Amazon's AI shopping assistant is generating clicks at 0.02% the rate of standard ads, and that gap tells you everything about where conversational commerce actually is.
The Summary
- Amazon's chatbot ads have driven just 88 clicks for Paladone since January, versus 500,000 from traditional Amazon ads in the same period
- The 0.02% performance ratio reveals conversational AI commerce is still in discovery mode, not transaction mode
- Marketers are treating these as data collection experiments, not sales channels
The Signal
Amazon and OpenAI both launched conversational ad formats this year, betting that chat interfaces would become a new commerce surface. The early returns from Amazon's experiment are instructive, and not in the way the platform hoped.
Paladone, a toys and lamps seller doing $20-30M annually on Amazon, ran chatbot ads alongside their standard Amazon advertising since January. Standard ads generated 500,000 clicks. The chatbot ads generated 88. That's not a rounding error. That's a 5,682x difference in engagement between proven and experimental formats.
The story here isn't that conversational ads don't work. It's that we're still figuring out what "conversational commerce" even means in practice. ChatGPT has users who deliberately chose an AI chat interface. Amazon's shopping assistant sits inside a shopping app where people already know how to buy things. Adding a chatbot to an optimized buying flow is like adding a phone tree to a checkout counter. Nobody asked for more steps.
What Amazon is getting instead of sales is behavioral data. How do people phrase product questions? What attributes do they mention first? Which product categories trigger chat versus search? That data has value, but it's research value, not revenue. The advertisers paying for these placements are essentially funding Amazon's agent training program.
The performance gap also suggests something about user intent. People on Amazon have buying intent. They know what they want, or they're browsing with purpose. A chatbot introduces friction. People on ChatGPT have discovery intent. They're exploring, asking, learning. An ad there might feel native. Context determines whether an agent interaction adds value or just gets in the way.
The Implication
If you're a brand experimenting with conversational ad formats, treat them as user research, not performance marketing. You're learning how customers talk about your products when they're not forced to use your taxonomy. That's useful, but budget accordingly.
For platform builders, this data point is a warning shot. Conversational interfaces don't automatically improve commerce. They need to solve a problem that search and browse don't already solve. Amazon's chatbot doesn't clear that bar yet. The companies that figure out where conversation actually adds value, rather than where it sounds futuristic, will own this space.
Source: The Information