futurism.com web signal

Movchan Agency ran fake AI publicists to pitch reporters

TL;DR

  • UK-based Movchan Agency operated at least 15 fake personas with AI-generated headshots to pitch journalists on behalf of paying clients.
  • Emails from personas 'June Barton' and 'Summer Casey' scored 100 percent AI-generated on the detection tool Pangram, per Futurism's testing.
  • Founder Nadya Movchan called it 'poor judgment' tied to Gmail deliverability, though fake-persona emails date back to September 2025.

A PR agency getting caught juggling fake pitch personas would be a regular industry embarrassment. Doing it with AI-generated headshots pulled from a public face-generator, at the scale Futurism reports, is something different: it is the first clean, named example of an agency using generative tools to fabricate the humans on the other end of a cold pitch.

Reporter Maggie Harrison Dupré traced at least fifteen distinct fake personas back to the UK-based Movchan Agency, including a 'June Barton' pitching a wellness app called RiseGuide and a 'Summer Casey' working from a fake firm domain. The giveaway was almost cartoonish. Several headshots still carried the watermark from this-person-does-not-exist.com, and the fake PR-firm domains all redirected to Movchan's actual homepage. When Dupré ran the pitches through Pangram, an AI-text detection service, the emails from Casey and Barton came back as 100 percent AI-generated, while others showed a human-AI mix.

The reason this matters beyond a single agency is that PR outreach still runs on a trust convention. A reporter assumes the person on the other end of the pitch is a real publicist with a real career they would not risk. That convention is what makes cold email answerable at all. Movchan's setup, whether or not you find founder Nadya Movchan's 'poor judgment' framing convincing, shows how cheap it now is to synthesize the human end of that transaction at scale. The clients Futurism contacted, including RiseGuide and Joi AI, said they had no idea and denounced the practice, which is credible and also beside the point for any journalist now looking at a pitch from a name they cannot verify.

The honest caveat is that this is one agency, exposed by one reporter, and Movchan's own account, that the personas were 'pre-warmed domains' set up 'a few months ago' to work around a Gmail deliverability issue, is not fully corroborated. Emails from the fake accounts reportedly date back to September 2025, which is a longer runway than 'a few months.' What the reporting does not tell you is how many stories actually landed off these pitches, or whether any client knowingly signed off.

The useful thing that comes out of this is a cleaner incentive picture. AI-detection vendors just got a real newsroom use case. Agencies that can document human-only outreach have something to sell. And reporters have permission to start reverse-image-searching a headshot before they open the pitch.

Shared on Bluesky by 2 AI experts