An AI content detection company based in New York, focused primarily on text and recently expanded into image and video. The text detector returns an AI-generation probability and claims a false-positive rate of 0.0041% (roughly one in 24,000 documents), with independent validation cited from The New York Times and a University of Chicago study; Substack uses it. Pangram Image 1.0 detects AI-generated images and localizes synthetic regions with a heatmap, and Image Flash 1.0 targets video, scoring 98.7% on the DF26 academic benchmark. Pricing: Free (2,000 words/day, no payment method), Individual $20/month or $15/month annual (300,000 words + 100 image checks), Professional $65/month or $45/month annual (1,500,000 words + 500 image checks), Teams/Enterprise custom. API access is billed separately at $0.05 per 100 words and is not included in any subscription. However, PCMag’s hands-on testing found the image detector could be fooled by simple edits (cropping, added noise) and mislabeled an AI-generated image as 100% human, and that a free “AI humanizer” pushed ChatGPT-generated poetry to 100% human. The company raised $9M in July 2026 (Menlo Ventures led), bringing total funding to roughly $13M.
**Pros & Cons**:
Pros:
Text detection has independent validation (NYT, University of Chicago) and a very low claimed false-positive rate, stronger external evidence than most competitors in this category; Substack adoption is a real production reference; image detection adds heatmap localization of synthetic regions, not just a binary verdict; video detection scores 98.7% on the DF26 academic benchmark; generous free tier (2,000 words/day, no card) supports real evaluation; company is funded ($9M, Menlo Ventures) and actively shipping new models.
Cons: Image and video detection are materially weaker than text—PCMag bypassed the image detector with cropping and added noise, and it mislabeled an AI image as 100% human; a free “AI humanizer” pushed ChatGPT poetry to 100% human, so text detection is not humanizer-proof either; API is billed separately ($0.05/100 words) and is not included in any subscription**, which is a common point of confusion; the product straddles text, image, and video detection with uneven maturity across them, so a single “AI detector” label overstates consistency; independent third-party reviews beyond PCMag and press coverage are limited, and no G2/Capterra ratings surfaced; the detection arms race means published accuracy can shift as evasion tools evolve.
**Who is it for?**:Educators, editors, and publishers who need a text-focused AI detector with independent validation and a low false-positive claim; platforms like Substack-style newsletters that need to flag AI-generated submissions at scale; researchers and journalists who need an API for bulk text detection and will pay separately for it; teams evaluating AI image detection who understand its current limits and will combine it with human review.