People say losing AI for a month is becoming more expensive
Generative AI may be creating far more consumer value than subscription revenue alone suggests. A Stanford Digital Economy Lab study asked representative samples of US adults how much compensation they would require to give up tools such as ChatGPT, Gemini, Claude or Copilot for one month. The Stanford Digital Economy Lab study found that the mean answer rose from $98 in 2025 to $124.50 in 2026.
The median was much lower, rising from $3.40 to $11.40. That gap between mean and median is important because it shows that a smaller group of highly attached users pulls the average upwards. The researchers also estimate that the adult user base grew from 98 million to 115 million across the two survey waves.
The headline number is not a market price
The $124.50 figure is not what the typical person pays for an AI service, and it should not be treated as a forecast of subscription prices. It comes from willingness-to-accept experiments, a standard economic method that asks what compensation would make someone surrender access to a service for a period of time. The point is to estimate welfare rather than revenue.
Using those choices and adoption estimates, the researchers calculate that aggregate consumer surplus rose from $116 billion to $172 billion. They argue that this exceeds estimated US generative-AI revenue, implying that users may be capturing much of the economic value rather than paying it all to providers.
Frequent users value the tools most
Usage frequency was the strongest predictor of valuation, followed by workplace use and whether someone already paid for a subscription. That is intuitive: a tool woven into daily tasks is harder to surrender than one used occasionally. It also helps explain why headline surveys about AI adoption can hide very different kinds of users.
LiveAIWire recently reported that half of US shoppers are using AI for shopping, while other research has shown that workers are spending their own money on AI. Those stories point to the same emerging divide. Some people are still experimenting with AI, while others increasingly treat it as infrastructure for work, shopping or everyday decisions.
Saved time at home may be part of the hidden value
Another Stanford study offers a clue about where some of that value comes from. Research tracking 200,000 US households found that generative AI can reduce time spent on digital chores at home, and that users often spend the saved time on leisure rather than formal skill development. Stanford research on AI and digital chores This matters because traditional productivity statistics are better at counting paid output than the value of making an annoying personal task disappear.
An AI tool that helps compare options, draft a complaint, plan a trip or summarize a document can feel valuable even when no money changes hands. The benefit is partly convenience, partly reduced cognitive effort and partly the ability to attempt tasks that someone might otherwise postpone.
Consumer value can rise even when trust remains uneven
High stated value does not mean users trust every answer. People can find a service useful while still checking important claims, avoiding sensitive tasks or disagreeing with how the product handles data. This is particularly relevant as AI moves deeper into commercial decisions. LiveAIWire has covered how AI persuasion can influence online shopping, which makes the relationship between convenience and influence worth watching.
The welfare estimate also does not tell us how value is distributed. Heavy professional users may gain much more than occasional users, and access can depend on income, devices, language support and digital confidence. Those differences matter if AI increasingly becomes part of how people navigate work and services.
The business question is how much of the surplus providers can capture
If the Stanford estimate is directionally right, AI companies face an unusual commercial opportunity. Users may already receive more value than providers collect in revenue, leaving room for higher prices, new premium tiers or paid specialist functions. At the same time, competition can keep prices low if consumers can switch between broadly comparable assistants.
That tension helps explain why providers are racing to add memory, agents, voice, file handling and integrations. The more an assistant becomes embedded in a person’s routines, the larger the switching cost can become. Consumer surplus may therefore be both a sign of genuine usefulness and a map of where the next wave of monetisation will be attempted.
A better way to think about AI adoption
The study offers a useful alternative to counting downloads, subscriptions or prompts. It asks what people would lose if the technology disappeared. That is closer to the economic meaning of a general-purpose digital tool, especially while business models are still evolving.
The finding should still be treated as an estimate from survey-based choice experiments, not a cash value sitting in household accounts. But the direction is striking: between 2025 and 2026, more adults used generative AI and many said it had become harder to give up. That is a strong signal that the technology is moving from novelty towards everyday utility.
The important point is that the result should not be read as a universal forecast. The evidence describes a particular setting, population or technical system, and the strongest conclusion is about what happened under those conditions. That distinction matters because AI stories often travel faster than their limitations. A useful reading keeps the headline finding intact while separating it from broader claims that the source did not test.
There is also a practical reason to watch this development. AI products are moving from isolated demonstrations into ordinary workflows, which means small design choices can have large effects once they are repeated across millions of interactions. The next phase will be less about whether a system can perform a task at all and more about reliability, human control, cost, access and what happens when the technology meets messy real-world behaviour.
For readers, the safest takeaway is neither enthusiasm nor dismissal. The evidence is strongest when it is used to identify a real change and weakest when it is stretched into a prediction about everyone. What matters next is replication, wider deployment data and whether the same effect survives outside the original conditions. Those are the tests that turn an interesting result into something people can reasonably use.
The wider pattern across AI is becoming clearer: capability alone is not the whole story. Context determines whether a tool helps, distracts, saves time, shifts power or simply moves effort somewhere else. That is why seemingly narrow findings can matter. They expose the conditions under which AI changes behaviour, and those conditions are often more useful than a single benchmark score or product claim.
The important point is that the result should not be read as a universal forecast. The evidence describes a particular setting, population or technical system, and the strongest conclusion is about what happened under those conditions. That distinction matters because AI stories often travel faster than their limitations. A useful reading keeps the headline finding intact while separating it from broader claims that the source did not test.
There is also a practical reason to watch this development. AI products are moving from isolated demonstrations into ordinary workflows, which means small design choices can have large effects once they are repeated across millions of interactions. The next phase will be less about whether a system can perform a task at all and more about reliability, human control, cost, access and what happens when the technology meets messy real-world behaviour.
For readers, the safest takeaway is neither enthusiasm nor dismissal. The evidence is strongest when it is used to identify a real change and weakest when it is stretched into a prediction about everyone. What matters next is replication, wider deployment data and whether the same effect survives outside the original conditions. Those are the tests that turn an interesting result into something people can reasonably use.
The wider pattern across AI is becoming clearer: capability alone is not the whole story. Context determines whether a tool helps, distracts, saves time, shifts power or simply moves effort somewhere else. That is why seemingly narrow findings can matter. They expose the conditions under which AI changes behaviour, and those conditions are often more useful than a single benchmark score or product claim.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.