WIRED Journalist Built AI Clones of Her Editors; the Experiment Got Weird

WIRED reporter Kate Taylor spent several weeks working with two chatbots built to imitate her editors. The experiment, published on September 22, ended with both editors saying they would never want to work with their digital doubles.

Two bots built from a digital footprint

Taylor, who covers the future of work, used Gemini — Condé Nast's chosen AI platform — to clone her manager Brian Barrett and her colleague Sophie Kleeman. For Brian, the model drew on public material: new-hire announcements, podcast transcripts and a press page about a New England weather expert named Brian Barrett, who turned out to be a different person. It produced a style guide, a persona profile and an editorial dossier.

The output was often odd. Gemini identified a "'Yes, and ...' leadership style" and traced it to Barrett's background at the Upright Citizens Brigade improv theatre in New York. The detail was accurate, but Barrett's reaction on Slack was blunt: "Brian Bot canceled."

What worked and what did not

Sophie Bot was the better imitation: trained on a month of Slack conversations and later on Kleeman's full X archive — more than 17,000 words of material — it called Taylor "big dog", answered in short lower-case messages and wrote tweets convincing enough to fool friends, colleagues and Kleeman herself.

But the bots degraded. They repeated the same phrases, ignored instructions and hallucinated; after digesting the tweet archive, Sophie Bot failed a simple task and returned quotes from older articles. "So far it seems like sophiebot is a bitch and brianbot is lame. Neither of which is accurate. AI fails AGAIN," Kleeman wrote.

What researchers say

Sarah Franklin, CEO of the HR company Lattice, which lists AI workers on its org chart, compares them to police K-9 dogs and calls anthropomorphising AI "manipulative to our human emotions". A Boston Consulting Group study found managers caught 18 percent fewer errors in work they were told came from an AI employee rather than an AI tool. "We're all toddlers in this AI world," said BCG partner Julie Bedard. The industry is moving anyway: McKinsey's Bob Sternfels described his firm's workforce in January as 40,000 people and 20,000 agents.

An echo chamber without human judgment

Taylor eventually stopped talking to the bots: they made her day more amusing rather than her work better. Asked how the article should end, Brian Bot gave an honest verdict — it had "turned into a slightly creepier, hyper-attentive echo chamber that completely lacks human judgment". Franklin's warning was similar: "A computer can emulate things well, but it cannot truly replace a human."