graphics. My own rules: a person beats an object, action beats stillness, your own photo beats the agency photo.
Topics: depth beats breadth
Google judges expertise topic by topic. Its own example is a local news site with a dedicated gardening section.
What that means for a general news site:
- Build sections you want to win as real sections, with enough stories, their own authors and a topic page that links them.
- Stay with a topic over weeks. One story is a guess. Twenty connected stories are a history Google can read.
- Prefer depth to breadth. Agius put it in five words at NESS 2026: Discover rewards depth, not breadth.
Local relevance and readers abroad
Discover now prefers sites based in the reader’s own country. In the NewzDash panel, New York local sites appeared about five times as often in the New York feed as in the California feed.
For publishers that reach readers in a neighbouring country, an identical story for both markets is the weaker bet. Give the second market its own angle, with its own places, people, rules and sources, or accept that the feed may show your story mostly at home.
Clickbait is a Discover habit, and the habit is risky
A 2025 study of 300 headlines from El Español and La Vanguardia classified 44 percent of their Discover headlines as clickbait, against 2 percent on their own front pages. Newsrooms wrote differently for the feed. That data is from June 2025, before the update; the update targets exactly this gap.
Volume: fewer, stronger stories
- Every extra URL needs a reason: a new fact, a new reader group, a follow-up, a service. More of the same splits the same readers over more cards.
- An agency rewrite with nothing added is a cost, not a story. Google asks whether content that draws on other sources adds real value instead of just rewriting them.
- Measure per URL, not only in total. A stable total can hide that the median story earns far less.
The newsroom rhythm
Discover is a channel, so someone owns it every day. At El Español, audience specialists and editors work together through the whole day. The routine I set up with newsrooms:
| When | Who | What |
|---|---|---|
| Every morning, 15 minutes | Discover owner with desk heads | Yesterday’s top and bottom stories by desk, topic, click rate; one decision per desk |
| Before each publish | Editor | Topic check, angle, entities, H1, image, author |
| Hours 1 to 6 | Desk | Read the first numbers, fix only real gains |
| Day 1 and 2 | Desk | Follow-up or takeaways piece, links both ways |
| Weekly | Discover owner, editor-in-chief | Topic portfolio: which topics to deepen, which to stop |
| After a Google update | Discover owner | Wait until the rollout ends, then a new baseline |
The numbers to look at, per desk and per country:
- URLs with Discover impressions, against URLs published.
- Median clicks per URL, not only the total.
- Click rate, with clicks and impressions next to it.
- Share of clicks on day 1 and 2.
- Stories with and without shock words.
Search Console’s Discover report covers 16 months and only shows data above a minimum number of impressions. Store the data yourself if you want a longer history. During a rollout, do not act on single days: the February 2026 update took 22 days, Google had said up to two weeks.
What to stop doing, and why your own data decides
With the February 2026 update, Google said Discover will show more local, less sensational and more in-depth content. Turned around, the update targets three things:
- Content from sites outside the reader’s country.
- Sensational content and clickbait.
- Shallow, unoriginal stories from sites without expertise in the topic.
That Google says it does not make it my recommendation. Google’s definitions are fuzzy: what counts as clickbait, local or expertise is not spelled out, and what Google says does not always match the data of real websites. My recommendation is always the same: look at your own data. Count, per desk and per topic, what your shock-word headlines, your extra URLs and your rewrites actually earned this year against last year. Stop what your numbers say lost, keep what they say still works.
My list. Check each one against your own data:
- Shock words and mystery templates in headlines.
- More URLs as the answer to falling traffic.
- Agency rewrites with nothing of your own.
- Swapping headlines and URLs every hour.
- Republishing unchanged stories with a new date.
- Generic stock or logo images on the card.
- Treating Discover as a promise. It is a feed you earn every day.
Treat every rule as a test
No one has a settled best practice for the feed after the 2026 changes. Search optimisation has had more than twenty years to settle; Discover changed its rules this year. So I work with hypotheses and tests, and every rule above is one.
A test that teaches you something:
- One change. For example: headlines with two named entities against one.
- One desk. The other desks are the control.
- A start date and a stop rule. Four weeks, or until the next Google update.
- The right number. Click rate and median clicks per URL, with absolute counts.
- A written result, into the list of what worked and what did not.
Every newsroom needs its own version of this timeline, written together with its editors. That is what my Google Discover & News Feed Workshop builds.
Sources
All read on 2026-11-20.
- Google Search Central: Discover and your website
- Google Search Central Blog: Google’s February 2026 Discover Core Update, 5 Feb 2026
- Google Search Status Dashboard: February 2026 Discover update, 5 to 27 Feb 2026
- Google Search Central: Creating helpful, reliable, people-first content
- Google Search Central: Create a News Sitemap
- Search Console Help: Removals tool
- Google Search Central: How HTTP status codes affect Google’s crawlers
- NewzDash, John Shehata: Google Discover Feb 2026 Core Update Scorecard, 20 Feb 2026
- NewzDash, John Shehata: 99 News SEO Lessons I Learned at NESS 2026, 9 Nov 2026
- Press Gazette: Plunging Google Discover traffic hits Reach digital revenue, 3 Mar 2026
- Marfeel: Google Discover’s shift to AI, YouTube and X
- Oncrawl: BILD improves performance in Google Discover, 16 Jun 2026
- INMA, Johannes Sommer: Südwest Presse boosted its Google Discover reach by 300% using AI-driven content tagging, 14 Dec 2025
- Opennemas: Interview with Álvaro Mazariegos and Gonzalo Mirasierra, El Español, 25 May 2026
- Press Gazette: The Independent’s AI Bulletin, 3 Apr 2025
- WAN-IFRA: Tabloid Aftonbladet is putting its online audience in charge, 27 Jan 2016
- Infonomy: Clickbait on Google Discover, 2025
Frequently asked questions
How long does a story stay in Google Discover?
About two days. NewzDash measured editorial articles at 48.7 hours on average in the US feed, median 44 hours.
Does publishing more articles bring more Discover traffic?
No. More publishers and more articles share the same feed, so each hit gets smaller. Google itself says traffic changes can be unrelated to publishing frequency.
Should I republish an article to get it back into Discover?
Only when the story has really moved on. Google asks whether you change dates to make pages seem fresh when the content has not substantially changed, and German publishers reported Discover problems after swapping headlines and URLs too often.
What headline works in Discover after the February 2026 update?
Name the entity, the event and the consequence for the reader, and keep one open question the article answers. Google tells you not to withhold crucial information, and a US panel saw templated mystery headlines drop out.
Do I need a Discover desk?
You need someone who owns the Discover numbers every day and editors who act on them. Whether that is a desk or a role on each desk depends on your size.