Who Gets Paid for Your Indifference?

In the first piece I argued that AI agents bring back the clerk, and that whoever sets the default for an unbranded request owns the brand. One line in it, about selling the eye-level shelf as an API setting, deserves its own piece, because the nuance is where the money is.

Paying for position is old. Grocery chains began charging manufacturers slotting fees for shelf space in the early 1980s, and the fees grew to cover keeping a product on the shelf and putting it at eye level or on the end of an aisle. Search engines rebuilt the same business online: pay, and your listing sits on top.

But both versions had an escape hatch. In a store, the paid product sits at eye level and the cheaper one sits just below it. You can look down. In search, the paid result carries a small label, and you can scroll. The bribe was visible, and the alternative was one glance away.

An agent gives you one answer. There is no shelf below eye level. There is nothing to scroll.

Sergey Brin and Larry Page saw this coming in 1998, while Google was still a Stanford research project. In an appendix to the paper that introduced their search engine, they called selling the top result for a query more insidious than ordinary advertising, because nobody can tell who deserves to be there. The search engine that tried it, OpenText, caused an uproar and stopped being a serious search engine. They also predicted that subtler bias would be tolerated, precisely because it is hard to see.

A paid default is the bribe you cannot see, attached to the choice you did not make.

Here is the first nuance, and most people will miss it. The paid default is not advertising.

Advertising sells attention: an impression, then a click, then maybe a sale. Each step leaks. The paid default skips the funnel entirely. The brand is not buying the chance to persuade you. It is buying the purchase itself, delivered, every time the household says the category word. That is not marketing. It is a supply contract wearing an ad’s clothes.

The second nuance follows. The default is priced by indifference. The less a household cares about a category, the more its default is worth, because nobody is going to override it. Your milk default may be worth a great deal. Your coffee default may be worth nothing, because you named the coffee years ago and your agent will not budge. Indifference, which has never had a price, gets one, household by household and category by category.

If the highest bidder simply won, the system would eat itself. Search learned this early. Around 2002, Google stopped ranking ads purely by price and began multiplying the bid by how often people actually clicked. In 2005 it formalized that as Quality Score. An ad nobody wanted stopped showing up, no matter what the advertiser paid, because ads people ignore train people to ignore ads.

The agent default will need the same second term, and it will be harsher. The signal is not a click. It is the override: the moment a household says “no, not that one, the other kind.” Every override is a brand name spoken out loud, which is exactly the 1898 reflex the default was supposed to put to sleep. Sell a bad default and you do not just lose the sale. You teach the customer to start naming brands again, and the asset you were selling evaporates.

The third bar is the reconciliation with the first piece. I said DoorDash would launch its own brands, and I said I might be wrong because renting the default to food companies could pay more. The auction says both happen. The house brand bids nothing, but it gives up the best outside bid every time it fills the slot. So it wins a category only where its own margin beats what the brands will pay. Milk and bread, probably. A cereal people half remember from childhood, probably not. The map of house brands and paid defaults gets drawn one category at a time, and it gets redrawn every week.

1. Cost per default becomes a price people quote.

   Brands pay per unit filled, not per thousand impressions or per click. Budgets that sat in marketing move toward sales, because what they buy is a purchase, not a chance at one.

2. Every paid default comes with a receipt.

   Agents will show a line that says why this item, and whether the slot was paid. It will arrive through regulation or through competition, but it will arrive, because it is the only version of the “sponsored” label that works when there is one answer instead of a list.

3. The override rate becomes the most watched number in consumer goods.

   It is the new quality score. A brand with a low override rate pays less to hold the slot. A brand with a high one cannot buy it at any price for long.

4. House brands and paid defaults split the store between them.

   Not one or the other. The platform keeps the categories where its own margin wins and rents out the rest, and the line moves as bids change.

5. Households start selling their own defaults.

   The indifference being auctioned is yours. A personal agent that works for you can put your paper towel slot up for offers and keep the proceeds as a discount. When that happens, the household becomes the shelf.

I keep coming back to one distinction: personal AI versus platform AI, and whether every business lets people bring their own. This is the cleanest illustration of why it matters that I have found. A platform’s agent sells your indifference and keeps the money. Your own agent can sell the same indifference and hand the money to you. Same auction, same brands, same paper towels. The only thing that changes is whom the clerk works for.

I wrote about your AI or theirs before any of this had a price on it, and about an agent without appetite. The default auction is where those ideas stop being philosophy and start showing up on a grocery receipt.

Already true

Grocery shelf space has been sold for decades. Search ranks paid results by price and quality together. DoorDash and Instacart both run advertising businesses that food brands already pay, and their merger announcement names those ad tools as part of the combined company. Agents already fill grocery carts from unbranded requests.

What has to happen

Agents have to handle most routine staples without the household watching each pick. Platforms have to be willing to sell inside the answer rather than beside it. And regulators have to allow a paid default at all, with disclosure rather than a ban.

Where I am probably wrong

The paid default may never become a default. If agents are required, by law or by habit, to ask before using a paid pick (“want to try this brand instead?”), the whole thing collapses back into ordinary advertising with a polite voice. That would be a better world for shoppers and a smaller business for platforms. If by 2030 every paid placement in an agent still arrives as a question rather than a decision, this piece was wrong about the mechanism, even if the first one was right about the clerk.

The roll on your counter works fine. It was always going to. That is why it was worth bidding on. You did not care which one came, and somebody turned that into money at two in the morning.

Your not caring has a price now. The only question is whose pocket it lands in.