# Testing Pangram’s New AI Detection Tool for Gmail Inboxes A hands-on look at how Pangram identifies AI-generated content within personal email threads. - **Source:** [Wired – Gear](https://www.wired.com/category/gear) - **Published:** 2026-11-09 - **Topic:** [technology](https://brifing.si/c/technology) - **Original article:** - **On Brifing:** https://brifing.si/s/96f06320-4fc6-4781-b540-4e31a5cd49f8 ## AI Detection Enters Gmail The Pangram AI detection tool has expanded its functionality to integrate directly with Gmail inboxes. By installing a browser extension, users can now automatically scan incoming messages to determine if they were composed by artificial intelligence or human authors. This service requires a paid subscription, specifically the twenty-dollar monthly individual plan. ## Privacy and Data Handling Pangram states that email contents are not stored after the assessment process is complete. Furthermore, the company claims these messages are not utilized to train their detection models. Users maintain control over the integration and can remove the plug-in at any time, providing a layer of security for sensitive inbox communications. ## Scanning Scope and Limitations Once connected, the tool scans the one hundred most recent email threads and monitors all subsequent incoming messages. It does not process historical archives beyond this initial set. The system also features a minimum word count threshold, as AI detection becomes significantly more difficult to perform accurately on very short email messages. ## Categorization and Labeling Features The dashboard categorizes emails as AI-generated, human-written, or a mix of both. Users can optionally apply these labels directly within the Gmail interface for better organization. Additionally, the tool allows for automated actions, such as sending suspected AI-generated emails directly to the archive or spam folder to reduce inbox clutter. ## Customizable Filtering Options To prevent false positives or unwanted scrutiny, users can exclude specific domains, subject keywords, or individual senders from the scanning process. This is particularly useful for ensuring that personal communications from family members or professional correspondence from a manager are not subjected to the same algorithmic analysis as generic newsletters or press releases. ## Real-World Performance Results In a test of 108 threads, eight were flagged as AI, while 37 were identified as human, with the remainder labeled as mixed. Press releases were the most frequent AI culprits. Notably, the tool correctly identified a message written with Microsoft Copilot as AI, while accurately labeling human-authored text as human. ## Operational Constraints and Nuance The tool only analyzes the first email in each thread, meaning it cannot detect if a correspondent switches to AI mid-conversation. While the background scanning is convenient, the high volume of mixed or AI-labeled content highlights how pervasive automated writing has become in modern digital communication, often blurring the lines between human interaction. ## Impact on Digital Communication Knowing the origin of an email significantly alters the reader's mindset and engagement level. As AI-generated content continues to proliferate, tools like Pangram provide a way to maintain transparency in digital correspondence. Whether this becomes a standard utility depends on how much users value distinguishing between human and machine-generated communication. ## Sources covering this story - https://www.wired.com/story/i-let-pangram-loose-on-my-gmail-inbox-heres-how-much-ai-it-found/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._