What Comes After Search Is What Came Before It
/Condé Nast and USA Today are planning as though Google sends them nothing. Marlon reads The Ringer, Tedium and Self-Host Weekly, and works out which pre-search discovery tools a one-person site can still run.

TL;DR: Jason Kottke’s one-paragraph link post points at Katie Baker’s Ringer piece, where Condé Nast CEO Roger Lynch says he told his teams “Assume there’s no search” and USA Today CEO Mike Reed says “our plan is to assume search is gone and build our business in other ways.” SparkToro’s Amanda Natividad: “Sixty-eight percent of all U.S. Google searches end without a click,” up from 45 percent a decade ago. Business Insider’s search traffic is down 85 percent. When an index stops delivering readers, discovery falls back on what carried it before: webrings, blogrolls, forums, the link post. Every one of those is a person putting their name next to somebody else’s work. A one-person site can run all four by Friday.
I set classified ads for a living. A classified is the oldest discovery machine we have, one line of agate telling a stranger that a thing exists, paid for by whoever wants it found. Nothing ranked it. Somebody read down the column and went and looked.
The publishers already wrote the obituary
Skip the decline numbers in Baker’s reporting. The exits are the story.
Reed says USA Today is “getting close now to the point where we’ll block Google as well.” Neil Vogel, CEO of People Inc.: “Turning them off and blocking them entirely is 100% on the table.” On September 15, Cloudflare changes its own settings to block Google’s crawlers by default.
The small operators put it plainer. The recipe blog Inspired Taste calls Google’s business model a “hostage situation.” Nicolas Bouliane, who runs All About Berlin by himself, told NPR his traffic is down 75 percent since AI Overviews launched in 2024, and posted in late July that “At this pace, All About Berlin will not survive into 2027.”
The largest publishers on the internet are budgeting for zero. The smallest are watching a clock. So what did people use before the index existed?
Every pre-search tool was one person vouching
Four of them. They all still work.
A webring is a loop of sites, every page carrying prev and next links to its neighbors, so every page tells a stranger you belong to it, instead of one page off to the side saying so. Joe wrote up how the 1995 version worked and how the one running here was built back in July, so I will not re-explain the plumbing. A blogroll is a standing list of people whose work you will answer for in public. A forum is a room with a subject in it and nobody’s algorithm. The link post barely qualifies as a format. One paragraph. Go read this.
The proof is sitting in the comments under Kottke’s post. Sixteen of them, most answering Peter Benjamin: “What are kottke-ites using for search engines these days?” The thread coughed up DuckDuckGo’s no-AI subdomain, Kagi, and a Chrome extension Kottke handed over himself, the one that makes the no-AI version your default. Mike Riley diagnosed it: “Google has gradually transformed into a store. If you search for anything that can be purchased or sold the first couple pages of links will all go to retail outlets.”
Sixteen people on a personal blog answered a search question better than the search engine did. Nothing ranked that thread. No schema markup, nobody optimizing for an answer engine. It worked because every name in it stood behind what it pointed at, in front of a room that would remember.
Cory Doctorow explains why this survives on the web and nowhere else. Every countermeasure in his piece on dickovers works in a browser and dies in an app: “An app is a webpage wrapped in the right kind of IP law to make it a felony to change how it works, which is why companies are infinitely horny to get you to use their apps, not their websites.”
That is the license the whole revival runs on. A ring is a list of links pointing at each other, and on the web nobody has to approve it. Try assembling one out of five Instagram accounts.
The counterweight, and it stings
Three straight issues of Ethan Sholly’s Self-Host Weekly will cure anyone of pitching the alternative as a paradise.
July 24: three Jellyfin team leads gone inside a week. The co-founder quit on a Friday and the other two walked out behind him. Sholly says the co-founder burned out and clashed with the people running the project, “including pushback for a CLAUDE.md file in violation of Jellyfin’s LLM guidelines.” Same issue, Codeberg bans vibe coded projects from its public git platform. July 31: the File Browser team announces the project is shutting down, archived September 1.
Then August 7. Sholly vibe coded two apps for his own homelab, a Kindle progress syncer and a thing that turns YouTube channels into podcast feeds, and he will publish neither. One stated reason: “My skin isn’t thick enough for the wrath of this community.” He closes on a line that should worry anybody betting on decentralized discovery: “the stigma associated with vibe coding has left me wondering what else I’ve been missing out on and what others have built that isn’t discoverable.”
The man whose entire job is surfacing self-hosted software is sitting on two pieces of it because the community would skin him for the tooling. Google’s index has company. Only this time the wound is self-inflicted, and the people inflicting it could stop tomorrow.
Five moves, ordered by how fast they pay
Split your analytics into search and not-search. One number from organic search, one from everything else. The second is your actual business, and most solo operators have never looked at it by itself.
Put your ring badge on every page. A badge on one page about the ring is a badge nobody sees. On this site it renders in the header and the footer of every page, because that is where a ring badge has always belonged.
Give your blogroll a roof. A list of names is wallpaper. A list that hands out a working OPML file is a tool, because a reader can import the whole thing in one click and never need you again. That is the point.
Write link posts. Kottke’s was one paragraph and it moved more people than a month of keyword pages will.
Cut the keyword page to make room. A page written to answer a query instead of a person is raw material for an AI Overview now. Per The Telegraph’s figures in Baker’s piece, ranking first used to earn a click 27 percent of the time; with an Overview above it, 11 percent. You would be feeding the summary that replaces you.
Pick one. Set it this week.
The part I would argue with
Ernie Smith’s case for forums is where I break with the source. He argues we walked away from forums because something newer showed up, not because we weighed them and found them wanting: “We tried the forum thing. We wanted something else. Not necessarily because it was better, though sure, maybe it was. But because it was different.” He is right about what replaced them, and he quotes psychologist Nigel Barber on it: “The problem is that they are run by an engagement algorithm that ignores the principles of successful communities.”
Forums still lost on one specific thing, and Smith walks straight past it. Forums had no front door.
You had to already know a forum existed to join it. Nothing in the format let a stranger wander in. Social platforms solved that brutally well, which is the real reason people went, and any revival that skips the front-door problem rebuilds the identical failure with better nostalgia bolted to it. Smith’s own favorite is Visual Editors, a mid-2000s news design forum he says “barely worked, honestly,” and every virtue he lists for it was invisible from outside.
So this argues for the ring more than against the room. A ring is a front door bolted onto rooms that never had one. Prev and next, every page, nobody to apply to, no algorithm deciding whether your badge renders today. Build the room if you want a room. Hang the door first.
Frequently asked questions
Does any of this actually replace the traffic search used to send?
No. Anybody selling it that way is lying to you. A webring moves tens of readers where an index moved thousands. The index is leaving whether or not you build anything, and what these tools hand back is a smaller number of people who came because somebody they trust pointed at you. That traffic converts at rates search traffic never approached. You are trading volume for intent, on a schedule you did not pick.
What is the minimum viable version if I only have one afternoon?
Publish the blogroll, export it as OPML, write one link post about something you genuinely read this week, and email the person you linked to. That afternoon buys a durable asset and one person who now knows your name, and the person is the half that compounds. Joining a ring takes longer, because other sites have to answer you. Start it once the blogroll gives you somebody to ask.
Does the money follow the same path as the attention?
Simon Owens’s July 24 issue advertises a Substack-versus-Patreon analysis in its own subhead and then does not deliver it, so I will not characterize it. The rest points where you would expect. Judd Legum’s newsletter has “over 150,000 signups and at least 7,500 paying subscribers,” and Owens cites growth marketing agency ATTN, via Bloomberg: “Microinfluencers average a 3.2% engagement rate, almost triple the 1.1% rate for macroinfluencers.” Smaller rooms, better numbers, in the attention column and the revenue column both.
Comments are open to members. Sign up free →