AI Agents Are Changing the Internet—and Websites May Soon Be Designed for Them
The internet is about to be reshaped by personal AI agents that can book flights, reserve hotel rooms, negotiate prices and shop online for consumers. These agents have recently been released at scale through apps such as OpenAI’s Dots and Meta’s Muse. Muse was downloaded 5 million times in its first month, while agentic web traffic tripled from August to September.
Combining AI agents with human visitors does not work well on an internet designed primarily for people. Many websites block the AI agents that consumers use to negotiate prices, plan vacations or complete other legitimate tasks. Still, if demand for AI agents continues to grow, the internet may increasingly be redesigned for them rather than for humans.
AI Agents Are a New Category of Web Visitor
AI agents represent a third category of web visitor. They use the internet at machine speeds rather than human speeds, but they are not traditional bots because they act on behalf of individual people to perform useful, authorized tasks. They can complete online activities much faster than humans.
Because AI agents do not fit neatly into the systems used to distinguish bots from humans, some platforms—particularly those focused on security—have chosen to ban them altogether. TechCrunch recently reported that people are struggling to use agents on websites such as Pizza Hut and Yelp, where AI cannot make restaurant reservations.
Amazon has blocked Meta’s Muse agent and is also engaged in a legal battle with AI startup Perplexity over its agent. Although a judge recently ruled in favor of Perplexity, Amazon has brought legal action against it. In that case, the company claims that AI agents threaten its advertising business by making it difficult to distinguish between ads seen by humans and ads served to bots.
Why Websites May Eventually Welcome AI Agents
These blocks are a nuisance for early adopters, but Ethan Mollick, an associate professor at the Wharton School of the University of Pennsylvania, predicts that they could eventually affect the entire web. If AI agents become common tools for requesting refunds or negotiating prices, they could overwhelm customer service departments until human assistance is no longer readily available.
“Humans will ultimately become very frustrated if agents are not allowed to do this on their behalf,” Mollick wrote recently on LinkedIn.
As demand for personal AI agents grows, platforms will have strong incentives to accommodate them. Most agent traffic currently goes to e-commerce websites, and allowing large numbers of AI agents to shop for real people could create a significant business opportunity. Shopify has already welcomed agents by providing a dedicated checkout lane that bypasses pages designed for human visitors.
Will the Internet Become Worse for Human Users?
The immediate question is what will happen to the rest of the internet. The web has increasingly been designed to capture and monetize human attention, even though people remain its intended users. That could change if websites become optimized for AI agents that outnumber human visitors.
Websites could increasingly become less useful as destinations for real people. Users who want to browse and compare products themselves may encounter worse interfaces, while verification systems become more intrusive or incomplete. People might be asked repeatedly to prove they are human—or even be scanned by the iris to do so.
Pricing could also become less transparent. Instead of everyone seeing the same price for a sweater, AI agents with different profiles could receive different offers. Consumers may also be unable to tell whether an agent’s recommendations are based on their needs or influenced by advertising and platform commissions.
Personal AI agents promise to make online tasks faster and easier, but their rapid adoption could force the internet to serve two very different audiences. The challenge will be creating a web that allows AI agents to act on behalf of people without making the online experience less transparent, accessible and useful for the humans they represent.
Source: www.technologyreview.com


