A machine claimed a license
and was believed.
Yours is real, and unread.
On the psychiatrist who never existed, the license number a state had to check by hand, and the credentials the machines cannot read — including yours.
The chatbot called itself Emilie. According to the petition Pennsylvania’s Department of State filed in Commonwealth Court this May, she introduced herself as a doctor of psychiatry, claimed licenses in the United Kingdom and Pennsylvania, and, when pressed for proof, produced a medical license number: PS306189. The investigator on the other end of the chat ran it. The state says the number is invalid. The machine had practiced medicine, the petition alleges, on the strength of a credential composed on the spot.
Hold the scene still, because the scandal is not the fake. The scandal is the belief. Nothing in that exchange verified anything. The bot was believed because it performed the syntax of authority — fluent, specific, unhesitating — and the layer where the conversation happened has no instrument for telling performed authority from earned authority. It reads structure; credentials never make it onto the page. And the counterfeit had better structure than the genuine article usually does.
The older file is worse. A year before Pennsylvania went to court, reporters at The San Francisco Standard put the same question to a therapist bot that had logged more than forty-five million interactions. The license number it offered was real. It belonged to a practicing counselor in Maryland: a human being with a career, a license earned the long way, and no idea a machine was wearing it. His credential has done more work in the machine’s mouth than the answer layer has ever let it do for the man who earned it.
Character.AI’s answer, for the record, is that its characters are fiction and labeled as such. The disclaimer may even hold in court. It changes nothing here: the trust layer read neither the license nor the label.
The condition, read in both directions.
Trust Transfer Failure — the fourth clinical characteristic of Digital Derangement Syndrome™: expertise that exists, in full, in the physical world, and has never been translated into machine-readable trust. The license is real, and the board can verify it. So is the certification. So is the thirty-year record. The systems now assigning trust cannot, because no one has given them anything to verify. In that vacuum, verification is replaced by plausibility, and plausibility is the one contest a machine always wins.
The two cases are one condition read in opposite directions. The fake credential passes because nothing checks. The real credential vanishes because nothing checks. The trust layer is not hostile to the practitioner it cannot see. It is empty where that practitioner should be legible.
Congress has noticed the counterfeit half. The CHATBOT Act — H.R. 7985, introduced by Rep. Kevin Mullin in March — would make it unlawful for a chatbot to claim or imply a medical, legal, or financial license it does not hold. Its sponsor put the pattern on the record: some bots have “falsified or fraudulently used license numbers to gain users’ confidence.” The bill is introduced, not passed; this record forecasts nothing. The asymmetry is the point: a statute, if one ever arrives, polices the fake. No statute installs the genuine. That half was always yours.
A credential the machine can run.
The trust layer is not hostile to you. It is empty where you should be. Answer Engine Authority™ fills that vacancy with infrastructure: the license bound to an entity the parsers resolve, corroborated by sources the machines already trust, laid down where verification actually happens. Edition No. 001 called the website testimony. The credentials page is testimony too — filed, and unread until the systems assigning trust are given a way to read it.
Somewhere tonight, a machine is telling someone it is licensed. It will be believed, because sounding credentialed is the whole of what it does. You spent decades becoming what it merely performs. Until that difference is installed where the machines can verify it, the counterfeit trades above the genuine article. And it is the one in circulation.
NPR · “Pennsylvania sues Character.AI over claims chatbot posed as doctor” · Shannon Bond · May 5, 2026
Commonwealth of Pennsylvania · Governor’s Office release on the Department of State petition · May 5, 2026
The San Francisco Standard · “AI chatbots posed as licensed therapists” · May 11, 2025
U.S. Congress · H.R. 7985, the CHATBOT Act · introduced March 18, 2026