{"data":{"id":2076,"lang":"en","title":"Substack Launches AI-Detection Tool With Pangram to Flag Machine-Written Posts","summary":"Substack partnered with Pangram to let readers detect how much of a post was AI-generated, launching on web and iOS today.","category":"ai","category_label":"AI","published_at":"2026-07-26T14:02:05Z","updated_at":"2026-07-26T14:02:05Z","url":"https://techtools.cz/tools/news/article/2076-substack-launches-ai-detection-tool-with-pangram-to-flag-machine-written-posts","api_url":"https://techtools.cz/news-api/v1/articles/2076?lang=en","markdown_url":"https://techtools.cz/news-api/v1/articles/2076.md?lang=en","image_url":"https://www.engadget.com/img/gallery/substack-is-adding-an-ai-detection-feature/intro-1784660001.jpg","word_count":152,"reading_time_minutes":1,"sources":[{"url":"https://www.engadget.com/2220064/substack-is-adding-an-ai-detection-feature/","domain":"engadget.com"},{"url":"https://www.axios.com/2026/07/23/substack-subscribers-ai-generated-content-pangram","domain":"axios.com"}],"translation":{"lang":"cs","id":2149,"title":"Substack spouští nástroj na detekci AI od Pangramu, který odhalí strojově psané příspěvky","url":"https://techtools.cz/tools/news/clanek/2149-substack-spousti-nastroj-na-detekci-ai-od-pangramu-ktery-odhali-strojove-psane","api_url":"https://techtools.cz/news-api/v1/articles/2149?lang=cs"},"content":"Substack has partnered with AI-detection startup Pangram to let readers see how much of a post, comment, or reply was written by artificial intelligence. The feature is live today on web and iOS (Android to follow) and scans text over 100 words published from launch onward.\n\nCreators can also add a statement disclosing how they used AI in a piece, run Pangram on drafts before publishing, or flag scan results they believe are wrong. In its announcement, Substack coined the term \"Claudefishing\" for using AI to fake a human connection with readers, taking a swipe at how platforms like LinkedIn have handled synthetic content.\n\nPangram says its latest detection model, Pangram 3.3, achieves a 0.01% false-positive rate, but Substack itself acknowledged that AI detectors remain imperfect — The Atlantic has reported that even leading tools like Pangram are \"far from perfect.\" Substack summed up its philosophy simply: \"people should know what they're getting.\""}}