AI-written op-eds are appearing in major newspapers, but detection is tricky

AI-written op-eds have quietly entered the opinion pages of The Wall Street Journal, The Washington Post, and The New York Times, according to a Semafor analysis using the AI detector Pangram. Of 310 guest submissions over the past month, 10 were flagged as at least 80% AI-generated, and another 40 showed partial AI involvement. For builders shipping AI writing tools, the real story is not the raw numbers but the detection biases, editorial policies, and credibility debates that will shape how these tools are adopted in professional publishing.

What the Pangram analysis found

The most prominent case was billionaire Stanley Druckenmiller's op-ed critiquing Treasury interventions, which scored 100% AI by Pangram. Druckenmiller told NOTUS he uses AI for all his writing, comparing it to using a calculator. The Wall Street Journal's opinion editor Paul Gigot defended the practice, saying AI is "a fact of modern life" and that what matters is the author's original argument and standing.

Other flagged pieces include an Aug. 2 New York Times guest essay by former CISA director Jen Easterly and an Aug. 9 Washington Post column by Dartmouth provost Santiago Schnell, which Pangram identified as 100% AI-written. Schnell acknowledged using ChatGPT to refine arguments and check grammar, noting the piece went through multiple rounds of Post edits. The Financial Times also appended a disclosure to a guest column after readers questioned AI involvement.

Why detection bias matters for builders

Stanford research has shown that AI detection tools sometimes flag work by people whose first language is not English. Schnell, a native Spanish speaker, pointed to this bias. Pangram claims a 0.5% false positive rate, but independent evaluations remain limited. For builders integrating AI writing features, relying on any single detector as a gatekeeper risks misclassifying legitimate human writing, especially from non-native speakers. The practical takeaway: detection is probabilistic, not definitive, and should be paired with human review and transparent disclosure policies.

Editorial policies are diverging

The three papers have notably different stances. The New York Times prohibits AI in developing and drafting guest essays. The Washington Post requires guest writers to confirm their submission was "not created or manipulated with artificial intelligence or editing software." The Wall Street Journal, by contrast, accepts AI as a tool as long as the author's argument and credibility are intact. These differences create a fragmented landscape for any AI writing product targeting opinion journalism. Builders should expect publishers to adopt increasingly specific AI policies, and tools that offer transparent disclosure features may have an advantage.

What remains unclear

The analysis rests on a single investigation using one detection tool, and Pangram's accuracy has not been independently verified at scale. The sample size is small, and the 80% threshold may miss subtler AI assistance. Additionally, Michigan researchers found that fine-tuned AI models can produce text that MFA-trained experts prefer over human writing, suggesting detection will only get harder. For now, builders should treat AI detection as a probabilistic signal, not a ground truth, and watch for policy shifts that could open or close distribution channels for AI-assisted content.