Amazon turns to AI to identify scammers this holiday season

On Sept. 1, Amazon rolled out a tool that lets shoppers ask its AI assistant Alexa for Shopping whether a message claiming to be from Amazon is legit or not.

Ask about an email, text, or call, and Alexa checks the details against Amazon’s full record of every communication it has ever sent using artificial intelligence.

Amazon’s its verification tool on Alexa for Shopping gives customers a definitive answer to whether a message actually came from Amazon or a scammer.

The tool targets impersonation scams like order confirmations or fake account alerts that have become “prolific” on the marketplace, Abigail Bishop, global head of scam prevention at Amazon, told Retail Brew.

“What we find is that it really doesn’t matter what the product is or what the category is,” Bishop said. “Whatever type of technique or whatever type of communication is very effective in their schemes, they will replicate that when they’re reaching out to customers.”

The verify tech uses AI to check the details of a message against a record of all communications Amazon has ever sent. If it came from Amazon, Alexa will confirm it and if not, Alexa will automatically report it to Amazon’s internal teams and offer users next steps.

“We absolutely had to leverage AI in order to be able to go through all those communications within seconds to tell you whether it’s really us or not,” Bishop said.

Amazon said it initiated takedowns of more than 70,000 phishing sites and 14,000 scam phone numbers in the last year, and can now kill phishing sites in hours and scam numbers the same day.

“What the scammer is after is not necessarily anything tied to the bait that they’re putting out there,” Bishop said. “What they’re interested in is getting your personal information and your money.”

Amazon has warned customers to be “wary” of messages relying on urgency and not to purchase gift cards as a payment form; the company said it will never ask for payment over the phone or via email.

According to Amazon, its controls block more than 99.9% of suspected infringing listings before brands have to report them. Its Counterfeit Crimes Unit has pursued over 32,000 bad actors since 2020, and in 2025 seized more than 15 million fake products.

Reviews on Amazon’s marketplace get similarly scrutinized. Each one runs through machine learning models before it gets posted, and Amazon said over 99% of products viewed on its store have only authentic reviews.