On June 3, 2026, Matthew Prince, the CEO of Cloudflare, posted a number from his own network’s live data. For the first time in the history of the internet, more than half of all web traffic was not human. Bots, crawlers, and AI agents made up about 57 percent of the requests. People made up the rest. Prince said he’d figured the crossover would come at the end of 2027. Agentic traffic got there a year and a half early.
Sit with that. The web you’ve spent twenty years learning to win is now, by volume, mostly machines. And the machines aren’t browsing. They’re fetching. An agent doesn’t scroll your listings and admire your photography. It queries your inventory the way you’d query a database, takes what it can read, and moves on.
Which leaves one plain question to decide the next few years. When the machine comes looking, can it read you?
The deal changed
Prince has been putting out a second number for a while, and it’s the one that should bother every operator. Ten years ago, Google crawled about two of your pages for every visitor it sent back. Fair trade: you let them index you, they sent you traffic. The AI crawlers don’t trade like that. The pure-LLM bots pull thousands of pages for every click they send back. They read everything you publish and send you almost no one.
That’s the part people wave off when they call AI “only one percent of traffic.” It isn’t a smaller version of search. It’s a different arrangement. Search read you to send you people. The agent reads you to answer someone else, somewhere else, on a question you never even saw. You can be the best fit for what a buyer needs and never surface, because the machine couldn’t make sense of what you had.
The barbell
I’ve written before that this industry isn’t heading toward a clean roll-up. It’s heading toward a barbell.
One end is consolidation, the ordinary kind. The auction business is hitting the same moment veterinary practices and dental groups hit a decade back: aging owners with no successor, fragmented ownership, capital ready to buy on cash flow. Ritchie Bros bought BigIron. The private equity calls are landing. That end of the bar keeps getting heavier.
The other end is new. The tooling that used to take real money and a couple of years to build, a bidder portal, a catalog system, a marketing engine, can now be stood up by one sharp operator with AI for a fraction of last year’s cost. The small end of the bar gets stronger, not weaker.
What gets crushed is the middle. The mid-sized firm with neither the scale of the big house nor the software self-sufficiency of the nimble solo. Everybody’s been telling that as a money story, an M&A story. It’s bigger than that. The thing that actually springs the trap on the middle is whether anyone, or anything, can find them.
What the last twelve months did
Here’s what moved, and it moved fast.
Zero-click searches, the ones that end without anyone clicking through to a website, hit 68 percent of Google searches this year, up from around 60 two years ago. Where Google now shows an AI Overview, organic click-through falls by nearly 60 percent. The front door you optimized for two decades is quietly swinging shut.
The new door is wide open and busy. AI-referred traffic tripled year over year. Shopify, which watches a huge slice of online commerce, reported that orders coming in from AI grew roughly thirteenfold in a single year, and that those AI-referred shoppers convert about 50 percent better than the ones who arrive from a normal search. Adobe measured the same pattern across a trillion retail visits and found the thing I can’t stop turning over. A year ago, traffic that arrived by AI converted worse than human traffic. This year it converts 42 percent better.
Look at that flip again. The buyers the machines bring aren’t the bottom of the funnel. They’re the top. They show up already knowing what they want, because the agent did the sorting before they ever appeared. So if your data isn’t readable by the machine, you don’t lose a little traffic at the edges. You lose your best buyers first, to whoever the agent could actually understand.
That’s what “underwater” really looks like. The middle doesn’t bleed out slowly. It gets skimmed of its best demand, then repriced.
The call
So here’s the call, with a date on it, so you can hold me to it.
Inside the next 24 to 36 months, a meaningful share of auction companies will find themselves underwater, and the cause won’t be a soft economy or a deal they didn’t chase. It’ll be that they went non-findable. The agents doing the discovering won’t be able to parse what they have, so they’ll route the buyers elsewhere. The firm will still have good inventory, good people, a name people trust in the room. It just won’t exist to the machine doing the looking.
You can check me on it. If three years from now AI-mediated discovery is a rounding error and the firms that ignored it are doing fine, I was wrong. I don’t think I’m wrong.
Three chairs
I’ll tell you why I’m willing to say it that flat. I’m sitting in three chairs on this one, and they all show the same picture.
I run an auction company close enough to the wave to feel the pull of it. I chair the board of the national association, which means I watch the whole field and not just my own corner. And I’ve been writing this argument down for months, long enough that I now hear it come back from peers who read an early version and started using the language themselves. When the operator’s seat, the industry’s seat, and the writer’s notebook all point the same direction, I quit calling it a hunch.
Plenty of this lands harder on the middle than on the big houses, and I know who’s reading. I’m not naming names or calling a top on anybody’s business. I’m telling you what the data says is coming while there’s still room to do something about it. That’s about the only useful thing a person in these chairs can do with a view like this.
Put it down where the goats can get it
So what do you do.
You make yourself readable. Not pretty. Readable. Our world has spent years building sites that look sharp to a person and say nothing to a machine: ten broad categories, three-word lot titles, photos with no real description underneath. A human can squint and work it out. An agent can’t, and the agent is the one shopping now. Structured data, clean categories, descriptions specific enough that a machine scanning for “commercial kitchen equipment under five thousand dollars within two hundred miles” turns up yours. That stopped being a marketing nicety. It’s whether you’re in the market at all.
And you sort what you own from what you rent. The bidder list you built over twenty years, you own. The brand, the contractor network, the reputation, you own. The platform you lease, the ad traffic you buy, the borrowed reach, you rent, and rented things get taken back when the terms change. The operators who come through this in good shape will be the ones who owned their data and made it legible to the machines on purpose, before they had to.
There’s an old line around farms and sale barns. If you want everything fed, put it down where the goats can get it. Don’t hang it up high where only the tall ones reach. Set it at ground level, plain, where anything that’s hungry can find it.
For a hundred years that meant talk straight and price it clear, so the buyer in the back row understood you as well as the one up front. It still means that. But the hungry thing in the room now isn’t only the buyer. It’s the machine doing the buyer’s looking. And it’s standing right down at ground level, reading, waiting to see whose feed it can actually reach.
Put it down where the goats can get it.