Blaze Test Finds Pangram Spots Most AI Content
The Blaze tested the AI detection tool Pangram on text, images, and video, and reported mostly accurate verdicts with one uncertain result.
Updated on Nov. 11, 2026 in Artificial Intelligence
Live Poll
Do you trust tools that claim to detect AI-generated content online?
On November 11, 2026, The Blaze published results of tests of Pangram, an AI detection tool, across text, image, and video content. Pangram correctly identified most AI-generated material but returned a lower-certainty verdict on a short Grok-generated poem.
Why it matters
AI-generated content accounts for more than half of all LinkedIn posts and up to 45% of TikTok posts, making reliable detection a growing concern. Earlier detectors have wrongly flagged human-written work as AI, leading to false accusations against students.
Pangram charges 1 credit per 100 words of text, 5 credits per image, and 1 credit per second of video, and some features require signing in to upload files or scan URLs. Its 0.01% claimed false positive rate and the 85% verdict on the short poem are the key metrics readers can weigh against the test results.
The players
Pangram
Pangram is an AI detection company whose tool analyzes text, images, and video to estimate whether content was machine-generated.
The Blaze
The Blaze is a conservative American news and commentary outlet founded by Glenn Beck.
The details
The Blaze submitted a Grok-generated poem, outputs from ChatGPT, Claude, Gemini and Siri, its own writing, real and altered photos, and an AI-generated video of a boat under the Brooklyn Bridge to Pangram. The tool flagged the AI-generated video with 100% certainty, identified an unaltered real photo as containing no AI, and marked a slightly altered real photo as AI-detected with a heat map.
Timeline
• November 11, 2026: The Blaze published its Pangram AI detector test.
The Tech Race
Earlier AI text detectors have wrongly flagged student essays written by humans, creating false accusations that shaped how schools approach the tools. Pangram's test extends that track record by checking whether a newer commercial detector avoids the same false-positive failures across text, images and video.
Students, employers and content platforms that rely on AI detectors face decisions about how much weight to give a verdict, especially a lower-certainty one. The test suggests readers should treat a single detector result as a probability rather than proof of how content was made.
The takeaway
A detector's confidence score is an estimate, not a verdict, so it should not be the sole basis for judging whether a person or post used AI. Checking the tool's certainty level alongside other evidence helps avoid the false accusations earlier detectors have produced.
Further reading
For more on how these tools are evaluated, explore Artificial Intelligence coverage on the site.
Source note: This article includes information reported by TheBlaze.
Live Poll
Do you trust tools that claim to detect AI-generated content online?
