Google's AI pricing plan. From each according to their ability…

Source: Cory Doctorow; Medium (Pluralistic) · 2026-01-21

Google is spending a lot on AI, but what’s not clear is how Google will make a lot from AI. Or, you know, even break even. Given, you know, that businesses are seeing zero return from AI:

The first string: Google’s going to spy on you a lot more, for the same reason Microsoft is spying on all of its users: because they want to supply their AI “agents” with your personal data:

The second piece of news is that Apple has partnered with Google to supply Gemini to all iPhone users:

Apple already charges Google more than $20b/year not to enter the search market; now they’re going to be charging Google billions not to stay out of the AI market, too. Meanwhile, Google will get to spy on Apple customers, just like they spy on their own users. Anyone who says that Apple is ideologically committed to your privacy because they’re real capitalists is a sucker (or a cultist):

But the big revelation is how Google is going to make money with AI: they’re going to sell AI-based “personalized pricing” to “partners,” including “Walmart, Visa, Mastercard, Shopify, Gap, Kroger, Macy’s, Stripe, Home Depot, Lowe’s, American Express, etc”:

Personalized pricing, of course, is the polite euphemism for surveillance pricing, which is when a company spies on you in order to figure out how much they can get away with charging you (or how little they can get away with paying you):

It’s a weird form of cod-Marxism, whose tenet is “From each according to their desperation; to each according to their vulnerability.” Surveillance pricing advocates say that this is “efficient” because they can use surveillance data to offer you discounts, too — like, say you rock up to an airline ticket counter 45 minutes before takeoff and they can use surveillance data to know that you won’t take their last empty seat for $200, but you would fly in it for $100, you could get that seat for cheap.

This is, of course, nonsense. Airlines don’t sell off cheap seats like bakeries discounting their day-olds — they jack up the price of a last-minute journey to farcical heights.

Google also claims that it will only use its surveillance pricing facility to offer discounts, and not to extract premiums. As Stoller points out, there’s a well-developed playbook for making premiums look like discounts, which is easy to see in the health industry. As Stoller says, the list price for an MRI is $8,000, but your insurer gets a $6000 “discount” and actually pays $1970, sticking you with a $30 co-pay. The $8000 is a fake number, and so is the $6000 — the only real price is the $30 you’re paying.

The whole economy is filled with versions of this transparent ruse, from “department stores who routinely mark everything as 80% off” to pharmacy benefit managers:

Google, meanwhile, is touting its new “universal commerce protocol” (UCP), a way for AI “agents” to retrieve prices and product descriptions and make purchases:

Right now, a major hurdle to “agentic AI” is the complexity of navigating websites designed for humans. AI agents just aren’t very reliable when it comes to figuring out which product is which, choosing the correct options, and putting it in a shopping cart, and then paying for it.

Some of that is merely because websites have inconsistent “semantics” — literally things like the “buy” button being called something other than “buy button” in the HTML code. But there’s a far more profound problem with agentic shopping, which is that companies deliberately obfuscate their prices.

This is how junk fees work, and why they’re so destructive. Say you’re a hotel providing your rate-card to an online travel website. You know that travelers are going to search for hotels by city and amenities, and then sort the resulting list by price. If you hide your final price — by surprising the user with a bunch of junk fees at checkout, or, better yet, after they arrive and put their credit-card down at reception — you are going to be at the top of that list. Your hotel will seem like the cheapest, best option.

But of course, it’s not. From Ticketmaster to car rentals, hotels to discount airlines, rental apartments to cellular plans, the real price is withheld until the very last instant, whereupon it shoots up to levels that are absolutely uncompetitive. But because these companies are able to engage in deceptive advertising, they look cheaper.

And of course, crooked offers drive out honest ones. The honest hotel that provides a true rate card, reflecting the all-in price, ends up at the bottom of the price-sorted list, rents no rooms, and goes out of business (or pivots to lying about its prices, too).

Online sellers do not want to expose their true prices to comparison shopping services. They benefit from lying to those services. For decades, technologists have dreamed of building a “semantic web” in which everyone exposes true and accurate machine-readable manifests of their content to facilitate indexing, search and data-mining:

This has failed. It’s failed because lying is often more profitable than telling the truth, and because lying to computers is easier than lying to people, and because once a market is dominated by liars, everyone has to lie, or be pushed out of the market.

Google says that it’s going to get every website in the world to expose accurate rate cards to its chatbots to facilitate agentic AI. Google is also incapable of preventing “search engine optimization” companies from tricking it into showing bullshit at the top of the results for common queries:

Google somehow thinks that the companies that spend millions of dollars trying to trick its crawler won’t also spend millions of dollars trying to trick its chatbot — and they’re providing the internet with a tool to inject lies straight into the chatbot’s input hopper.