AI detection startup Pangram has introduced new tools to help users identify AI-generated images and videos, offering a potential way to separate authentic content from the growing flood of ‘AI slop’.
Known for its widely used tool to detect AI-generated text, Pangram said that users can now upload videos and get an assessment of whether AI tools were used to create or edit the clips. The New York-based company said it has also upgraded its AI image detection tool with new capabilities.
Pangram’s new AI video detection and latest image detection capabilities are currently only available to paid subscribers. However, free tier users can run a handful of images through its detector even if they have not signed up for a paid plan.
Its expansion of AI detection capabilities comes at a time when generative AI continues to permeate more and more forms of communication. Roughly 50 per cent of news articles and blogs on the internet are AI-generated, according to a recent analysis by digital marketing firm Graphite.
Pangram is among the growing crop of startups competing to be the best at distinguishing AI from human-generated content. While the accuracy rates of AI detectors tends to vary, Pangram’s tool has proved effective at detecting whether a full piece of text was written using an AI tool or not.
It beat rivals such as Originality.ai and GPTZero by maintaining near-perfect accuracy in assessing texts that were roughly 500 words or more, according to a 2025 study by the University of Chicago.
Noting the challenges in building an image detection model, Pangram CEO Max Spero wrote in a post on X, “There are way more AI image/video models than frontier LLMs, and their outputs are more diverse. Most phone cameras, including iPhones, do some level of AI postprocessing to add detail to real photos. We had to build a model that’s robust to this postprocessing while understanding the difference between AI edits and fully AI-generated visuals.”
We’ve been working towards AI image and video detection for months, and now it’s finally here.
There were a lot of challenges along the way. There are way more AI image/video models than frontier LLMs, and their outputs are more diverse. Most phone cameras, including iPhones, do some level of AI postprocessing to add detail to real photos.
We had to build a model that’s robust to this postprocessing while understanding the difference between AI edits and fully AI-generated visuals.
I think the heatmap feature is a really important step forward and a huge differentiator over other models which typically report scores as a binary confidence.
Please try it out!!
— Max Spero (@max_spero_) November 4, 2026
Spero further said that the company delayed the release of its AI image detection model so it could test it on several iPhone 18 Pros because its model initially flagged images from the phones’ cameras as AI.
Story continues below this ad
Upgraded AI image detection model
While its image model, Pangram Image 1.0, was previously released in research preview, the company said the model is now available via its Chrome extension. The extension can automatically scan images as users browse and flag ones that are likely to be AI-generated or AI-modified.
In terms of accuracy, Pangram Image 1.0 was able to flag more than 99 per cent of AI-generated images with a false positive rate of 0.27 percent in its testing, as per the company. However, there are some caveats to this.
When it comes to images with resolutions below 512 x 512, “results will come with a reduced confidence disclaimer,” it said.
New AI video detection tool
The tool is capable of distinguishing between clips that are ‘partially AI’ and those that are entirely AI-generated. It also shows users a heatmap that points out exactly which aspects of the video are likely to be AI-modified.
Story continues below this ad
The underlying AI model behind its video detection tool “particularly excels at frontier AI video detection,” Pangram said. In internal testing, the video detection tool was able to flag AI-modified content with more than 98 per cent accuracy, according to the company.