For years, the internet treated visitors as either people or bots. Now there is a third kind: AI agents that browse and handle tasks on someone’s behalf. They can search for a flight, compare prices or try to dispute a charge—and they can do it far faster than a person.
The trouble is that many websites don’t know how to handle them. Some block AI assistants over security concerns or fears of losing control of sales and advertising. That means a tool designed to save time may hit a wall before it can complete a booking or purchase.
A new way to shop and handle everyday tasks
Use of these agents is growing. According to MIT Technology Review, Meta’s Muse app reached five million downloads in its first month, while web traffic generated by agents tripled between August and September 2026.
The friction is already showing up in everyday services. Some users have run into obstacles when asking assistants to interact with restaurant and retail websites. Amazon has also blocked Muse agents and is in a legal dispute with Perplexity over its assistants’ access to the store. Among other arguments, Amazon says these systems can make it harder to distinguish advertising shown to people from automated activity.
At the same time, some companies are adapting. Shopify, for instance, is opening dedicated checkout options for agents instead of making them navigate pages built for people sitting in front of a screen.
What could change for ordinary users?
The promise is appealing: hand off tedious tasks and use an assistant to compare insurance, find a better rate or dispute an incorrect charge. But the outcome will depend on the rules governing what these programs can do—and how much access websites give them.
If websites are designed increasingly for machines, browsing on your own could become more frustrating. People may also face more checks to prove they are human. And if prices or recommendations vary from one agent to another, customers could find it harder to tell whether they are seeing the best deal or a suggestion shaped by advertising or commissions.
These are possible risks, not a foregone conclusion. Safeguards matter too: an agent, for example, should not share payment details without permission or make a purchase its user never approved. The challenge is to balance security with the ability to use assistants and access services fairly.
The rules will determine whose side AI is on
Agents could help people save time and negotiate better deals. But that benefit is not guaranteed. It will depend on how they interact with businesses and whether their recommendations are transparent.
The central question is simple: will the internet let these assistants work for their users, or will it prioritize the interests of platforms? The answer will affect not only people who already use AI, but how all of us shop, find information and handle tasks online in the years ahead.
