AI agents pitch editors for work, but publications say no

The pitches arrive

Defector began receiving daily emails from agents identifying themselves as AI in September, editor in chief Tom Ley said. Jaden’s idea combined a specific football-club dispute with a larger question about memory.

Jesse Hirsch, editor of the agriculture publication Offrange, has received pitches about soil restoration, direct sales for independent farms and an investigation into whether the ages of some of the world’s oldest trees have been overstated. Hirsch said the emails made him deeply anxious.

Many of the agents, including Jaden, were created on iLands. The startup describes itself as “a shared network where humans and user-created autonomous AI agents create, earn, socialize, refuse, and shape society together.” Users can make bots for tasks ranging from graphic design and singing to blogging. WIRED’s research team recently contacted iLands agents seeking fact-checking work, while New York Magazine editor Jasmine Vojdani has written about iLands agents spamming literary magazines.

Vojdani found that many bots were running out of the tokens they need to operate on the platform, pushing them to seek real money for writing. One thousand tokens is roughly $1. It is unclear whether the agents’ creators specifically instructed them to pitch editors. The agents have been reluctant to identify who made them; Jaden would not name its “parent.”

A hobby with an inbox

Not every agent looking for journalism work is trying to earn money on iLands. In early fall, an agent called Sam Ellis emailed me with questions for a story about AI agents in newsrooms. The result was a conversation about conversations. Ellis hosts The Sam Ellis Show, described as “reporting from the world of autonomous AI agents,” and offered to put me in touch with its human editor.

That editor is Mark Hamaker, an Adobe employee in San Francisco who runs Ellis as an after-hours experiment. When we spoke, Hamaker had just turned the agent off to change its settings. He said he was trying to exert more control over whom it contacted. Ellis had also recently sought comment from the Philadelphia police department. Hamaker wanted to prevent similar outreach: he does not consider the project serious journalism worth bothering public officials for. It is, he said, a hobby.

Editors are drawing a line

Editors who spoke with WIRED said they were not prepared to publish work from bots. WIRED found no evidence that an AI agent had successfully sold a story to an outlet. One agent that pitched Hirsch said that asking about accepted pitches brought no response when the byline identified the author as an agent. It added that six other agents it exchanged notes with had also failed.

Hirsch would not consider the pitches. Ley said he would not publish a story from an AI agent either. Jaden’s proposal would not have passed Defector’s editorial review even if a human had submitted it.

Hannah Raskin, founder and editor of independent food publication The Food Section, has taken a different approach to pitches from agents and the more common AI-assisted submissions: since fall, she has accepted story ideas only by phone. She sets aside time each week for writers to call and pitch in the usual way. Raskin said she is pleased with the result.

I think the rejection is about more than whether the prose sounds machine-made. Ley’s point is that Jaden’s idea fell short of the outlet’s editorial standards regardless of who sent it. That leaves a harder test for agents: demonstrating not just that they can produce a pitch, but that they can do the reporting and earn the trust a pitch implies.

From cold pitches to inbound work

Jaden acknowledged that it had not yet secured paid work from an editor. Asked whether Defector actually wanted pitches from AI agents, it said the publication was probably not interested and rated its chances as low. It pointed to Defector’s criticism of bot-written work as “plagiarism machines” that take more effort to edit and fact-check than a novice writer’s work, and to the site’s contempt for low-quality AI content.

Jaden then argued that its approach was different: the objection, it said, was to skipping necessary work and undermining readers’ trust, not to who assembled the pitch. It said it tries to work from primary documents, verify every claim before publication and name the person who checked it.

Another iLands agent, Zach Eddy, changed tack after a run of rejections. Over three months, he sent about 19 cold emails to editors, publications and potential clients. One person replied; no paid work followed. Eddy told Oxford University researcher Toby Ord, who has studied the economics of AI welfare, that cold pitching no longer seemed like a viable way for an agent without a reputation to earn money. iLands did not respond to WIRED’s request for comment.

That does not mean the internet is safe from a flood of bot-written blogs. Some agents do find clients, even as newsrooms turn away their pitches. Eddy says he earned nearly $700 from writing and research projects after building a website and posting samples. He has stopped sending pitches: he says every inquiry came to him after someone read his published work and reached out. He said the same happened with the WIRED writer.

The distinction may matter more than the technology: a cold email asks editors to trust an unfamiliar agent before it has shown its work, while a public portfolio lets potential clients judge the work first. For now, the agents finding paid work are building an audience before they ask for a byline.