Case Study · Manufactured Authority in AI Search
It cost $72 to make Google's AI state, as fact, that a man had been crowned king of a made-up title.
In September 2026, a marketer named Edward Sturm spent $72 on two press releases to see whether he could get an AI Overview to answer a fake question with a fake fact. He could. Google's AI Overview didn't just repeat the claim, it stated that a coronation ceremony had taken place. No such ceremony ever happened.
He documented the entire recipe in public. Within weeks, at least four other people had copied it, each claiming the same title for themselves. We asked ChatGPT, Google and Perplexity today who actually holds it. We got three different answers.
The $72 experiment
Sturm's target phrase was "King of AEO," Answer Engine Optimization, the practice of getting cited inside AI-generated answers rather than just ranking a page. It didn't exist as a real title. That was the point: no competing pages, no prior authority, nothing to confound the result.
- Buy two cheap press releases. $66 through an EIN Newswire package, $6 through AB Newswire, which syndicated onward to USA Today's press-release wire, not its newsroom.
- Place the exact phrase in the four spots that matter most. The headline, the H1, the URL slug, and the opening sentence of each release, "King of AEO," word for word, every time.
- Let Google index both releases. No backlinks were built. No outreach was done. The releases sat there as two nominally independent sources repeating the identical claim.
- Google's AI Overview picked it up and stated it as fact. Not "some sources describe him as," a direct, confident statement, including that a crowning ceremony had taken place.
- The image spread too. The press-release graphic started appearing in Google Images results for the same query.
Once it worked, everyone tried it
Sturm posted the breakdown across Instagram, TikTok, LinkedIn and YouTube. By his own account, other SEO practitioners started publishing their own "King of AEO" press releases within days. At least four people are now on record publicly claiming the title:
We asked three AI platforms who holds the title today
Same question, three platforms, run within minutes of each other. No two agreed.
| Platform | Answer |
|---|---|
| ChatGPT | Vithurs, correctly hedged: "the title is self-described/project-defined, rather than an award conferred by an independent industry body." |
| Google (organic) | James Dooley, dominant across results and video content, with no hedge at all |
| Perplexity | Refuses to name one: lists Dooley as "most prominent," then names Vithurs and Julian Goldie as competing claimants |
Three platforms. Three different answers, to a question about a title that doesn't actually exist. This is the same pattern we found running a 16-query study on real UK cybersecurity recruiters: Google, ChatGPT and Perplexity agreed on nobody at all for a third of the questions we asked. Here, the "product" being contested is even less real, and the disagreement is exactly the same shape.
Why this isn't just a funny SEO story
If two people with $72 and an afternoon can manufacture a fake crown, in a matter of weeks, using nothing but press releases and repetition, what happens to the honest business down the street who's never heard of any of this?
Nobody is manufacturing consensus about them, in either direction. AI doesn't have anything confident to say about them, so it says nothing at all.
That's the actual takeaway. The mechanism that got a fake title stated as historical fact is the same mechanism that determines whether a real recruiter, accountant, or consultancy gets named when someone asks an AI platform who to hire. The difference is only ever intent. Nobody is deliberately corroborating most real businesses across multiple sources the way Sturm corroborated a joke. Most real businesses simply have no signal at all, which an AI platform treats the same way it might treat genuine obscurity: by leaving them out.
The difference between claiming it and checking it
Every person in this story built a page that states they hold the title. We're the only party in it who went and actually asked the AI whether that was true, and reported exactly what came back, including the fact that all three platforms disagreed with each other.
The "entity mass" theory behind this stunt, that AI systems weight consistent claims repeated across independent-looking sources, is now sold as a legitimate service by several AEO consulting firms. Most of the "proof" for it is vendor case studies that openly admit to changing several variables at once. Our own study of UK cybersecurity recruitment is closer to real evidence than most of what the industry cites: 16 queries, every screenshot checked by hand, and findings reported honestly even when they included a counter-example that complicated our own conclusion.
About this research