The dream of AI agents comparing prices and booking for you assumes everyone else will step aside for it, but ecommerce profits from opacity, so platforms are mass-blocking the bots.
In 2001, the tech critic Cory Doctorow wrote 'Metacrap', an attack on the Semantic Web: assume first that everyone in the world will set aside their own interests and describe the world in my categories, and then my plan will work.
Two decades on, the same thinking has a new skin, called 'agentic AI'.
When one bot can't finish the job, send a swarm
The original pitch was simple: tell a chatbot to go out onto the open web, compare flights, and book the best one.
Each step turned out too specialised, so it became a swarm of bots, each handling one slice: one finds sites, one reads pages, one compares prices, one pays.
Specialisation raised accuracy at each point but added cost and complexity. Error rates multiply: two low-failure bots wired together fail more often than either alone. Add more and the system gets more fragile.
The semantic fights were never technical
Semantic Web people believed in objectively correct descriptions that everyone would eventually adopt.
Doctorow's counterexample is blunt: some outlets refuse to call trans women women not because they lack a taxonomy, but because they are bigots.
Ecommerce makes money by keeping you confused
Ecommerce sites are hostile to bots on purpose. Surveillance pricing, drip pricing, junk fees, price fixing — nearly all the profit growth at the biggest firms comes from these, and all of them need opacity.
These tactics are facially illegal, but fines are cheaper than compliance, so they continue. Expecting those same firms to embrace a price-comparing bot is like expecting enough words to wake up a superintelligence.
Google itself sells merchants tools to combine its surveillance data with AI to gouge shoppers. So the current wave of ecommerce platforms mass-blocking AI agents is no surprise.
Any plan that begins with 'first, everyone else sets aside their interests' does not deserve to be taken seriously.
Why it matters
Whether AI agents can save shoppers money was never about how smart the model is, but about whether the other side is willing to be seen. As long as the money is made from murky pricing, a better agent just hits a wall built to lean the wrong way.
Curated from high-quality sources, with concise summaries and key takeaways.
