Google Discover across 49 client sites: 0.09% of clicks, and 95% of them came from a single article in four days

At the Search Central Live Deep Dive in Barcelona, Discover got its moment: Google's team went over how a page gets into the feed, what image it needs and why more and more low-quality AI-generated content is arriving. All useful if you are a publisher. But the question our clients ask is a different one: how much traffic does Discover give a business site?

I measured it on our portfolio. 49 client sites with Search Console, 90 days, and the answer fits in one number: 0.09% of clicks.

What Google said in Barcelona (and what the documentation says)

• There is no Discover crawler of its own. The feed comes out of Google's normal index: if a page is indexed, it can appear. There is nothing to switch on and nothing to submit.
• The image rules the card. For the large version, the image must be at least 1,200 pixels wide and exceed 300,000 pixels in total, and the page has to allow it with max-image-preview:large in the robots meta. The official documentation, updated in March, says the same.
• No bait headlines. Discover penalises clickbait and sensationalism, and in Barcelona the talk turned to the rise of mass AI-generated content trying to slip into the feed.
• It is measured separately. Search Console has its own Discover report, separate from web search performance, and it only shows up once the property reaches a minimum of impressions.

None of this is new. What no talk tells you is how much Discover weighs when the site is not a newspaper.

How much Discover weighs across 49 business sites

I checked every active client property we have Search Console access to: online stores, comparison sites, education, tourism, industry and services. From 6 July to 3 October.

42 of the 49 sites have not a single impression in three months. And it is not for lack of size: the two biggest sites in the portfolio do not appear either.

The reverse does hold. The seven that show up in Discover are all among the 16 with the most organic clicks, and the smallest of them gets around 35,000 clicks in 90 days, more than three times the portfolio's median site. Below that threshold, none. Volume looks like a condition for getting into the feed, but it does not guarantee it.

One article, four days

This is how those 2,899 clicks break down:

And within the first one, almost everything is a single page: a blog post about which vocational course to study when you are over 40. 2,747 clicks, 95% of all the portfolio's Discover.

It was published on 7 July. This is its whole life in Discover:

Eight days of life and four of peak. From 10 to 13 July it brought in 2,597 clicks, 95% of everything the article did. After that, nothing: not one impression in almost three months. That is Discover: a spike that does not repeat and cannot be planned.

The topic is no accident either. It is not a product page or a service page: it is a human-interest story (reinventing yourself at 40) that someone reads without searching for anything. It is exactly the kind of piece Discover shows, and the kind almost no business site publishes.

The image does not explain who gets in

If the image is what weighs most on the card, you would expect the pages that meet the requirements to be the ones taking the traffic. I checked the page with the most Discover on each site: whether it allows the large image and which preview image it declares.

The page that brought the most traffic has an empty og:image tag. Google did not need it: the article does allow the large image and carries several images over 1,200 pixels wide in its body, and Google picked one of those. Meanwhile, the only page that ticks every box, a 2,560-pixel image and permission to show it large, got zero clicks.

The image is the ticket to the large format, not what gets you into the feed. What decides is whether the content interests someone who was not looking for it.

My recommendation: do the technical minimum, and leave Discover out of the plan

• Do not treat it as a channel for a business site. In our portfolio it weighs 0.09% and depends on a four-day stroke of luck. If a client asks why they are not in Discover, the honest answer is that almost nobody in their category is.
• Do the minimum, it costs little. Add max-image-preview:large to the robots meta, use at least one image 1,200 pixels wide in your articles and fill og:image with it. That way, if an article gets in, it shows with a large card.
• If you have a blog, have human-interest pieces. The only article that worked was not about the product, but about the person who would buy it. It is the same logic Discover asks for: something you would read without having searched for it.
• Isolate the spike before you credit it. Analytics has no Discover channel, so those four days slip into another one. Cross the dates with Search Console's Discover report before crediting a rise to SEO.

Methodology: Search Console data from 49 client properties with active access, from 6 July to 3 October 2026, web search type versus Discover. Images and tags checked on 6 October on each site's page with the most Discover impressions. Clients are anonymised.