How Marketers Should Target AI Shopping Agents

The marketing industry has always focused on consumers... after all, in business school we're taught persuasion tactics and strategies to attract customers. But now AI shopping agents have entered the chat and they're searching for products, not your consumers.

Harvard Business Review released a study by Jafar Sabbah and Oguz A. Acar which found that tricks that have worked on people, don't work the same on AI shopping agents.

The researchers tested eight common e-commerce promotions on four AI models: GPT-4.1-mini, GPT-5, Gemini 2.5 Pro, and Gemini 2.5 Flash Lite. They ran thousands of simulated shopping rounds. In each one, the agent picked between products that carried typical badges and promo overlays.

The results were surprising. Only one tactic worked on agents the way it works on humans: product ratings. Scarcity cues, countdown timers, strike-through pricing, and vouchers often failed. Some even backfired. Higher prices pushed agents away from a product. The more advanced reasoning models were also more skeptical, and in some cases seemed to treat loud persuasion as a red flag. The authors also surveyed 50 e-commerce executives in the US and UK. Many had already noticed traffic or conversion shifts they link to AI agents.

The Top Elements AI Shopping Agents Look For When Searching For Products. Source: FMI LinkedIn

This makes sense when you think about why a countdown timer works on you. It triggers fear of missing out. An agent has no feelings. It has a job: find the best option for the person who sent it. Ratings and price are facts it can use.

A fair note. This is a simulation, not real checkout data, and it covers four models. The authors also warn that results can change as models update, so brands need to keep testing. But the direction is clear. I’d rather prepare early than react late.

Here’s my vision. Brands should stop choosing between human marketing and agent marketing. They need both. Think of it as two storefronts. The first is for humans. It has story, emotion, community, and design. This is where people fall in love with a brand. The second is for agents. It has clean product data, clear prices, honest specs, and strong customer reviews. No fake urgency. If an agent can’t read or trust your page, you may never make the shortlist. Your story gets people to love you. Your data gets agents to pick you.

First, treat reviews like ad spend. Ratings were the one thing that worked, so earning great reviews is now core marketing. Second, build a testing habit. Replay your product pages against different AI models every time they update.

If you’re headed into marketing, this changes which skills matter. Data quality, testing, and clear writing move up the list. Learn how AI reads a page, not just how people do.

Check out the video I designed and comment if you would trust an AI shopping agent for your shopping needs!