How much would someone have to pay you to give up AI for a month? Stanford put real cash on the table to find out, and the average answer hit $124.50 in 2026. The median? $11.40, about two coffees.
The question Stanford asked
Economists at Stanford’s Digital Economy Lab, led by Erik Brynjolfsson, ran what researchers call a willingness-to-accept experiment. The setup is blunt. They asked a representative sample of US adults: if you had to give up AI for a month, starting tomorrow morning, what’s the smallest payment you’d take?
Survey takers saw randomized price points: $1, $10, $20, $50, $100, $200, $500. Accept one, and you’d lose access to ChatGPT, Gemini, Claude, and Copilot for 30 days. All of it. The answers build a demand curve, and from that curve the researchers pulled out what AI access is actually worth to ordinary people.
It’s the same trick economists use to value clean air or an extra hour of sleep. You can’t read a price tag off a free product. So you make people choose between the product and money, then watch where they flinch. The full paper, What is Generative AI Worth?, is free to read if you want the raw math.
The numbers behind giving up AI for a month
Two survey waves, one in July 2025 (1,491 people) and one in March 2026 (1,908 people), both sampled through Prolific. Here’s what changed in eight months:
| Metric | July 2025 | March 2026 | Change |
|---|---|---|---|
| Average price to give up AI for a month | $98.00 | $124.50 | +27% |
| Median price | $3.40 | $11.40 | +235% |
| US adults using generative AI | 98M | 115M | +21% |
| Total consumer surplus per year | $116B | $172B | +48% |
Look at the median row again. In 2025 it was $3.40. People would hand over a month of AI for the price of a gas station burrito.
By 2026 it climbed to $11.40, and that 235% jump is the fastest-moving number in the whole table. It tells you AI went from “neat toy” to “thing I’d actually miss” for regular users, not just tech workers.
The $172 billion total is the headline, though. That’s the value US adults place on these tools each year, and here’s the kicker: it’s bigger than the revenue the entire US generative AI industry collects. Users capture more of the value than the companies selling the tools. For context, the same authors’ earlier work pegged the surplus near $97 billion in late 2024, so the curve keeps bending upward.
Why the average and the median live in different zip codes
An average of $124.50 sitting next to a median of $11.40 looks like a mistake. It isn’t. It’s the shape of how people actually use AI.
The mean gets hauled up by a minority of power users who’d demand serious money to lose their tools. For them, AI is load-bearing: workflows, client work, half the job. Ask those users to give up AI for a month and they want hundreds. Ask the median user and you get lunch money. That second group opens ChatGPT a few times a week for recipes, emails, and settling arguments at the dinner table.
Which one are you? Quick test: open your chat history. If the last 20 conversations include at least a few work tasks, you’re closer to the mean. If it’s mostly meal plans and gift ideas, congrats, you’re the median, and AI is still optional fun for you.
The predictors the researchers found, in order: how often you use AI, whether you use it at work, and whether you pay for a subscription. Paid users valued access higher than free users. Not shocking, but it’s nice when data agrees with common sense.
What it means for your $20 subscription
Here’s where it gets personal. If heavy users pull more than $124 of value out of these tools monthly, then $20 for ChatGPT Plus or Claude Pro is a steal for them. The math flips for light users: if a month without AI is worth $11 to you, paying $240 a year is hard to defend.
So audit your own usage the way the study did. One honest week of tracking beats any spec sheet. Count the times AI actually saved you real work, not just answered a curiosity question. Then price that.
We did a version of this math earlier this year when Anthropic’s plans came out far cheaper per task than the competition. Same logic applies here: value depends on how much you actually use the thing, not on benchmark scores.
And remember your AI bill isn’t the only one drifting up. Memory prices keep climbing because of the AI boom, so the laptop you’d run local models on costs more too. Free tiers are looking better by the month.
The catch in the Stanford study
Some honest caveats before you quote that $124.50 at your boss.
Willingness-to-accept numbers always run higher than willingness-to-pay. People demand more to give something up than they’d ever pay to get it. Economists have argued about that gap for decades, and it means this study measures loss aversion as much as pure value.
Also, giving up AI for a month means all of it. No ChatGPT at work, no Gemini on your phone, no Copilot in your docs. A total embargo feels scarier than losing one tool, so answers skew high. Someone who’d shrug at losing ChatGPT alone might still demand real money to give up AI for a month across every app they touch.
And respondents knew they’d never actually lose access. It was hypothetical money on hypothetical terms. Real behavior under a real deadline might look different.
None of this sinks the study, by the way. Two waves, big samples, representative populations: it’s the best measurement anyone has done. The honest read is “AI is worth at least this much” rather than “exactly this much.”
The takeaway
If you’d refuse $500 to give up AI for a month, stop agonizing over subscription prices. You’re getting a bargain, and the Stanford data backs you up. If $11 sounds about right, cancel the paid plan, keep the free tiers, and pocket the difference. Either way, you now know what the tools are actually worth, because economists asked, and thousands of people answered with real numbers.