Inside the AI systems Amazon uses to protect every part of your shopping experience

Key takeaways Since 2020, Amazon's Counterfeit Crimes Unit has pursued more than 32,000 bad actors through litigation and criminal referrals to law enforcement, across 14 countries.

Amazon's AI early warning system successfully anticipated a bad actor attack on a viral, new branded product, blocking the infringing listings a full eight days before the brand owner even shared their IP with us. In 2025, Amazon’s legal actions led to the shutdown of more than 100 websites attempting to facilitate fake reviews and scams targeting our store.

For the past five years, Amazon has published an annual Brand Protection Report —a detailed look at how we combat counterfeits, protect intellectual property, and safeguard the brands that sell in our store. But global retail has never been more connected—or more complex.

Bad actors are constantly evolving their tactics, criminal networks operate across borders, and the threats facing retail extend well beyond counterfeits.

At the same time, advancements in AI are giving us capabilities that didn't exist a few years ago—allowing us to move from proactive to predictive, analyzing billions of signals simultaneously and detecting threats before they ever reach customers.

That’s why today we're releasing Amazon’s Trustworthy Shopping Experience Report, providing a comprehensive look at how we're working to protect customers, selling partners, and brands across our global store.

The report expands our commitment to trust and safety, continuing to cover brand protection and anti-counterfeiting, but now also encompassing organized retail crime, product safety, scam prevention, and trustworthy reviews. Our goal is to protect the store for customers, brands, and sellers alike.

But we understand policies designed to protect customers can sometimes create friction for sellers trying to grow their business. That’s why we’ve invested in tools such as Amazon's Account Health Dashboard, which gives sellers transparency and control into their adherence to policies, performance targets, and more.

Ensuring that legitimate selling partners can thrive on Amazon is central to our mission, and this report reflects that commitment.

How Amazon's Counterfeit Crimes Unit takes action to stop counterfeiters The Counterfeit Crimes Unit takes direct action against bad actors, either through civil lawsuits or criminal referrals to law enforcement agencies worldwide.

Our approach is built on four interconnected pillars: 1) proactive controls that stop issues before they reach customers, 2) powerful tools that anticipate risks, 3) holding bad actors accountable, and 4) protecting and educating consumers. Below are a few highlights illustrating each pillar in action: 1.

Stopping problems before they reach customers We make it straightforward for legitimate businesses to set up a selling account, but bad actors attempting to misrepresent themselves encounter intentional friction at every step. All new sellers are required to complete Amazon's robust verification process before they are allowed to sell in Amazon's store.

We then continuously monitor activity throughout a seller's journey, with our automated technology and artificial intelligence (AI) scanning billions of attempted changes to product detail pages daily for signs of potential abuse.

Multimodal systems also analyze billions of signals simultaneously—including text, images, seller behavior, and supply chain patterns—to help identify potential abuse before it reaches customers.

We also launched our direct product validation program to allow us to work directly with accredited safety laboratories to verify the compliance of products in our store.

We deployed Omniscan—an advanced machine learning system that verifies the readability and language of essential safety details at scale before products are listed—across our global fulfillment network in the U. S. , Canada, UK, Türkiye, Saudi Arabia, and Europe. The system generated image sets for more than 12 million products.

Omniscan is an advanced machine learning system that verifies the readability and language of essential safety details at scale before products are listed.

To protect the integrity of customer reviews, our systems analyze thousands of data points across billions of reviews before a review appears in the store, drawing on review data that dates back to 1995 to help detect fake or abusive content. In 2025, we proactively blocked hundreds of millions suspected fake reviews from our store. 2.

Anticipating risks before they emerge We developed an early warning system that can detect emerging threats to new brands and products before they reach our catalog.

By integrating real-time signals from social media and other retailers, we can anticipate potential infringement in some cases even before brands share their new intellectual property (IP) with us. In 2025, our teams successfully anticipated a bad actor attack on a viral, new branded pr