WIRED

I tested AI detection on my Gmail inbox

Pangram's AI detector scanned 108 email threads and found more than half were probably machine-written, mostly press releases.

By ·

Nov 9, 2026

7 cards · 1m 50s · tap to begin

From · · · 1m 50s

I tested AI detection on my Gmail inbox

Pangram's AI detector scanned 108 email threads and found more than half were probably machine-written, mostly press releases.

In brief

Pangram's AI detector scanned 108 email threads and found more than half were probably machine-written, mostly press releases. Pangram's Gmail extension detected AI-written emails in a user's inbox with apparent accuracy, flagging over half of messages as likely machine-generated. Originally reported by WIRED.

AI is everywhere now

AI-generated content fills restaurant menus, student essays, movies and social media. In some digital spaces it overwhelms human-made content entirely.

Detecting AI in your inbox

Pangram makes a browser extension that sits in Gmail and runs incoming messages through algorithms to detect AI writing. The twenty dollar per month individual plan includes Gmail integration.

How to set up Pangram

How to set up Pangram

Connect your Gmail account through Pangram's integration page. The tool scans your one hundred most recent email threads and any new messages after that. Email scanning doesn't count against your monthly scan credit limit.

What Pangram labels show

After scanning, your Pangram dashboard shows how many emails are suspected AI, probably human, or mixed.You can apply these as Gmail labels, send AI emails to archive or spam, and exclude specific domains or senders.

The inbox results

Out of one hundred eight threads scanned, eight were labeled AI, thirty-seven got the human label, and the remainder were marked as mixed. Press releases were the most common AI culprits.

Accuracy on known emails

Emails from friends, family and colleagues all passed the human test. An email written with Microsoft Copilot was correctly flagged as AI. A section of the article written by hand was marked human.

Limits of the detector

Pangram won't decide on emails below a minimum word count. Only the first email in each thread is scanned, so if someone switches to AI halfway through a conversation it won't be spotted.

In short

Pangram's Gmail extension detected AI-written emails in a user's inbox with apparent accuracy, flagging over half of messages as likely machine-generated.

Read the original on WIRED

How we make summaries

React

Sign in to react and comment.

Comments (0)

Life is short. Keep it sweet. Respect others' opinions and be kind!

    Recommended next

    More decks on pangram and related topics.