Amazon's AI shopping assistant can now spot fake Amazon emails and texts

Amazon has rolled out a new AI feature that helps customers figure out whether an email, text message, or phone call genuinely came from the company, according to The Verge. Built into the Alexa for Shopping assistant, the tool is aimed at impersonation scams in which fraudsters pose as Amazon to trick people into handing over account details or one-time passwords.

How the verification works

Customers can ask Alexa for Shopping about a specific message they received. The assistant then checks it against a log of all the communications Amazon has actually sent, and also examines the message's contents, its formatting, and who sent it, The Verge reports.

The system is deliberately conservative. Amazon says the assistant will only vouch for a message when it is entirely sure of the result. In an example the company provided, a customer asks whether Amazon really texted a one-time password from the number 98626 at around 4:50pm; the assistant responds that the message was a genuine Amazon one-time password.

What happens with suspicious messages

When the assistant concludes a message did not come from Amazon, it does not leave the customer guessing. It tells them to check their orders inside the Amazon app and to contact Amazon's support team directly, steering them away from the suspicious message itself.

Part of a broader effort

The Verge notes that Amazon introduced a related tool earlier this year that lets people forward questionable messages to [email protected] for an authenticity check. The new assistant capability is also one of several recent additions to Alexa for Shopping, which gained an "Update Me When" feature that can track new product releases.

Why it matters

Brand-impersonation scams are among the most common forms of phishing, and a retailer of Amazon's scale is an obvious target. Fake delivery updates and bogus one-time-password requests work precisely because they imitate the real messages customers are conditioned to expect.

This feature attacks the problem with a signal no scammer can replicate: Amazon's own record of what it sent and when. Instead of training users to hunt for typos or inspect sender addresses, the company is performing the comparison itself with AI, and its rule of only confirming authenticity when fully certain is meant to prevent the most damaging failure mode for a scam-checking tool, namely giving false reassurance about a fraudulent message.

There are limits. The check only covers messages claiming to be from Amazon, not the countless scams that impersonate banks, couriers, or government agencies. Still, as AI assistants take on larger roles in shopping, using them to authenticate the exact kind of message that most often fools customers is a practical application of the technology, and a template other large brands under similar pressure from impersonators could follow.