It is not always easy to tell whether the messages filling your inbox were composed by a real person or generated by a machine. From store newsletters to emails sent by friends, family, and colleagues, artificial intelligence has become a common tool for daily communication. Understanding how much of your digital correspondence relies on automated text is becoming increasingly difficult without specialized help.
Extending AI Detection to Gmail
The Pangram AI text and image detector tool has now expanded its reach directly into your Gmail inbox. The company produces a browser extension that sits inside Gmail, running incoming messages through its specialized algorithms to spot artificial writing. To use this integration, users need at least the individual plan priced at $20 per month, as the feature is not included in the free tier.
Connection Process and Privacy
You can activate the integration by heading to the Gmail integration page on Pangram and clicking the button to connect your account. Alternatively, you can locate the Gmail Integration option on the left-side navigation pane within your main web dashboard and complete the standard authorization prompts. Pangram states that the contents of your emails are not stored after being evaluated and are not used to train models designed for text detection. While this requires a certain level of trust, the plug-in can be removed from your Gmail account at any time.
Once connected, Pangram begins scanning your 100 most recent email threads along with any new messages that arrive afterward. It cannot process the entire historical archive of every email you have ever received. Fortunately, running these scans does not count against your monthly scan credit limit, ensuring you are covered regardless of inbox volume. After the initial scan, your dashboard interface updates to show a breakdown of suspected AI messages, probable human messages, and those flagged as a mix of both. These categories can optionally be applied as labels inside Gmail, which remain even if you uninstall the tool later.
Filtering and Management Options
Users can configure the system to automatically send incoming AI-written emails straight to the archive or spam folder, keeping automated content out of sight. You can also exclude specific domains, subject keywords, sender addresses, and Gmail labels from the scanning process if you prefer certain correspondence, such as messages from family or managers, to remain unmonitored.
When put to the test across 108 scanned threads, eight were labeled as AI, 37 received the human designation, and the rest were marked as mixed. Discovering that a significant portion of incoming messages likely originate from artificial intelligence can be disheartening, with press releases acting as the most frequent culprits. In terms of accuracy, none of the messages verified as human-written were incorrectly flagged as artificial, showing reliable performance on a small sample size.
Testing with self-sent messages created using Microsoft Copilot resulted in a correct AI label, while a manually typed excerpt from this article was accurately marked as human. There is a minimum length threshold below which the detector cannot make a determination, and only the initial message in each thread is scanned. The background operation of the tool makes it a seamless addition for long-term use, fundamentally changing how one perceives digital correspondence in an era of automated replies.
