A publisher I advised had a post do 340 visits in its first three weeks from search. On day 24 it did 51,000 in eleven hours. Then it stopped, permanently, and the article went back to doing forty visits a day.
That is Google Discover. It is a feed, not a search engine, and treating it like search is why most brands never crack it and why the ones that do get destroyed when it turns off.
I want to give you both halves: how to actually earn Discover placement, and why you should never let it become more than a third of your traffic.
Discover is answering a different question
Search answers "what did this person ask for." Discover answers "what would this person want to see right now, given everything we know about them, without them asking."
That is an interest match, not an intent match. Which flips almost every content instinct you have.
Keyword targeting is largely irrelevant. Nobody typed anything. Search volume is irrelevant. Ranking position does not exist. What matters is whether the piece is topically adjacent to things this specific person engages with, whether it feels current, and whether the presentation earns a tap in a scrolling feed.
What actually gets picked up
From the accounts I have watched closely, five properties show up over and over.
Timeliness without being news. Pure breaking news is dominated by publishers with authority you will not out-earn. The sweet spot is the piece that is clearly of this moment but not a wire story: an explainer on something that just changed, an opinion on something everyone is discussing, a practical response to a new product or policy.
A genuinely interesting entity. Discover leans heavily on entities the user follows or engages with. Articles that are about a well defined thing (a named company, product, person, place, event) get distributed to people interested in that thing. Vague thought pieces about "the future of work" have no entity to attach to and go nowhere.
A large, high quality, non stock image. This is not a minor factor. Discover is a visual feed. The guidance is a wide image at least 1200 pixels, and in practice the difference between a stock photo and a distinctive original image is the difference between a 2 percent tap rate and something several times that. Screenshots, charts you made, real photographs. Anything that does not look like it was chosen from a dropdown.
A headline that promises a specific payoff. Not clickbait, which gets suppressed, but specificity. "What Changed in the New Update" loses to "The New Update Quietly Broke Scheduled Posts." I run headlines through a headline analyzer not because the score is gospel but because it forces me to generate eight options instead of shipping the first one.
Mobile page experience that does not fight the reader. Discover traffic is overwhelmingly mobile and overwhelmingly impatient. Interstitials, autoplay video, and a slow hero image kill the session before it starts. The performance baseline in Core Web Vitals in practice is the price of entry here, not a nice to have.
The E-E-A-T layer that gates everything
Discover is a recommendation surface pushed to people who did not ask, which makes Google conservative about who it trusts there. Sites with anonymous authorship, no clear ownership, and no track record in a topic mostly do not get in.
Real author pages with real credentials. A clear About page with a named organization. Consistent publishing in a defined subject area rather than random topic hopping. Site level entity clarity, which is the same work I argue for in E-E-A-T explained.
I have seen sites go from zero Discover impressions to consistent daily impressions after nothing changed except adding real bylines and cleaning up their About page. That is not a growth hack, it is a threshold being crossed.
What gets you removed
Misleading headlines relative to content is the fastest way out, and the punishment is site level rather than article level.
Aggressive ad density above the fold. Discover users bounce fast on ad walls and that behavior feeds back.
Topic incoherence. A site that publishes finance, celebrity gossip, and gardening confuses the interest model and gets distributed to nobody.
Recycling the same angle repeatedly. The feed heavily discounts repetition. Ten posts on the same news event from slightly different angles trains the system that you are farming.
The business risk nobody warns you about
Here is the part I care about more than the tactics.
Discover traffic converts terribly. It is interrupt traffic from people who were not looking for you. In the accounts I have looked at, Discover visitors convert at a fraction of the rate of search visitors on the same page, often several times worse. It is real audience but it is cold audience.
Worse, it is unstable in a way that no other channel is. Search rankings decay. Discover just stops. There is no diagnostic, no ranking to check, no gradual decline to react to. One day the impressions go to near zero and there is no support ticket to file.
I have watched a founder hire against Discover revenue and have to unwind the team eleven weeks later. That is a business failure, not a Google failure. They built fixed costs on a variable channel.
How to use it correctly
Treat Discover as an audience acquisition top of funnel and nothing else.
Every Discover-optimized post should have exactly one job: convert a fraction of that spike into an owned relationship. Email signup, most of the time. Not a demo request, not a purchase, because the audience is too cold. Something with a low commitment threshold and a real reason to say yes.
Instrument it so you can see it. Tag the internal links and campaigns coming off those posts consistently so that six months later you can answer the only question that matters: how many subscribers did the spike produce, and what were they eventually worth.
Then move those people into a channel you control. This is exactly the argument for building a newsletter operating system rather than renting attention, and it is the same lesson as the death of organic reach on social platforms, arriving one platform later.
My allocation rule
No client of mine is allowed to plan a business where a single algorithmic feed provides more than about a third of sessions. When it creeps past that, we deliberately invest in the slower channels: evergreen search content, email, direct.
That rule has cost clients traffic in good quarters. It has also meant that when the feed turned off, and it always eventually turns off, revenue moved by single digit percentages instead of collapsing.
Chase the spike, bank the subscriber
Discover is worth pursuing. Original images, timely topics with clear entities, honest specific headlines, real authors, fast mobile pages. That is the whole recipe and it is not complicated.
What is hard is the discipline afterward. The correct response to a 51,000 visit day is not to publish twelve more posts on the same topic. It is to look at how many of those people you converted into something durable, and if the answer is under 1 percent, fix your capture before you chase another spike. The spike is not the asset. The list you build from it is.