A broken tool usually tells you it’s broken.
Last month, a man was crowned king in a stadium in England. There were robes. A throne. A full ceremony. 50,000 people in attendance.
None of it happened.
Three weeks later, if you asked the largest AI systems in the world who held that title, they gave you his name. And they told you about the ceremony. Not as a rumor. Not as a claim somebody was making. As background — the way you’d mention where a person went to college.
Hi, I’m Jeff Payne. You’re listening to The Jeff Payne Show, Episode 83: First a Claim, Then a Fact.
Here’s the context, and it takes about thirty seconds.
There’s a corner of the marketing world fixated on a question that gets bigger every month: when somebody asks an AI system something, whose name comes back in the answer? Not whose website ranks. Whose name gets said.
A few people in that world decided to run an experiment. They picked a title that didn’t exist — a made-up crown for a made-up category — and set out to see if they could get the machines to hand it over.
So they published. A press release. Then another. Then videos, articles, a few websites built for the occasion, and a reference page written in the calm, neutral voice of an encyclopedia.
Within hours, the answer engines were repeating it.
That’s the part my industry is talking about this week. It’s the least interesting part.
Here’s what actually happened.
The first release made a claim. It said, “This man is the king of this thing.” A claim is something a system can weigh. It can look for support, look for contradiction, decide how much confidence to attach to it.
The second release didn’t make the claim. It announced something else entirely — and on the way there, it described him, in passing, as the man known across the industry as the king.
The title wasn’t the point of the sentence. It was the setup for the sentence.
That’s the whole trick.
Claims get examined. Premises don’t. The moment something stops being asserted and starts being assumed, nothing in the chain goes back and reopens it. It’s furniture now.
And here’s the detail I’d sit with. That write-up said fifty thousand people attended. The stadium holds twelve thousand. It’s the all-time attendance record, going back to the day it opened: 12,005.
Four times the capacity. Four times the record. You could check that on your phone in ten seconds.
Nothing checked it. It got repeated instead.
Because these systems aren’t built to verify. They’re built to notice agreement. When enough independent-looking sources say the same thing, that consensus is the signal. And consensus is a perfectly reasonable stand-in for truth — right up until somebody decides to manufacture the consensus on purpose.
The whole thing cost less than dinner for two and took an afternoon.
Now hold that against something that happened slowly.
Burbank Dental Lab has been a client of mine for years. A while back, they built a set of guides for dentists — practical, genuinely useful reference material on restoration work. Not marketing. Reference. Over time, those guides generated more than $2.4 million in revenue.
Then they rebuilt the whole thing as an interactive 3D implant education platform.
Here’s the part that matters today. A company in their industry — not a partner, not somebody they paid — linked to it. Decided on their own that it was the resource worth pointing at.
Nobody published a press release announcing that Burbank Dental Lab was the authority in implant education. Somebody else concluded it, and said so in public, without being asked.
On any given Tuesday, those two things look identical to a retrieval system. A manufactured crown and an earned link are both just third parties saying something about you.
They are not identical over time. One of them has something underneath it.
So what’s the executive takeaway? I want to be careful here, because I want to be honest with you about what this story doesn’t prove.
That crown was won in an empty room. Nobody searched for that title. No customer ever typed it. No business lost a dollar failing to rank for it. There was no competition because there was nothing there worth competing over.
Nothing your company wants to be known for is an empty room. The phrase your buyers actually use has incumbents, budgets, and people who’ve been at it for a decade. That cost profile does not transfer, and anyone telling you it does is selling you something.
But the mechanism transfers completely.
The mechanism is that the record about your company can be written by somebody who isn’t you — cheaply — and then quietly promoted from claim to premise. After which it stops being argued about and starts being assumed.
That cuts both directions, and the unflattering direction is the one nobody’s planning for. There’s no correction desk. There’s no form. You can’t file a dispute with an answer engine.
So here’s a question.
Ask the answer engines about your own company. Not “do they mention us” — you’ve probably done that one. Ask something different: what do they state about you in passing? What gets treated as settled background, said without hedging, on the way to saying something else?
That’s the part nothing is going back to check. Worth finding out what it says while it’s still yours to influence.
Thanks for listening. I’m Jeff Payne. I’ll see you next time.
Last month a man was crowned king in a stadium in England. Robes, a throne, a full ceremony, 50,000 people in attendance. None of it happened — and three weeks later the largest AI systems in the world were describing the ceremony as settled background.
The stunt is the boring part
A small group of marketers ran an experiment. They invented a title — a crown for a category that didn’t have one — and set out to see whether they could get answer engines to hand it over. They published a press release. Then another. Then videos, articles, a handful of purpose-built websites, and a reference page written in the flat, neutral register of an encyclopedia.
Within hours, the answer engines were repeating it. That is the part the marketing industry is discussing this week, and it is the least interesting thing about what happened.
First it’s a claim. Then it’s a fact — and nothing between those two points is doing any checking.
Claims get examined. Premises don’t.
The first release made a claim: this person holds this title. A claim is something a system can weigh. It can look for corroboration, look for contradiction, and decide how much confidence to attach.
The second release made no such claim. It announced something else entirely—and on the way there described the same person, in passing, as the man known across the industry as the king. The title was not the assertion of the sentence. It was the setup for the sentence.
That is the entire mechanism. Once something stops being asserted and starts being assumed, nothing in the chain can go back and reopen it. It has been promoted from claim to furniture.
This is one of the best written, most informative email newsletters I’ve ever read!
Thank you for sharing your expertise, and hope you’re doing well.
Beth Zimmerman
Founder, Executive Director
The number nobody checked
The original write-up reported a crowd of 50,000 people. The stadium mentioned in that write-up holds an all-time attendance record of 12,000. Going back to the day it opened, it was 12,005.
Four times capacity.
Four times the record. A fact any person could verify from their phone in under ten seconds. Nothing in the chain that repeated it did so.
That is not a bug in one system. Retrieval systems are not built to verify — they are built to notice agreement. When enough independent-looking sources say the same thing, the consensus is the signal. Consensus is a reasonable proxy for truth right up until someone deliberately manufactures it.
What the other version looks like
Set that against something that happened slowly. Burbank Dental Lab, a long-standing client and one of the top 5 dental labs in America, built a set of practical reference guides for dentists on restoration work — not marketing material, reference material. Over time, those success guides generated more than $2.4 million in additional revenue. They are now being built as an interactive 3D implant education platform.
The part that matters here: a company in their industry — not a partner, not a paid placement — linked to it, having decided independently that it was the resource worth pointing at. No press release announced that Burbank Dental Lab was the authority in implant education. Someone else concluded it, in public, unprompted.
To a retrieval system on any given day, those two things look the same. A manufactured crown and an earned citation are both third parties saying something about you. Over time they are not the same at all.
A manufactured crown and an earned link look identical on any given Tuesday. One of them has something underneath it.
Why this doesn’t transfer the way it’s being sold
One limit is worth stating plainly, because it is what the experiment did not prove. That crown was won in an empty room. Nobody searched the phrase. No customer ever typed it. No business lost revenue by failing to rank for it. There was no competition because there was nothing worth competing over.
Nothing a real company wants to be known for is an empty room. The phrases buyers actually use are used by incumbents, budgets, and operators who have been working them for a decade. The cost profile does not carry over, and anyone claiming otherwise is selling something.
The mechanism, however, carries over completely: the record about a business can be authored by someone who isn’t that business, cheaply, and then quietly promoted from claim to premise — at which point it stops being argued about and starts being assumed.
There’s no correction desk. There’s no form. You can’t file a dispute with an answer engine.
The question worth asking
Most businesses have already run the obvious test: Do the answer engines mention us? That one is easy and largely answered.
The harder question is what they state about you in passing — what gets treated as settled background, said without hedging, on the way to saying something else. That is the part nothing is going back to check, and it is worth knowing what it says while it is still yours to influence.
Source: The general thesis was adapted from the Edward Sturm newsletter. Stadium capacity and record attendance independently verified against Leigh Sports Village published figures.
COMPLETE THE FORM TO
BOOK A STRATEGY CALL
"*" indicates required fields
COMPLETE THE FORM TO
BOOK A STRATEGY CALL
"*" indicates required fields
Subscribe and Share – WE APPRECIATE YOUR SUPPORT
