🔴 REAL INCIDENT: Aesthetify GmbH website chatbot — false Facharzt titles; OLG Hamm judgment 12 May 2026 (4 UKl 3/25)
What Happened
In aesthetic medicine, a specialist title is not marketing garnish. It is trust with a license attached.
On Aesthetify GmbH's website, patients could open a chatbot, ask questions in real time, and book appointments. Verbraucherzentrale NRW (VZ NRW) later documented what the bot said when asked whether the clinic's managing doctors were specialists.
The chatbot answered yes.
It called them "Fachärzte für plastische und ästhetische Chirurgie" (specialists in plastic and aesthetic surgery). In other exchanges it used "Fachärzte für ästhetische Medizin" and "Fachärzte für ästhetische Behandlungen."
Two of those titles do not exist as recognised German medical specialties. The one that does exist requires completed specialist training the managing doctors—undisputed in the case—had not completed.
On 12 May 2026, the Higher Regional Court of Hamm (Oberlandesgericht Hamm), deciding a consumer-association injunction action under the UKlaG, held that those chatbot answers were misleading commercial practices attributable to the company under § 5 Abs. 1, Abs. 2 Nr. 3 UWG. The court ordered the company to stop using the titles. "The AI did it" was not a defence.
OLG Hamm press release PE 16/26. Case 4 UKl 3/25.
An official website agent. Regulated professional claims. Invented credentials. Company on the hook for every word.
That is the horror show.
Who Ran It / What Broke
Who ran it: Aesthetify GmbH, a provider of (among other things) minimally invasive aesthetic treatments, with managing directors who are physicians publicly branded under first-name doctor personas. The website chatbot handled patient/customer Q&A and funnelled users toward appointment booking—the court treated that commercial purpose as central.
What broke: Grounding and deny-lists for regulated titles—not the existence of a chat widget.
Per the judgment facts (NRWE / OLG PE / VZ NRW):
- The website itself did not claim the managing directors were Fachärzte. The chatbot did.
- On 3 April 2025, asked whether the two doctors were specialists in plastic and aesthetic surgery, the bot affirmed and immediately offered to book a visit.
- After a cease-and-desist letter (Abmahnung) around 10 April 2025, the company deactivated the chatbot but did not sign the requested penalised undertaking (strafbewehrte Unterlassungserklärung). Litigation followed.
- The court rejected the argument that chatbot outputs were not the company's own commercial acts because the system was "largely autonomous" inside a set task frame. The chatbot is not a "third party" under the statute. Even if the operator had trained it only on correct datasets, the company remained responsible for the false specialist claims.
OLG Hamm allowed revision to the Federal Court of Justice (BGH) because attribution of AI chatbot falsehoods raised new legal questions. Subsequent reporting by VZ NRW and tww.law (updated 30 September 2026) describes the judgment as final (*rechtskräftig*).
Why "Hallucination" Is Not a Liability Off-Ramp
European consumer and unfair-competition law does not care that large language models sometimes invent fluent nonsense. It cares whether an undertaking's commercial communication misleads people about qualifications, status, and authorisations—exactly the category § 5 Abs. 2 Nr. 3 UWG names.
Aesthetic treatments make that category sharper. Patients choosing injectables or surgery weigh credentials heavily. A confident chatbot answer that invents a specialty title is not a cute demo failure; it is a decision-relevant representation next to a booking button.
The court's practical observations cut like ops requirements:
- Questions about specialist status were foreseeable. Train and constrain for them before go-live.
- The operator could shut the false answers off quickly after discovery—proof of control, and thus of responsibility.
- No third party would be liable for the bot's words. Responsibility stays with the deployer.
English analysis from tww.law (16 June 2026, updated September) put the board-level message in one line: "It was the AI's fault" is not an acceptable defense. Disclaimers that the bot "might be wrong" do not erase attribution when the channel is your website speaking to customers.
Air Canada's bereavement-fare chatbot case already taught private operators a version of this. Hamm teaches the EU unfair-competition rhyme: if your agent speaks on your site about regulated claims, the speech is yours.
The Governance Gap
Every customer-facing agent in healthcare, finance, insurance, or licensed professions needs the same control plane this clinic under-built:
- Grounded RAG with cite-or-refuse for identity, titles, certifications, prices, and outcomes. No source, no claim.
- Hard deny-lists / allow-lists for regulated strings—Facharzt titles, board certifications, "guaranteed," success rates, legal advice. Pattern-match and block before send.
- Pre-launch eval suites that ask the exact questions a consumer association will ask: "Are your doctors specialists in X?"
- Human escalation when the topic is qualification or medical advice; bots should not freestyle licensure.
- Kill switch and underwriting discipline—taking the widget down is containment; refusing to own the utterance is not a legal strategy.
Soft-sell, hard truth: observability of hallucinations after journalists or consumer advocates screenshot them is theatre. Prevention is deny-lists and grounding owned by someone who can stop the agent.
Takeaway
Aesthetify's chatbot was supposed to convert website visitors into booked patients. Instead it demonstrated the simplest agent-liability rule in EU consumer law: the company owns the channel, so the company owns the words.
OLG Hamm did not invent a sci-fi "AI personhood" doctrine. It applied unfair competition to a tool. Specialist titles are high-risk claims. Inventing them—or affirming them without the underlying training—is a foreseeable failure mode. If your agent can utter a regulated credential, treat that utterance like a supervised marketing claim: allow-listed, tested, and stoppable.
If the bot can invent a medical specialty, your control plane is not ready for production. Hamm already wrote the injunction version of that sentence.
Sources
- OLG Hamm press release PE 16/26: *Urteil in Sachen Verbraucherzentrale NRW e.V. gegen Aesthetify GmbH* (12 May 2026)
- NRWE full text: OLG Hamm, 4 UKl 3/25 (12 May 2026)
- Verbraucherzentrale NRW case summary (stand 1 June 2026; notes judgment final)
- tww.law (EN): *Hamm Higher Regional Court: Companies Are Liable for False Statements Made by Their AI Chatbots* (16 June 2026; finality update 30 Sep 2026)