← A Thousand Words
As early as 1971, the American Nobel laureate Herbert Simon identified a fundamental problem of the digital world: an abundance of information creates a scarcity of attention. The platform economy turned that insight into a business model. Whoever aggregates attention can sell it in small units – as search ads, feed posts, video spots or display banners. Reach, frequency and available “inventory” set the pace.
Google and Meta are the most successful machines of this attention economy. In 2025, Alphabet generated more than 70 percent of its revenue from online advertising. At Meta, advertising accounted for $196 billion of $201 billion in revenue, or roughly 98 percent. The purpose of these systems is not merely to answer a question, but also to create as many relevant advertising opportunities as possible: on search results pages, in feeds, videos, apps and partner offerings. The more usable attention there is, the larger the market.
Generative AI is changing the interface – and perhaps the economic unit behind it. A traditional search engine gives users a list of possibilities. Someone talking to an AI system expects a synthesis, a single answer. A traveller does not want ten links to travel sites, but a suitable itinerary. An employee does not want fifty documents, but a decision-ready brief. A researcher does not want the largest possible number of hits, but a reliable answer with sources, uncertainties and open questions.
This is where the intention economy begins. Its scarce resource is no longer my attention alone, but a precise understanding of what I want to achieve. A search query is only a signal. Behind it lies the intention: to compare, decide, learn, buy, program, diagnose or avoid a risk. AI systems can ask follow-up questions, absorb context, plan intermediate steps and use tools. When everything works as it should, this brings them progressively closer to the desired result.
Until now, what mattered was how long we looked. In future, what will matter is how precisely a system understands what we want.
The promise of an “exact answer”, however, requires precision. For many tasks there is no single objectively correct solution. A good assistant, whether digital or human, must recognise trade-offs: Should a trip be especially affordable, climate-friendly or convenient? In an analysis, is speed more important than completeness? The intention economy is not mind-reading. It is a dialogue in which a system must “understand” and make visible the objective, constraints and uncertainty before it acts.
For the platform economy, this marks a paradigmatic shift. In the attention economy, the winner is whoever creates and monetises the greatest number of opportunities for contact. In the intention economy, the winner is whoever delivers the most useful result with the least friction. The key metrics would no longer be impressions, clicks or time spent, but tasks solved: Was the answer correct? Could it be verified? Did it save time? Did it actually lead to a better decision?
This also shifts the value chain. Under the old logic, a search leads to several page views, each of which creates new advertising inventory. A good AI answer can compress that journey into a single dialogue. This is efficient for users; for publishers, comparison sites and retailers, it is likely to erode reach, perhaps relevance, and certainly monetisation in the traditional sense. The contest of the future will no longer be about eyeballs and simple reach, but about whose data is cited, whose offer is recommended and whose transaction is completed directly inside the assistant.
This does not mean that advertising will disappear. AI answers, too, can contain recommendations, sponsored options or paid priorities. Google and Meta in particular have enormous advantages: data, distribution, advertisers and established auction systems. The shift could therefore become the next stage of their business model. Alphabet itself describes how large language models enable more natural questions and better results; the same technology can also make a user’s intention far more precisely marketable.
This is the dangerous side of the term. In the Harvard Data Science Review, Yaqub Chaudhary and Jonnie Penn warn of an intention economy in which not only attention, but signals of human intentions are collected, predicted, influenced and sold to interested parties. The advertising auction could become an auction of intentions. An assistant might then do more than decide which answer suits me; it might also try, imperceptibly, to change what I am supposed to want in the first place.
“Intention” is therefore not automatically a friendlier alternative to “attention”. A sound, sustainable intention economy must be designed differently. It needs a clear separation between answers and advertising. Paid influence must be visible. Sources, assumptions and uncertainty must remain verifiable. Users must be able to determine which context is stored. And a system should optimise for the user’s stated goal – not for the margin of an invisible third party.
For companies offering content as well as products, the strategic question shifts accordingly. Until now it has been: How can we reach our target group as comprehensively and efficiently as possible? In the intention economy it could become: At what moment can we demonstrably fulfil a specific intention better than legacy systems? This requires less campaign logic and more content and product logic. Brands do not need to appear everywhere. They need to be trustworthy, machine-readable and verifiable wherever an AI is preparing a decision.
We will not leave the attention economy overnight. But AI is changing its centre of gravity. The internet of advertising spaces is increasingly being overlaid by a layer of assistants that select, condense and act. Anyone who owns only inventory risks becoming interchangeable. Anyone who understands intention and fulfils it loyally in the user’s interest can win the more valuable relationship. The decisive question, then, is not whether AI recognises our intentions. It is to whom the AI is accountable when it acts on them.