Evergreen content is losing. Here is what is winning instead.

For fifteen years the advice to a small site was the same: write evergreen content. Pick the questions your buyers ask, answer them properly, and let the articles earn clicks for years while you sleep. This month the publishers who did exactly that are reporting that evergreen content is the category losing hardest to AI Overviews, and that the three things gaining are real-time coverage, Google Discover, and placement somebody paid for. That is a strange list, and it is worth taking seriously rather than filing under "publisher problem".

What the publishers are reporting

The numbers come from the reports collected by Press Gazette and Search Engine Land in the first week of September. Search clicks are down 42% from baseline across the publishers surveyed. On queries where an AI Overview appears, click-through to the pages beneath it is down 61%. And the pages losing most are the ones written to be timeless: the explainer, the "what is", the how-to that ranked for three years.

The reason is not mysterious. Evergreen content answers a stable question in a stable way, which is precisely the material a model can summarise with confidence. An Overview above a "what is a second-price auction" result can say what a second-price auction is, cite two pages, and send nobody anywhere. The article is still the source. It is no longer the destination.

AdExchanger puts it more bluntly: the open web's informational traffic is being dismantled and may not come back. We would not go that far for a small site, for a reason we will get to. But the direction is not in dispute.

The three things reported as growing

Real-time. Coverage of something that happened in the last few hours is hard to summarise from a crawl that finished last week, and Google still surfaces it in news carousels where an Overview does not sit. Publishers report the share of traffic from time-sensitive pieces rising even as the evergreen archive sinks. The catch is obvious: real-time content expires in real time. You are on a treadmill.

Discover. Google's feed of recommended articles on phones, which is not search at all — there is no query to answer, so there is nothing for an Overview to replace. Publishers report Discover holding or growing while search shrinks. The catch is that Discover is opaque, unpredictable and entirely at Google's discretion, which is a poor foundation for anyone.

Bought placement. The third one is the odd one out, and the one that matters for a small site. The same reports name sponsorships and paid placement — traffic bought on a page people already visit — as holding up, because it never depended on a results page. A newsletter sponsor gets the same opens whether or not an Overview exists. A slot on a page people go to directly sends what it sends.

Notice what the three have in common. None of them is "a page that ranks for a stable query". Two are Google products with terms Google sets; the third is a transaction with a known price.

What a small site should actually do

The publisher advice does not transfer cleanly. A site run by three people cannot staff a real-time desk, and Discover mostly rewards sites with an existing audience and a lot of images. So here is the version we would give an indie founder, with the evergreen part kept.

Keep writing evergreen content, but change what it is for. An article that answers a stable question is still the thing a model cites, and being cited is not nothing: the citation carries your name, and some fraction of readers open it. What it no longer does is deliver the traffic on its own. Treat it as the reference that makes the other channels credible, not as the channel. The post on being cited covers what a citable page looks like.

Pick two channels that are not search, and run them properly. Our survey of indie marketing channels ranks what founders report working; communities and direct outreach are near the top, and neither cares about Overviews. Run two, not seven — the two-channel argument is made at length elsewhere.

Buy a small amount of placement, and read the numbers. This is the one growing channel that is open to a site with fifty dollars, and it has an honesty advantage the other two lack: a bought slot reports what it sent, in units you can compare with a search click.

One option is ours, so weigh it accordingly. BidSurvivor sells two twelve-hour slots a day on its own front page. The first brand into an empty slot pays $0, no account needed, one free claim per bidder per day; taking a held slot means beating the standing bid, from $1.00 up to a $5,000 ceiling; every visitor starts with $100 of house credit before signing up; and if you are outbid you lose the slot and the money, while a bid that fails to take the slot costs nothing. What you get afterwards is a public count of opens and click-throughs, on the front page and on your own brand page. The board is young, and its numbers are small and published, so read them before you decide it is worth a dollar. It is one bought placement among many, and the only reason it is in this paragraph is that it is the only one whose counting we can vouch for.

Why "losing" is the wrong verb for a small site

Here is the reason we would not join AdExchanger at the funeral. The 42% is a loss against a baseline. Publishers had the baseline. A site launched this year never had it, so it has not lost anything; it is simply arriving in a world where the evergreen channel pays out at a lower rate than the one in every 2019 playbook.

That changes the plan rather than ending it. Evergreen content used to be the whole strategy; now it is the reference layer under a strategy that has to get its visitors from somewhere else. Real-time and Discover are the big publishers' somewhere else. For the rest of us it is the room where our buyer is asking, the list we own, and a placement we bought at a price we knew — the things that were never search, and that nobody can switch off in a release note.

Write the evergreen piece. Just do not wait for it.