Item logo image for Local AI Detector

Local AI Detector

ExtensionTools
Item media 4 (screenshot) for Local AI Detector
Item media 1 (screenshot) for Local AI Detector
Item media 2 (screenshot) for Local AI Detector
Item media 3 (screenshot) for Local AI Detector
Item media 4 (screenshot) for Local AI Detector
Item media 1 (screenshot) for Local AI Detector
Item media 1 (screenshot) for Local AI Detector
Item media 2 (screenshot) for Local AI Detector
Item media 3 (screenshot) for Local AI Detector
Item media 4 (screenshot) for Local AI Detector

Overview

Free, open-source, fully local AI-content detector. No servers, no API keys.

Local AI Detector checks the text and images on the page you're reading for signs of AI generation, entirely on your device. No servers, no API keys, no accounts, no data collection. WHAT IT DOES • Scores the text on the page with open-source detection models that run in your browser, highlights the sentences most likely to be AI-written, and shows a calibrated probability rather than a made-up score. • Click a highlighted sentence to see its score and what each model said. • On comment threads, reviews and forums it scores each comment separately, and an optional filter dims the ones flagged as likely AI. • On YouTube it checks the video's transcript and, experimentally, the narrator's voice. The audio never leaves your device. • Checks images for signed Content Credentials (C2PA), AI-generator metadata and open-source invisible watermarks, and says plainly which watermarks can't be checked locally. • Flags hidden or invisible characters in text, shown separately from the AI score. • Runs automatically on a set of popular sites with lots of user-written text; on any other page, click the toolbar button or right-click to check it. • A fast check runs by default; a one-click Deep check reads the whole page with the most accurate model pair. PRIVACY All analysis runs in your browser. The only downloads are the detection models themselves (once, after you agree, 35–400 MB depending on your choices) and, only on a site you've explicitly allowed, the bytes of an image you ask it to check. Nothing you read is ever sent anywhere. ACCURACY This is not a forensic tool. On held-out web text the default models flag about two-thirds of AI-written texts and about 5% of human-written ones. Paraphrased or edited text often passes, and a low score means "no strong AI signal", not "written by a human". The full measurements are published with the source code. OPEN SOURCE MIT licensed. Every bundled model and library is openly licensed. Source code, documentation and the privacy policy are linked from the Options page.

Details

  • Version
    0.2.0
  • Updated
    October 2, 2026
  • Offered by
    fredgaffey08
  • Size
    9.9MiB
  • Languages
    English
  • Developer
    Email
    fredgaffey08@gmail.com
  • Non-trader
    This developer has not identified itself as a trader. For consumers in the European Union, please note that consumer rights do not apply to contracts between you and this developer.

Privacy

Manage extensions and learn how they're being used in your organization
The developer has disclosed that it will not collect or use your data. To learn more, see the developer’s privacy policy.

This developer declares that your data is

  • Not being sold to third parties, outside of the approved use cases
  • Not being used or transferred for purposes that are unrelated to the item's core functionality
  • Not being used or transferred to determine creditworthiness or for lending purposes

Support

For help with questions, suggestions, or problems, visit the developer's support site

Google apps