AI agents that promise to shop, book flights, or make reservations on your behalf keep hitting a wall: the websites they need to use don't want them there.
The problem is structural. Sites have spent years building anti-bot defenses to stop scrapers and fraud, and those same defenses can't tell the difference between a malicious bot and your agent trying to book a flight. The result is users stuck in the middle, watching their AI assistant get bounced by a CAPTCHA or silently blocked, with no clear way to prove it's acting on their behalf. A new standard is now being proposed to give websites a way to let legitimate agents in without reopening the door to abuse.
This is the access problem eating every consumer AI agent launch this year. It doesn't matter how good the underlying model is at planning a trip or filling a cart if the checkout page refuses to render for a non-human visitor. Whoever controls the gate between agents and the open web effectively controls which agents work at all, which is a lot of leverage to hand to whichever side writes the standard.
Standards like this tend to move only as fast as the biggest sites agree to honor them, and plenty of retailers have no commercial incentive to make scraping-adjacent traffic easier to wave through.