February 2026 Discover Update: 67% Traffic Shifts Explained

If your Google Discover traffic did something dramatic in early February 2026, either falling off a cliff or suddenly spiking, you weren't imagining it. On 5 February, Google rolled out the first Discover-only core update it has ever announced, and it reshuffled the feed for a lot of publishers before finishing on 27 February.

We run Discover traffic for clients across news, local services and ecommerce, so we watched this one land in real time. This is what actually changed, who it hit, and what we'd do about it if your numbers took a knock. No invented percentages, no scary made-up stats. Just the verified detail and our honest read on it.

What was the February 2026 Google Discover update?

It was a core update aimed only at Discover, not Search. Google confirmed it as "a broad update to our systems that surface articles in Discover", which is corporate-speak for "we've changed what the feed shows you and your readers".

Three things Google said it set out to do: show people more locally relevant content from sites based in their own country, reduce sensational content and clickbait, and surface more in-depth, original, timely content from sites with genuine expertise. That's straight from the announcement, not our paraphrase.

The rollout took its time. Search Engine Land confirmed the update started on 5 February and completed on 27 February 2026, a 21-day run. It launched first for English-language users in the United States, with other countries and languages coming in the months after. So if you're outside the US, you may have seen the effects roll towards you on a delay rather than overnight.

One detail we love, because it kills a lazy myth. Google said expertise is judged topic by topic. Their own example: a local news site with a dedicated gardening section can build gardening expertise even though it covers everything else, while a movie review site that publishes one gardening article probably can't. You don't have to be a single-subject blog to win here. You have to actually know your stuff in the lane you're writing in.

How much did traffic really move?

Here's where we have to be blunt. You'll see "67% of sites lost traffic" type numbers floating around (an earlier version of this very page quoted one). We can't find a credible primary source for figures like that, so we're not repeating them. When a stat has no traceable origin, it's usually because someone made it up and everyone else copied it. That's exactly the kind of low-trust content this update was built to punish, so it would be a bit rich for us to lean on it.

What Google actually told publishers to expect was refreshingly plain: "Some sites might see increases or decreases; many sites may see no change at all." That matches what we saw across our own client set. Plenty of sites barely moved. A handful swung hard in one direction.

The pattern in the swings is the interesting bit. In our campaigns, the sites that gained were the ones publishing genuinely local, first-hand content with a clear author behind it. The sites that dropped were the ones running engagement-bait headlines and thin rewrites of other people's stories. That lines up neatly with what Google said it was targeting, which is reassuring, because Google's stated intent and its actual behaviour don't always agree.

If your Discover numbers cratered while your normal Search traffic held steady, that's a useful diagnostic. It usually points to a content-quality or headline problem rather than a technical one. We've written before about how brutal these drops can be in our survival playbook for small publishers who lost 60% of their Google traffic, and a lot of that thinking applies here too.

The clickbait crackdown, and where Google drew the line

Google didn't just change the algorithm in the dark. On the same day, it quietly rewrote its Discover guidelines, and the edits tell you exactly what it's now grading.

According to Search Engine Journal's breakdown of the documentation changes, the word "clickbait" now appears explicitly in Google's Discover docs for the first time. Previously the guidance just said "avoid tactics to artificially inflate engagement". Now Google has split the old combined bullet into two: one telling you to use titles and headlines that capture the essence of the content, and a separate one telling you to avoid clickbait outright.

The current official Discover documentation spells out what clickbait means to Google: using misleading or exaggerated details in your preview content, meaning the title, snippet or image, to inflate appeal. It also warns against withholding the crucial information someone needs to understand what the content is even about. The classic curiosity-gap headline, in other words.

The sensationalism section got tightened too. Google folded its old wording about "morbid curiosity, titillation, or outrage" into a cleaner line about avoiding tactics that manipulate appeal. Same intent, sharper teeth.

What actually works in headlines now

We tell clients the test is simple. Could a reader predict what they'll get from your headline and image alone, and would they feel the page delivered it? If yes, you're fine. If the headline writes a cheque the article can't cash, you're exposed.

• Be specific over clever. "Where Chicago deep-dish actually comes from, and the three spots locals still argue over" beats "You won't believe the truth about pizza".
• Match the image to the story. Google explicitly counts your preview image as part of the promise. A misleading thumbnail is now treated the same as a misleading headline.
• Don't hide the payoff. Withholding the core fact to force a click is the exact behaviour the new guidance names.

If you've ever wondered whether you can outsmart this stuff, we have a slightly embarrassing case study. We once published a post about a completely fake Google update and watched it rank on page one. Funny, yes. Sustainable in a feed that now actively grades trust signals? Absolutely not.

Local relevance and expertise: the part everyone misreads

"Local relevance" got misread the moment the update landed. It does not mean stuffing a city name into your title and calling it a day.

From what we see, Google is weighting whether the content reads like it came from someone who actually understands the place or the subject. A site based in the reader's country, covering things that matter to people there, written by someone with evident first-hand knowledge. That's the signal. Geographic name-dropping without real understanding doesn't move the needle, and we've tested that more than once.

The expertise side is where E-E-A-T does the heavy lifting. Google's guidance on people-first content sets out the questions it wants you to ask: does the content provide original information, reporting or analysis, is it written by someone who demonstrably knows the topic, and does it give substantial value beyond simply copying other sources. Google is also unambiguous about priorities, stating plainly that of Experience, Expertise, Authoritativeness and Trustworthiness, "trust is most important". The rest exist to support it.

In practice that means real author bylines with real credentials, original data or first-hand experience, clear sourcing, and no easily-checked factual errors. None of that is exotic. Most sites just never bothered because the old feed rewarded clicks over credibility. That bill has now come due.

This is also why we're sceptical of pure volume plays. Spinning up a hundred near-identical articles to blanket a topic is the opposite of what this update rewards, which is why we're cautious about how Google treats AI-generated content at scale. The tooling isn't the problem. Publishing thin, expertise-free content with a human-shaped byline slapped on top is.

Does the Discover update affect normal Search rankings?

Short version: not directly, but the underlying quality signals overlap, so don't expect a firewall between them.

Discover and Search run on different systems with different jobs. Google's own documentation describes Discover traffic as less predictable than keyword-driven search and tells you to treat it as supplemental, because it fluctuates with user interests and feed changes. A Discover swing on its own doesn't mean your Search rankings have moved.

That said, the helpfulness and trust signals feeding both systems are largely the same. If your content got filtered out of Discover for thin expertise or clickbait, those same weaknesses are unlikely to be helping you in Search. We saw this clearly in the wider volatility around the March 2026 core update winners and losers: the sites that fixed content quality after February tended to hold up better through the next round.

The reverse doesn't run automatically. A local-relevance boost in Discover won't necessarily lift your Search positions for the same topic. Different algorithm, different priorities. But fixing your content properly for Discover usually improves it everywhere, which is the closest thing to a free lunch this job offers.

A 30-day recovery plan that doesn't rely on guesswork

If you took a hit, resist the urge to panic-rewrite everything at once. Work it like a diagnosis, fix the cause, then measure. Here's the sequence we actually use with clients.

1. Days 1 to 7: find the damage. Open the Discover report in Google Search Console and segment your last 90 days. Identify exactly which articles lost Discover impressions and clicks, and look for the common thread: headline style, topic, content depth, author presence. Be honest about which pieces used curiosity gaps or mismatched images. If you want help reading the data, the newer features inside Google Search Console make this segmentation a lot less painful.
2. Days 8 to 21: fix the cause, not the symptom. Start with your highest-traffic losers. Rewrite headlines so they describe the content accurately and specifically. Swap any misleading preview images. Add genuine expertise markers: a credentialed byline, original data, first-hand detail. Check your images meet Google's Discover spec while you're in there, since the docs ask for at least 1,200px wide and a roughly 16:9 ratio.
3. Days 22 to 30: measure and double down. Watch the Discover report daily for the revised pieces. Recoveries here often show within a few days of republishing, not weeks. Track time on page and scroll depth too. If engagement improves but Discover impressions don't, your content got better but your headlines or images still aren't pulling. Iterate on whatever's recovering and apply the winning pattern to new pieces.

If you're 30 days in and still flat, the problem is usually deeper than headlines, and it's worth treating it like a full algorithm recovery. We laid out that broader approach in our guide on what to do after the March 2026 spam and core update, and most of it transfers cleanly to Discover.

Why this matters more than a normal feed tweak

Here's the part that makes the February update worth taking seriously rather than shrugging off. Discover is no longer a nice-to-have side channel for a lot of publishers.

When Google confirmed Discover is coming to desktop search, Search Engine Journal noted that 52% of news publishers now prioritise Discover and 56% had seen recent traffic increases from it. Martin Little of Reach plc went further, calling Discover their "single largest traffic referral source". So an update that resets the rules of the feed, right before that feed expands to desktop, is a big deal for anyone leaning on it.

Our take, for what it's worth: this is one of the more sensible updates Google has shipped. It rewards roughly the things you'd want a content discovery system to reward, real expertise, honest headlines, content that's relevant to where the reader actually lives. If you've been doing the work, it's good news. If you've been gaming engagement, the runway just got a lot shorter, and desktop will only widen the gap.

This is the work we do every day, so if your Discover or Search traffic moved and you can't pin down why, that's squarely our SEO service territory. You can get in touch with our team and we'll tell you straight whether this update hit you or whether something else is going on. We'd rather give you an honest read than sell you a recovery you don't need.

The February 2026 Discover update separated publishers who know their subject from those who were renting clicks. The genuinely useful, genuinely local, genuinely expert content won. Build that, and the next feed change becomes a lot less frightening.