Antiwar.com prides itself on being an online home to the Antiwar movement. Yes, many of our staff are libertarians, but we run articles from authors across the political spectrum, as we have since our founding in 1995.
We take submissions from anyone who meets our standards, such as word count, grammatically clean, and well written, and within our editorial position. However, we have recently encountered a new issue with submissions: AI-generated articles.
Starting in the spring of 2026, Antiwar.com began receiving submissions that we later linked to an operation involving AI-generated articles submitted under deceptive identities. Some were published before we identified the problem and were subsequently removed.
After initially being alerted to articles from two authors, Antiwar.com began running articles through AI-detection programs. But we found those programs to overestimate the use of AI, particularly for articles that were grammatically clean, logical, and well-written.
So we could not simply use those programs to check articles and simply reject anything that came back as AI-authorship; rather, we began contacting authors whose submissions came back as a high likelihood of AI-authorship.
Some of these authors refused to speak on the phone, which raised further concern that there may be something more nefarious afoot.
In July, we received a submission from one author from a different author’s email. We began to suspect that Antiwar.com was the target of a person or AI agent that was generating articles by multiple authors and sending submissions to Antiwar.com under multiple bylines.
Antiwar.com then used ChatGPT to run a scan of all articles published as viewpoints at Antiwar.com over the past year, checking for AI-authorship and confirming the identity of the authors. ChatGPT identified two clusters of authors whose articles and submission histories raised concerns about AI authorship and possible coordination.
The Antiwar.com team then uploaded the original .docx submissions into ChatGPT to analyze the metadata in the articles. ChatGPT identified similarities in writing styles and submission patterns among the suspicious authors.
Cluster A was the two authors who published at Antiwar.com and were previously identified earlier this year. Cluster B was made up of at least three authors, and possibly a fourth. The investigation did not uncover any evidence linking the two clusters.
An analysis of the original Word documents also revealed striking similarities in the metadata of articles attributed to three authors in Cluster B.
The ChatGPT review also uncovered that the authors appearing in the two clusters were published at a half dozen other outlets. Some of the authors had bylines dating back to 2021.
Antiwar.com now runs all submissions through an AI-authorship check at ChatGPT. The check reviews the metadata, writing style, and biography of the author. Not only does it look for abnormalities consistent with AI authorship, but it also compares those submissions to past articles to look for patterns.
Articles that get flagged are either rejected or require a phone or video conversation with the author before publication. We do not reject the articles automatically if flagged because ChatGPT regularly states that it can only make an assessment, not a conclusion. The goal of these conversations is not only to confirm the author’s identity, but also to confirm that the author actually wrote the article.
Earlier this month, Antiwar.com received an email from OpenAI’s investigations team alerting us to a coordinated operation involving AI-generated opinion articles submitted under deceptive identities. We shared our findings with each other.
“We identified a covert influence operation from Iran that was using AI to help pitch English-language editorials to different online publications,” Ben Nimmo, principal investigator on OpenAI’s investigations team, told Antiwar.com. “They seemed to have decided that just spraying content across social media wasn’t the most effective way of reaching their audience, so they were trying to hijack real publications to get in front of readers.”
Two of those authors were the two we identified in Cluster A; none of the other authors sent submissions to Antiwar.com. In response to the email, Antiwar.com reached out to OpenAI to offer the findings of our investigation that identified several other authors, including Cluster B, which appears to be a similar operation not previously identified.
“Influence operations from Iran are a tangled web, but this looked like a different operation, trying a similar kind of approach,” Nimmo said. “It’s important to watch out for operations like this, and it’s also important to share information about them. The more the defenders stay aware and share, the harder it is for attackers to get through.”
The submission of AI-generated articles has been a major headache for Antiwar.com. As with all independent outlets, we rely on many small donations from our readers to fund our journalism. We operate with minimal staff so we can ensure the money is put to good use.
Reviewing each article for potential AI authorship has placed a burden on us to ensure that not only are the articles published at Antiwar.com factual and well-written, but also authentic.
Kelley Vlahos, Editor-in-Chief at Responsible Statecraft, told Antiwar.com that RS is also dealing with an increased number of AI-generated submissions. “We were committed to laying down a marker after the number of AI-generated pitches and stories started flooding into our inbox last year,” she said.
Vlahos added that RS maintains a strict anti-AI content policy. She explained RS “came to the conclusion that our profession as journalists was at risk.” Vlahos added, “Such tools render content bloodless, bury nuance, and are devoid of personality. These are two standards for which we will not yield: the integrity of each and every story we publish, as well as providing writing that is fresh, independent, challenging, and overall — human.”
Some publications have started taking a similar policy. The American Conservative magazine recently published a new writing policy in an article explaining to their readers why it is important to have all articles written my humans:
A computer can never be held accountable. No matter how proficient AI tools become at generating aesthetically acceptable prose, they cannot supply authority and responsibility.
The problem of AI-generated articles will likely only get worse as AI advances, and we believe that fighting against it is worth the burden that it places on our small staff. We hope that editors at other outlets will also be vigilant and work against such content to preserve the integrity of independent media and the human written word in the age of AI.


