When you wanted to buy something online your first port of call used to be a search engine. Now you’re just as likely to turn to an AI agent that will rank options and maybe even start the purchase process for you.
A new joint study from DataDome, Botify, and AWS looks at 6,000 nationally representative consumers across the UK, US and France, to find insights to help retail leaders navigate the transition to agentic search and discovery.
It finds 38 percent have used AI assistants for shopping tasks like product ideas, suggestions and comparison. At the same time AI-driven bot activity increased more than five fold in 2025 as platforms scape retail websites for model training, live retrieval and agent-led discovery.
A more worrying finding is that around 80 percent of websites are not protected against AI agent spoofing, exposing retailers to distorted analytics, inflated AI referral signals, misinformed commercial decisions, and fraud. 80 percent of AI agents fail to properly identify themselves too.
There are advantages to retails in allowing agent access, increased visibility when trusted agents surface accurate product information and more seamless discovery journeys that reduce friction for the shopper for example. But there are drawbacks too such as a risk of malicious crawling that exposes sensitive or strategic data, and distorted analytics when automated traffic is not separated from human behaviour.
The report’s authors conclude:
Today, AI systems are shaping how shoppers search, discover, compare and purchase products. Their influence will grow as tools become more capable and trusted.
Importantly, retailers need a mindset shift in the age of agentic retail. They need to develop an understanding of how their digital estate operates in this new era, and must now consider investing in structured, trustworthy data, and clear traffic policies for AI agents to remain visible and relevant as more power shifts to algorithms.
Early-mover advantages are emerging, but brands that delay action risk becoming harder to find, harder to trust, and easier to replace.
You can find out more on the DataDome site.
Image credit: BiancoBlue/depositphotos.com
