Market Commentary

The Ask Economy: how a generation that never learned to search is repricing attention

A cohort raised on chat, not search bars, is resetting the default channel for demand. This note maps who is moving, where the rails are forming, and the cost hidden inside the convenience.

Commissioned by · Rafael Pires · Behavioural Economics

1. The interface that stopped being asked for

A generation is coming of age that never learned to type a keyword into a blank box and sift ten blue links for the answer. It learned to describe an outcome in a sentence and expect delivery — the itinerary booked, the outfit assembled, the essay drafted, the deal closed. Search was a skill; asking is a reflex, and the difference is not cosmetic. It relocates effort. Where search asked the person to browse, evaluate and choose, asking hands that work to whatever sits on the other side of the sentence.

The blank search box is a relic; the blank chat window is the interface — and whoever receives the sentence now does the choosing that browsing used to reserve for the person typing it.

This is not a story about a new app. It is a story about who holds the interpretive step between want and outcome — and that step used to belong to the person doing the wanting.

2. Why the shift is showing up in the numbers now

The scale is no longer speculative. OpenAI has reported that ChatGPT's in-chat commerce feature processes roughly 50 million shopping queries a day, connected to more than a million Shopify merchants — a discovery channel with no homepage and no ad unit in the traditional sense. Google spent 2025 into 2026 folding an answer-first AI Mode directly into Search, moving the default experience from a page of links to a single composed answer. Neither move is a feature update. Each is a redraw of where demand first makes contact with supply.

Once the answer arrives pre-assembled, the ten blue links stop looking like choice and start looking like homework — and homework is exactly the thing this cohort has been trained, by the product itself, to skip.

3. Two skill sets, two defaults

Searching and asking are not the same competence wearing different clothes. Searching rewards someone who can decompose a need into keywords, hold several open tabs of imperfect information, and weigh them against each other — a slow, comparison-heavy default that behavioural economics would recognise as effortful System 2 work. Asking rewards someone who can state intent precisely and trust the assembly to a system whose evaluation logic is invisible to them. That is a different default entirely, and it changes who the interface favours by design — a case of choice architecture doing the choosing before the person arrives.

The prompt-native default is not lazier — it is a different allocation of cognitive load, and load that leaves the asker's hands does not disappear. It relocates to whichever system now curates the answer, along with the leverage that curation carries.

4. Who is laying the rails

Commerce built entirely inside a conversation needs infrastructure the open web never required — a way for an agent to read a catalog, compare it and transact, without a human ever loading a storefront. That infrastructure arrived fast. OpenAI and Stripe launched the Agentic Commerce Protocol in September 2025; Google's Universal Commerce Protocol and Anthropic's Model Context Protocol followed within months, each staking out a piece of how an agent is meant to discover and buy on a person's behalf.

None of these bodies regulate in the legal sense — there is no elected authority here yet. But each is writing the de facto rules for which brands an agent can even perceive, and a brand a machine cannot read has quietly become a brand a customer never sees.

5. Not one generation — three postures

The prompt-native cohort is not a single bloc moving in lockstep. It splits along how much of the decision people are willing to hand over, and that split tracks closely with loss aversion — the well-documented tendency to weigh a potential loss of control more heavily than an equivalent gain in convenience. Some readily trade the loss for the gain; others will not, no matter how good the assembly gets.

The gap between these three is not age. It is exposure — how much control each has already handed over, and how reversible that felt the first time it happened.

6. The craft the asking quietly retires

There is a cost inside the convenience, and it is not merely nostalgic. The comparison shopper who reads five reviews before buying is doing more than finding a product — they are building a private model of the market, one purchase at a time. Ask instead of browse, and that private model stops accumulating. The skill does not vanish overnight; it atrophies the way any unused capacity does, one small default at a time.

A generation fluent in asking and unpracticed in evaluating is, by construction, more dependent on whoever designs the interface between the two — a dependency that compounds the way hyperbolic discounting compounds: the convenience is felt today, the narrowed option set is felt only much later, and by then it looks like the market itself, not a design choice.

7. Where the exposure concentrates

Handing the interpretive step to an agent does not remove risk from the transaction — it relocates and concentrates it. Four exposures stand out as the rails get built: who controls the default answer, who audits it, who bears a bad one, and how reversible any of that turns out to be once habits set.

None of these four is hypothetical — each is already visible in how fast OpenAI, Google and Anthropic moved to own a piece of the answer layer before anyone else could.

8. What this resets, point by point

Pull the threads together and a short list of resets emerges — not predictions, but the structural consequences already visible in how the rails are being laid. Each one follows directly from the shift already under way, not from a guess about where it might go.

Each reset compounds the others: fewer people practising evaluation, more of the market's attention routed through a small number of assemblers, and a checkout still being fought over by the parties best placed to win that fight.

9. What stays open

The shift is clear enough to map; its endpoint is not. What remains unresolved is less about whether asking replaces searching — that is largely settled — and more about what happens to judgement once it is rarely exercised, and who is accountable when an assembled answer turns out to be wrong.

That is the question this shift leaves open, and it will not be answered by better assembly. It will be answered by whoever first decides to measure what the asking generation stopped practising.

Grounded in Folka’s corpus: The Prompt Native · Post-Language UX · Agentic Commerce

How we read this

This read draws on named platform moves (OpenAI, Google, Stripe, Anthropic, Upwork) and their reported figures for 2025–26, plus an editorial reading of protocol launches and labour-market data. Figures are attributed to the party that reported them, not independently audited. The one number that would most change this read: a measured comparison of consideration-set size in an agent-mediated purchase versus a self-directed search, across a representative sample of the prompt-native cohort.