The internet has traditionally been designed around one basic assumption: a person is eventually sitting behind the browser.
Cloudflare believes that assumption may soon become outdated. The company’s CFO, Thomas Seifert, predicts that within five years, non-human internet traffic could become up to 1,000 times larger than human traffic. Humans would become what he described as a “rounding error” online, not because people stop browsing, but because automated activity grows much faster.
Bots have already passed an important milestone. Cloudflare data showed earlier this year that automated requests had overtaken human web traffic. Now, AI agents could dramatically accelerate that change.
For e-commerce, this raises a practical question: what happens when the main reader of a product page is no longer necessarily the shopper but an AI acting on their behalf?
Traditional shopping creates relatively limited web traffic. Someone looking for a camera might visit several retailers, compare a few models, read reviews, and make a decision.
An AI shopping agent can behave very differently.
Cloudflare CEO Matthew Prince offered an example in which a person might check five retailers, while an agent could examine 5,000 websites on the shopper’s behalf. A single request to find the best product could therefore generate thousands of interactions across retailers, marketplaces, review sites, and other sources.
Multiply that by millions of consumers using agents for product research, travel, price comparisons, and other tasks, and the scale changes quickly.
This is not the same bot traffic businesses have dealt with for years. Traditional crawlers generally follow predictable patterns. Agentic systems can browse more like people, but operate at machine speed and repeat actions on a much larger scale.
Retailers have spent years optimizing e-commerce experiences for humans. Product pages need attractive images, useful descriptions, intuitive navigation, and persuasive content.
Those elements remain important. However, AI agents introduce another audience with different requirements.
An agent comparing hundreds of laptops does not experience a beautiful product page in the same way a person does. It needs information that it can reliably identify, interpret, and compare: model numbers, dimensions, compatibility, technical specifications, availability, prices, and other structured attributes.
That creates an interesting change in e-commerce. Product information increasingly needs to work simultaneously for human shoppers and machines.
The storefront still needs to convince the customer. The underlying product data needs to convince the agent that it matches the customer’s request.
This also connects with an issue we have seen in recent AI mistakes across retail.
AI can produce strange results when it works with incomplete information, weak context, or poorly governed data. When AI systems operate at machine scale, those weaknesses can spread much further.
Imagine an incorrect specification appearing in a catalog. A human shopper may notice the inconsistency or check another source. An AI agent comparing thousands of products could instead ingest that information, use it in comparisons, and potentially pass it to other systems.
Scale therefore increases the value of accuracy.
If agents become responsible for a larger share of product discovery, structured and verified product information will matter not only for presenting products correctly but also for making them understandable to automated systems.
There is another side to Cloudflare’s prediction: websites need to handle the traffic itself.
Thousands of automated requests for a single shopping task create costs for retailers and publishers. Businesses will increasingly need to distinguish among useful AI agents, search crawlers, malicious bots, and automated systems that consume resources without providing value.
That could change how websites manage access to their content. Instead of simply asking whether traffic is from a human or a bot, businesses may need to determine which type of bot is requesting information and what it intends to do with it.
For e-commerce, legitimate shopping agents could become valuable visitors. Blocking them completely could mean disappearing from AI-powered product discovery.
Cloudflare’s 1,000-to-one prediction is not a certainty. Seifert himself acknowledged that previous forecasts have been wrong. However, Cloudflare previously expected bots to surpass humans later than they actually did, and AI agents are creating a fundamentally different type of automated demand.
For brands and retailers, preparing for this environment does not mean abandoning human-centered e-commerce. It means recognizing that machines are becoming participants in the shopping journey.
Product content will increasingly need two qualities at once: rich enough to help people understand and trust a product, and structured enough for AI agents to interpret it accurately.
If agentic traffic grows anywhere close to Cloudflare’s prediction, optimizing e-commerce only for human visitors will eventually mean optimizing for only part of the audience.
Read further: News, AI, Cloudfare, e-commerce, ecommerce, Icecat