Personal AI agents can now book flights and hotels, and their numbers are rising fast. Meta’s Muse app hit 5 million downloads in its first month, pushing agentic web traffic up threefold between August and September 2026.
This surge is clashing with a web built for humans. Sites are increasingly blocking these tools, yet demand keeps growing. If this trend continues, the internet will be redesigned for machines instead of people.
A third category of visitor
These agents form a distinct group. They are not humans, nor are they traditional bots. They act on behalf of individuals to perform legitimate tasks, but they do so at machine speed. Current systems for distinguishing users do not account for them, leading many security-conscious platforms to ban them entirely.
Some major sites have already taken action. TechCrunch reported in October that users struggled to deploy agents on websites ranging from Pizza Hut to Yelp, which refuses to let AI make restaurant reservations. Amazon blocked Meta’s Muse and is fighting a legal battle with Perplexity to stop its agents as well. Although a judge ruled for Perplexity, Amazon has filed an appeal arguing that these agents threaten its advertising business by blurring the line between ads seen by humans and those served to bots.
Why blocks will fail
While these restrictions annoy early adopters, they are unlikely to hold. Ethan Mollick, an associate professor at the Wharton School of the University of Pennsylvania, predicts the problem will spread to the entire web. Once it is common knowledge that agents can request refunds or negotiate prices, customer service departments will become overwhelmed. Humans will eventually be annoyed when their tools are barred from acting on their behalf.
Platforms will find ways to allow agents in, especially since most agentic traffic targets e-commerce sites. These sites stand to make money if they let AI shop for real people. Shopify is already welcoming the tools by creating separate checkout lanes that bypass pages designed for humans.
The cost for human users
The question remains what happens to the internet for everyone else. We are used to a web where humans remain the intended users, even as sites try to capture our attention. That dynamic could shift toward one optimized for agents, which will eventually outnumber us, leaving the remaining humans with a worse experience.
Websites may become neglected destinations for real people. Those who wish to browse and compare products will face inferior interfaces. Verification systems are already imperfect; imagine being asked constantly to prove you are human, perhaps requiring iris scans. A single sweater might show different prices to different agents. You may never know if your agent’s recommendation is influenced by ads or platform commissions.
This is a stark contrast to the future promised by Google, Anthropic, and OpenAI. They envision a world where AI agents give consumers more power and bargaining leverage. That promise is not outlandish. Agents are capable of genuinely helping people fight bogus medical bills or research cheaper car insurance.
Whether these tools can be used for such purposes depends on the rules set for how they interact with websites. These rules are still being written. Some will benefit everyone, such as restrictions preventing agents from sharing credit card information freely. It is the contested rules, like whether businesses can block agents or stop them from negotiating better deals, that will decide if the web becomes a highway for agents working for us or against us.
