The thesis
Stay calm and monetize
SEO is not dead. Display ads still make money. People still go to websites. Three sentences, each one currently unfashionable, each one measurable. This is the essay behind the name on the door: the case for treating 2026 like an engineering problem instead of a funeral, argued from our own Search Console and AdSense records.
The numbers behind the claims live at the receipts. The placement principle the name comes from is the manifesto.
By Mario Bailey · Published
The eulogy
The funeral is also a sales funnel
SEO has been pronounced dead for as long as we have run websites, but this year's eulogy has better props than usual. An AI answer really does sit above the results now, and on some query classes the clicks really did fall. We are not going to argue with the grain of truth. We are going to size it.
First, notice who is speaking at the funeral. The people declaring SEO dead are, with remarkable consistency, selling something only a panicking publisher buys: the AI-proof course, the pivot framework, the fast exit at a distressed multiple. Panic is a market, and it pays better than patience. That does not make the eulogy false by itself. It does mean the argument should be settled with data, and the sellers never bring any of theirs. So here is ours.
The honest half
What actually died
Commodity information is losing the click. A generic question that a model can answer from its training data increasingly gets answered above the results, and the best published measurement of the cost is blunt: in Pew Research Center's study of real browsing behavior, users who saw an AI summary clicked a traditional result on 8% of visits, against 15% when no summary appeared. Half the clicks is a real loss, and we do not soften it. How to measure your own exposure, query class by query class, is in the AI Overviews guide, along with that study's caveats.
The other casualty is the page that never deserved the visit. We measured this on our own sites, in as close to a controlled experiment as the field ever hands you: two sites, same owner, same stack, same months of 2026. One was a benefits-calculator site where nearly every page, about two hundred of them, was a permutation of the same calculator, pages that existed because a keyword tool said the phrase got searched. Four months in the index earned that site two impressions. Not two thousand. Two. The other site published lender-level data from a federal small-business loan program: real entities, verifiable numbers, pages that exist because the thing they describe exists. On a brand-new domain, hundreds of its pages had impressions within weeks, dozens of them holding top-ten positions.
That is what died: the commodity answer and the keyword permutation. The eulogies rarely say so, because the easy part was the product most of the industry sold.
The receipts
SEO is not dead
The lender-data result is the part worth staring at. It happened in 2026, in the middle of the AI Overviews era, on a domain with no age, no links, and no reputation. Google still found the pages, still indexed them, still ranked them, and still sends people to them. Whatever SEO is now, a brand-new site can still do it, in the hardest year anyone has had to do it in.
The pattern in our data is consistent: queries survive in proportion to how hard they are to answer from training data. Local, fresh, numeric, entity-specific. A model does not know which lenders actually wrote loans in your county last quarter, or what a state's fee schedule says this morning. The searcher with that question still needs a page, and rankings on those queries still pay in clicks, not just impressions.
Even on the query classes the summaries hit, read the measured numbers instead of the mood. Eight percent is not zero. Clicks per search fell; the click did not go extinct. If your traffic dropped this year, the calm move is to find out which thing actually happened to you, because a core update and AI absorption leave different fingerprints and demand different responses.
The demand side
Display ads still make money
Advertisers did not leave the open web. They stopped paying for inventory nobody sees, which is a different event, and one you control. The clearest receipt we own: the same ad inventory moved from the page bottom, where it measured 28% viewable, to a dismissible sticky anchor, where it measured 88%. Fewer ads on the page, more revenue from them. That result is the first receipt on the proof page, and the gap has held across our network.
The demand is not just present, it is rising where the audience is worth reaching: on the B2B data site in our network, RPM better than doubled between mid-June and mid-July, on our own AdSense records. And when one of our sites sent the wrong classification signal and RPM fell by roughly three quarters, the fix was not more ads: removing the bad signal brought the category and the rate back. The demand had been there the whole time, bidding on someone else's correctly labeled pages.
Display alone has a ceiling, and this site says so plainly elsewhere. But a ceiling is not a corpse.
The visit
People still go to websites
Underneath the other two claims is the one that sounds too obvious to defend, so almost nobody defends it. A person with a real problem still ends up on a website, reading. Our own Search Console says so: one of our B2B sites finished this spring's update cycle, the gains and the givebacks included, with weekly clicks at a multiple of where it entered it.
The deeper reason is intent. The reader whose question is urgent, personal, and specific does not stop at a one-paragraph summary. They read everything, cross-check it, and come back. Those readers were always the ones worth having, and they are the ones a summary cannot satisfy.
And the sentence has a second half the funeral crowd skips: people still go to websites they know. The visit that starts from an email list, a bookmark, or a channel you own was never Google's to take. A website is a place, not a ranking, and the calm build has always included a second door.
The posture
What calm looks like
Calm is not denial. Denial skips the measuring; calm starts there. In practice it is a short list.
- Measure before mourning. Segment your own Search Console by query class before accepting anyone's story about your traffic, including ours. The method is written down.
- Diagnose before deciding. An update loss, AI absorption, and an RPM collapse can look identical in a revenue chart and have three different fixes. The triage quiz names the failure family in six questions.
- Build what still earns the visit. Real entities, verifiable data, honest editorial on top. Retire the commodity pages without ceremony.
- Place calmly. Viewable, reserved, dismissible, and few. The test is two bars, and every unit clears both or does not ship.
- Own a second door. Email from day one, and one channel that is yours.
Stay calm and monetize is not a slogan of denial, and it is not nostalgia for 2019. It is an operating instruction: the panic is somebody else's product, your own data beats their eulogy, and the boring, measurable work still pays. It paid this quarter, on our sites, and we showed the receipts.
Methods: every figure here is from our own Search Console and AdSense records, observed in 2026, with sites described by niche and never named, the same convention as the receipts. RPM movements are stated as relative changes only. The Pew Research Center numbers, their caveats, and the live-checked source are on the AI Overviews page.
If your year has been the bad kind
The calm posture is easiest to recommend from a good quarter, so the last word goes to the reader having the other kind. Six questions will name what actually broke, and the audit takes your setup apart on your own numbers, not on anyone's mood.
Get the audit Take the triage quiz