Making the internet better for AI agents might make it worse for everyone else
This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here.
Traditionally, websites have treated all visitors as either bots—automated programs performing tasks like indexing web pages—or people. Bots follow strict rules to avoid disrupting human experiences, while humans navigate sites designed for their needs. However, this division is changing with the rise of personal AI agents, which act on behalf of individuals to perform tasks such as booking flights or hotels. Unlike bots, these agents are not fully automated; they operate under human instruction but at speeds far exceeding human capability. Their introduction creates a third category of web visitor, complicating existing systems that distinguish bots from people.
The use of personal AI agents has surged since the launch of apps like OpenAI’s Dots and Meta’s Muse. Muse was downloaded 5 million times in its first month, and agentic web traffic—traffic generated by AI agents—tripled from August to September 2026. This growth reflects increasing consumer demand for AI-assisted tasks. However, the influx of agents has led to conflicts, as many websites block them from accessing services like price negotiations or reservations. Platforms such as Pizza Hut, Yelp, and Amazon have either restricted or taken legal action against AI agents, arguing that they disrupt business models, particularly advertising revenue.
Websites block AI agents for several reasons, including security concerns and the potential to overwhelm customer service systems. For example, Amazon has blocked Meta’s Muse agents and is engaged in a court battle with Perplexity, a startup developing AI agents, over whether agents should be allowed to access its platform. Amazon claims that agents threaten its advertising model by making it difficult to distinguish between ads viewed by humans and those seen by bots. These disputes highlight the tension between accommodating AI agents and protecting existing business practices.
Experts like Ethan Mollick, an associate professor at the Wharton School of the University of Pennsylvania, warn that if AI agents become widespread, they could overwhelm websites designed for human users. For instance, agents requesting refunds or negotiating prices might clog customer service channels, making them unusable for real people. As demand for agents grows, platforms may prioritize them over human users to capitalize on e-commerce opportunities. Already, Shopify has created separate checkout lanes for AI agents, bypassing pages designed for humans.
The internet may evolve into a space optimized for AI agents rather than humans, leading to a less user-friendly experience for people. Potential changes include neglected websites, worsening interfaces for human browsing, and increased verification systems to distinguish humans from agents—such as iris scans. Prices could become personalized based on the agent accessing them, and users might struggle to identify whether recommendations are influenced by advertisements or platform commissions. This shift contrasts with the promised benefits of AI agents, such as helping consumers fight medical bills or find affordable insurance.
The rules governing how AI agents interact with websites are still being developed. Some regulations, like preventing agents from exposing credit card information, are widely supported. However, contentious issues remain, such as whether businesses can block agents or restrict their ability to negotiate better deals for consumers. These decisions will shape whether the internet becomes a tool that benefits humans through AI assistance or a space where agents work against human interests. The outcome depends on the balance between business needs and consumer rights.
