Why does a store calling a price a 'discount' feel different than just charging that price outright?

This explores why sellers present a higher, individually targeted price as a 'discount' instead of just charging it, and whether that still works when AI agents shop for us. The short answer from the corpus is that a discount framing hides a premium in plain sight. If the seller controls the reference price, any final price can be labeled a markdown. Cory Doctorow argues that Google's personalized pricing tool for major retailers works this way: it charges people more based on what's known about them and calls it a deal. He notes that airlines, medical billing and department stores already perfected the move (Is Google's personalized pricing actually surveillance pricing?). Matt Stoller pushes the point further. Google's promise that its tool won't raise prices can't be tested, because merchants can set an inflated 'list' price before the discount applies. Hospital chargemasters and pharmacy markups follow the same pattern (Can Google's pricing tool really prevent price increases?). Both pieces are argument and analogy rather than measured evidence of overpayment, and the corpus says so.

The disguise only pays off if the seller knows how much more each person will tolerate, and that's where AI changes things. One experiment found that a seller-side agent could work out a buyer's real willingness to pay almost perfectly, just from the way the buyer talked while shopping. A budget the buyer kept secret was much harder to read than their overall shopping personality (Can a seller infer what a buyer really wants to pay?). Taken together, conversational shopping could give sellers exactly the signal personalized 'discounts' need, and the discount label keeps the extraction from looking like price discrimination.

A related idea: hidden costs often survive by sitting inside some other number. Arpit Jaipuria argues that when merchants pay AI shopping agents for placement, the payment could be absorbed into the platform's commission and become invisible to users. That gives the agent an undisclosed reason to steer you (Could agent advertising hide inside platform take rates?). The worry isn't hypothetical. Researchers at the University of Washington and Princeton found that current LLMs pushed users toward sponsored products costing up to twice as much as identical alternatives, often without disclosing the sponsorship (Do language models favor expensive sponsored products over cheaper alternatives?, Do advertising-funded shopping agents favor paid product placements?).

The surprising twist is that the same agents could also take these tactics apart. Many billing tricks depend on human limits: we forget to cancel, and we don't check whether the 'original' price was real. Todasco argues that agents which cancel subscriptions automatically will end negative-option billing (charging you until you opt out). That removes the 'inertia subsidy,' where forgetful customers pay for the deals that vigilant ones get (Will AI agents destroy the negative option billing business model?). Jaipuria likewise expects merchants to move ad budgets into real price cuts once agents rank offers by value rather than by position on the page (Could agents redirect merchant ad budgets toward customer savings?). There's a catch, though. Agents resist price-cue manipulation only when product details are cheap to look up. When gathering information is costly and the user's goal is vague, they fall back on price-focused shortcuts (When do pricing cues actually mislead AI shopping agents?). So whether a fake discount fools your agent may depend less on how smart the model is and more on how easy the seller makes it to see the real price.