Consultation
Case · anonymised under NDAEN · AI

ChatGPT confused a developer with his namesake: the two identities separated in 12 weeks

AI models attributed someone else’s criminal case to the client. The namesake dominated Google, and there were no English-language sources about the client.

ClientDeveloper with partners in the EU
TriggerPartners’ questions in meetings
CauseIdentities mixed up in AI and search
Timeline12 weeks
Outcome0 of 4 models confuse them
Top 10 before
Top 10 after
IMG 0121:9
Case cover for “ChatGPT confuses the client with a namesake”: an atmospheric shot on the theme, with no faces or identifying details

The situation

IMG 024:3
The moment the problem became visible: a letter from the bank, a pause in negotiations, an AI answer

Foreign partners began asking in meetings about a “case” the client had nothing to do with. It turned out they had been asking ChatGPT about their prospective partner.

A namesake with the same patronymic had been involved in a criminal case and held the top positions in English-language Google.

Main service in this caseReputation in AI answersWhat ChatGPT, Alice, Perplexity and Gemini say about you

What the audit found

  • 3 of 4 AI models attributed someone else’s case to the client
  • The namesake took 6 of 10 results in English-language Google
  • Not a single English-language source about the client
  • No published photograph of the client, so models could not tell the two apart
IMG 033:2
Audit excerpt: search results for the name before the work, sensitive data blurred
IMG 043:2
Excerpt from the risk map in the client report

What we did

IMG 053:4
A working moment: documents, correspondence, a week-by-week plan
  1. Week 1

    Answer snapshot

    40 prompts across 4 models, plus a source map.

  2. Weeks 2–6

    English-language profile

    A biography with dates, photographs, the company and its projects.

  3. Weeks 3–9

    Media

    Coverage of the client’s projects in international trade media.

  4. Weeks 4–8

    Identity markup

    Structured data with distinctive identifiers.

  5. Weeks 6–12

    Tracking

    An answer snapshot every 2 weeks.

The result

0 of 4AI models now confuse them
6 of 10English-language Google results about the client
40prompts in each snapshot
12 wksto full separation
IMG 0616:9
Search results / AI answer / database record before the work
IMG 0716:9
The same view after the project
A model confuses people when it has nothing to hold on to. Distinctive identifiers — role, company, photograph, dates — work better than any denial.Estena’s position

What we did not do

  • We did not try to “remove” the namesake: he is not our client and that is not our job
  • We published no denials that would have linked the client’s name to the case

Tell us about your situation. We reply within one business day.

The first meeting is free and carries no obligation. We sign an NDA before you share any details.

TELEGRAM  @estena
E-MAIL    hello@estena.agency
OFFICES   Moscow · Tbilisi · Dubai

By clicking the button, you agree to our privacy policy.