Wave 10: Benefits

The following report summarizes responses from 180 experts, in addition to 53 superforecasters and 595 members of the public collected between Jun 16, 2026 and Jul 07, 2026. Within expert respondents, 32 computer scientists, 37 industry professionals, 42 economists, and 69 research staff at policy think tanks participated.

Our wider website contains more information about LEAP, our Panel, and our Methodology, as well as reports from other waves.

Scenarios

Throughout this wave's survey, we ask respondents to consider three scenarios for AI progress by 2030 that we call "slow", "moderate", and "rapid".

In Wave 1 of LEAP, we asked respondents to forecast the percentage of LEAP panelists who would select each of these scenarios as best representing reality in 2030. In that wave, the mean aggregate forecasts by group (and the standard deviation of responses) are displayed in the table below (source).

Insights

1. If AI progress is rapid, experts forecast 500,000 annual deaths from tuberculosis, HIV/AIDS, and malaria in 2050, compared with 1.65 million deaths if AI progress is slow.
   We asked how many people will die globally each year from tuberculosis, malaria, and HIV/AIDS, three prominent diseases currently considered non-eradicable that caused about 2.5 million deaths in 2023. If AI capabilities progress slowly, the median expert forecasts approximately 1.65 million combined deaths by 2050; if AI progress is rapid, that falls to 500,000, a reduction of about 1.1 million deaths a year, or almost 70%. The same pattern holds across groups but with smaller gaps: the median public forecast falls from 1.7 million to 750,000, and the median superforecaster from 1.5 million to about 1 million. The effect is not limited to the rapid AI progress: moderate progress still brings the expert median down to 1.0 million. Unconditionally, all three groups forecast roughly 1.2 to 1.5 million deaths by 2050.

   In their written rationales, many forecasters were skeptical that AI-discovered treatments would be a key factor in reducing deaths from non-eradicable diseases in the near term. They cited inadequate health systems and poverty as the reasons high death levels persist. On the other hand, several forecasters pointed to the potential for AI to alleviate poverty and, by extension, deaths from disease in the regions where these deaths are highest.

2. Experts forecast that rapid AI progress would increase U.S. life expectancy by almost 16 years in 2100, compared with just six years if AI progress is slow.
   US period life expectancy rose 13.2 years (from 66.1 to 79.3) in the 77 years between 1946 and 2023. In a world with slow AI progress, experts forecast life expectancy to be 85 years in 2100 (77 years later), or an increase of only 5.7 years. In a world with rapid AI progress, experts forecast life expectancy to be 95 in 2100, or an increase of 15.7 years. The 10 year spread between scenarios suggests that experts predict that in a rapid-AI world U.S. life expectancy will increase by a full decade compared to a slow-AI world. Even in the moderate AI progress scenario, experts predict life expectancy in 2100 to be about 5 years above the slow AI progress scenario, itself equivalent to roughly 30 years of historical progress.

   Although almost all forecasters thought period life expectancy would rise between 2050 and 2100, many expressed through their written rationales that the gains would be unevenly distributed across the income spectrum. Optimists, on the other hand, thought that AI would address the top sources of mortality in a way that was broadly accessible to all Americans. Others pointed to the life expectancy in countries such as Monaco and San Marino, arguing that the U.S. could catch up to the existing frontier.

3. Under rapid AI progress, experts forecast that annualized GDP growth between 2045 and 2050 will be 3 percentage points higher in advanced economies and 2 percentage points higher in emerging and developing economies than under slow AI progress.
   We asked forecasters for annualized real GDP growth under different AI progress scenarios in two parts of the world economy: advanced economies, and emerging and developing economies. Their forecasts suggest AI accelerates growth more in advanced economies than in developing ones, even though developing economies are still expected to grow faster in absolute terms. Advanced economies have grown about 2% a year for two decades (1.9% in 2025), and forecasters see AI as capable of transforming that: the median expert expects 2050 growth to reach 5% under rapid AI progress, more than double the recent trend and a pace advanced economies have essentially never sustained. Superforecasters are more cautious at 3%. This nearly matches what the panel forecast in Wave 6, where experts put US real GDP growth at 5.0% by 2050 under rapid progress and superforecasters at 3.4%. Emerging and developing economies look different: they grew 4.4% in 2025 and averaged nearly 6% in the 2000s, and the median expert predicts 2050 growth to reach just 6% under rapid progress, meaningful but within the historical range for these economies. Under the unconditional and moderate scenarios, forecasts for both regions stay near recent norms, around 2 to 3% for advanced and 4 to 5% for emerging economies.

   In their written rationales, some respondents argued that AI would destroy the traditional development paths of emerging economies, preventing workers from making the transition from low-productivity to high-productivity jobs. They also argued that robotics and AI could bring manufacturing back to advanced economies, to the disadvantage of emerging economies.

4. Experts and superforecasters expect US users to value a month of generative AI access at about $450 in 2026, rising to $1,500 by 2030 if AI progresses rapidly.
   We asked for the mean willingness to accept (WTA), the payment a typical US adult chatbot user would require to give up all generative AI tools for a month. Unconditionally, the median expert and superforecaster both put the 2026 value at about $450 a month, roughly $5,400 a year, while the median public forecast is lower at about $400 with a quarter of the public predicting a willingness to accept under $150. For comparison, willingness-to-accept experiments by Brynjolfsson, Collis, and Eggers (2019) valued search engines at about $17,530 a year, email at $8,414, and digital maps at $3,648¹, placing predicted 2026 generative-AI value in the range of established internet services, below search but near email and maps. The predicted value increases with time and AI progress: by 2030 the unconditional median forecast reaches $750 for experts and $875 for superforecasters, and under rapid progress both reach $1,500 a month, or roughly $18,000 per year. The public expects far more modest growth: about $408 by 2030 unconditionally, and only about $500 even under rapid progress.

5. Even if AI capabilities progress rapidly, forecasters expect global happiness to stay nearly flat.
   Forecasters do not expect people to report being much happier under rapid progress in AI capabilities. The global population-weighted score on the Gallup World Poll's 0-to-10 Cantril ladder, a measure of how people evaluate their lives overall, was 5.44 in 2025 and has barely moved in over a decade. Forecasters expect that near-flat path to continue regardless of how fast AI advances. Across every group and scenario, median forecasts stay within a narrow band, rising only to about 5.5 to 5.8 by 2040, with the single exception of the public under rapid progress at 6.1. The public is the most apt to see improvement, with a 2040 median prediction of 6.1 under rapid AI progress. Experts stay close to flat, moving just a tenth of a point (5.5 to 5.6) between the slow and rapid scenarios. Experts are also split: a quarter put the 2040 rapid-progress score at 5.3 or below, beneath today's level, while a quarter put it at 6.0 or above. Even as forecasters expect AI to deliver large, measurable gains in health and economic benefit, they doubt these gains will translate into how satisfied people feel with their lives. Respondents frequently point to the stability of the historical trend line, noting that the 2008 financial crisis, COVID-19 pandemic, and recent inflation had done little to shift self-evaluations of life satisfaction globally. Another possible explanation of this is that while respondents expect gains from rapid AI progress (as documented in this wave), they also expect harms from AI on democracy, catastrophe, inequality, and job loss as documented in Wave 9. Some also noted that the benefits of AI would not be broadly shared, and that AI could cause a “crisis in human purpose.”

Questions

• Deaths from Currently Non-Eradicable Diseases: How many deaths will occur globally from tuberculosis, malaria, and HIV/AIDS combined during the following years and conditional on the following scenarios? ⬇️

• US Life Expectancy: What will the period life expectancy in the U.S. be in the following years and conditional on the following scenarios? ⬇️

• Global Economic Growth: What will be the annualized change, in percent, in the global real gross domestic product (GDP) for the following groups and conditional on the following scenarios between the following years? ⬇️

• Consumer Value of Generative AI: Consider US adults who use generative AI chatbot tools in a given year. If they were offered a cash payment to give up access to all such tools for one month, what would the mean payment they would require — known as their willingness to accept (WTA) — be, in 2026 USD? ⬇️

• Self-Reported Happiness: What will be the global average score (0–10) on the Gallup World Poll Cantril ladder — the population-weighted mean across all surveyed countries, as published in Our World in Data’s World series — in the following years and conditional on the following scenarios? ⬇️

For full question details and resolution criteria, see below.

Results

In this section, we present each question, and summarize the forecasts made and the reasoning underlying those forecasts. More concretely, we present background material, historical baselines, and resolution criteria; graphs, results summaries, results tables; as well as rationale analyses and rationale examples. We analyse these rationales alongside predictions to provide significantly more context on why experts believe what they believe, and the drivers of disagreement, than the forecasts alone.

Deaths from Currently Non-Eradicable Diseases

Question. How many deaths will occur globally from tuberculosis, malaria, and HIV/AIDS combined during the following years and conditional on the following scenarios?

Results. All three groups expect combined annual deaths from tuberculosis, malaria, and HIV/AIDS to decline from its 2023 total of 2.5 million and all three groups expect that decline to be faster under the rapid AI progress scenario. Unconditionally, the median forecast is roughly 2.1 to 2.2 million deaths in 2030 (experts 2.19M, public 2.14M, superforecasters 2.20M), dropping to 1.2 to 1.5 million by 2050 (experts 1.20M, public 1.47M, superforecasters 1.30M). Under the rapid AI progress scenario, experts expect deaths to fall to 500,000, superforecasters to 1.0 million, and the public to 750,000 by 2050. However, there is disagreement within these groups. For example, 25% of experts expect there to be fewer than 150,000 and 25% expect more than 1.0 million.

Rationale analysis

• Poverty as the primary driver: Most forecasters express pessimism that AI-discovered treatments will be a key factor in reducing deaths from these maladies in the near term, citing inadequate health systems in the developing world and poverty as the real reasons high death levels persist: “We have near-cures for all of these now. The problem is health delivery, not medical capability.”; “The core of the problem is poverty-driven barriers to healthcare and prevention, and AI is unlikely to solve these issues.” Several optimists, however, point to the potential for AI to alleviate poverty, and by extension death from disease, in the regions most afflicted: “AI makes the developed world wealthier, and (even if AI does not result in new treatments), more resources would probably be spent on frontline treatments and prevention”; “The decline of malaria and TB is driven less by new drugs/vaccines than by rising income and the associated improvements in public health, sanitation and nutrition. These diseases are essentially a function of development.”
• The historical trajectory: Optimists frequently anchor on the long downward trend and expect AI to bend it further, noting that deaths “have continued to fall since 2000, even without AI interventions.” Pessimists tend to read the same trend as decelerating and unlikely to persist, with many citing “the growing challenge of drug-resistant tuberculosis” and others pointing to climate change’s potential to “expand the range of malaria, causing it to spread significantly.” Several pessimists also note that deaths from these maladies are predominantly in regions experiencing robust population growth. One writes, “Africa’s population is expected to increase by nearly 1 billion people by 2050 [and] the pool of potential victims is highly relevant to the total number of deaths.”
• The 2025 aid shock: For many pessimists, the recent cuts to global aid is the dominant factor, particularly when considering the 2030 horizon: “The thing that actually moves my 2030 number isn’t AI, it’s the 2025 foreign-aid shock. The Lancet modeling on the PEPFAR [funding decline] runs anywhere from 0.77M to 2.93M additional HIV deaths through 2030”; “By 2030, any possible impact of AI...will still be vastly outweighed by the curtailing of US aid programmes globally in 2025.”; “IHME projected for 2025 that child mortality under 5 had reached an estimated 4.8 million deaths in 2025, an increase of 200,000 preventable deaths compared with 2024, after child mortality had decreased every other year since 2000.” A few optimists treat the shock as likely to be transient, with one arguing that funding crises "tend to resolve (budget reversals, other donors, domestic substitution),” and another arguing that it is likely to be outweighed by “GDP improvements in hotspot regions.”
• Time horizons: Many pessimists express near certainty that little will change for the better by 2030, citing the long timelines typically needed to develop and release drugs and the inability for AI to change this: “Even if AI can create custom drugs by 2029, there's no time for testing, manufacturing, dissemination.”; “Clinical trials cannot be done on GPUs and cannot be sped up by bigger clusters.” Several optimists, however, argue that AI could act via domains in which much faster progress is possible, such as “mapping and containing outbreaks, deploying doctors and medical professionals, [and] improving logistics of treatment deliveries.” Others conclude that while AI-discovered treatments are unlikely to have much effect on the 2030 outcome, on a longer timeline, “AI will accelerate the medical advances that will address these diseases, including prevention and treatment.” This argument is frequently cited by forecasters considering 2050 under a rapid progress scenario, with one optimist predicting, “AI-discovered medicines and cures will plausibly cut combined deaths by 90% or more.”
• AI’s impact on inequality: A recurring theme among pessimists is that health gains are unlikely given that gains from AI progress are unlikely to be evenly distributed, with one writing that gains are “more likely going to wealthy individuals in wealthy countries,” and another predicting that “there will be a global underclass left behind...for decades to come.” Several pessimists expect this dynamic to materialize even in a rapid progress world, with one noting that “a rapid progress scenario refers to the capabilities of AI, not to the funding priorities of global health needs.” A few optimists point to the potential for philanthropists, newly wealthy as a result of AI, to contribute more to global health initiatives: “There are a bunch of people who are likely to get rich from worlds with rapid progress who care a lot about global health interventions...I expect that there will just be much more money going towards these causes.”

US Life Expectancy

Question. What will the period life expectancy in the U.S. be in the following years and conditional on the following scenarios?

Results. Period life expectancy in the U.S. was about 79 years in 2023, and all three groups expect steady gains that widen under the rapid AI progress scenario. The unconditional median climbs to 81 to 83 years by 2050 (experts 83, public 81, superforecasters 82) and 85 to 88 years by 2100 (experts 88, superforecasters 87, public 85). Under the rapid AI progress scenario, the 2100 median rises to 95 years for experts and roughly 90 for the public and superforecasters. However, there is disagreement, especially in the long-run tails. For example, under the rapid AI progress scenario in 2100, 25% of experts forecast below 88 years and 25% forecast above 105.

Rationale analysis

• Distribution of the gains: Almost all forecasters foresee period life expectancy rising over the two time horizons, but many are pessimistic that the gains will be either exceptional or widely shared: “AI-accelerated medicine mostly raises a ceiling the rich already enjoy, and if AI widens inequality it could drag the floor further.”; “My optimism for AI benefits being spread across the economic spectrum is fairly low, reflected in my predictions and higher upper tails.” Optimists tend to expect gains will be widely shared: “AI is directly addressing the top sources of mortality (and others e.g. transport accidents) in ways that are likely to be broadly accessible, and so I don’t think economic inequality will hold it back.”
• Human behavior as the bottleneck: Optimists typically expect that AI, particularly in a rapid scenario, will lead to “accelerating the development of new and better treatments” and that this will extend lifespans. Many pessimists don’t view access to treatments as the key consideration: “It’s lifestyle choices, and it’s a lot to ask of technology to salvage the livers, kidneys, and pancreata of those pushing the curve down.”; “The things actually killing Americans early are behavioral and social, not waiting on a lab. Obesity, deaths of despair, overdoses, gun deaths.”
• The trendline: Optimists and pessimists typically anchor their predictions on an extrapolation of the historical trendline but differ in what additional factors—beyond AI—they think will bend the trendline up or down. Optimists often point to the potential for the U.S. to catch up to other developed countries, along with an existing-drug tailwind: “Switzerland is already at [84]. [The] US just needs to get closer to the frontier!”; “OWID indicates that some of the leading countries have greatly exceeded the US in life expectancy…Monaco and San Marino: 86.4 years and 85.7 yrs respectively.”; “The U.S. is already on a path toward improvement because of GLP-1s and related metabolic drugs.” Pessimists tend to view the trend as having already flattened—“[The] U.S. has stagnated for the better part of two decades while peer countries kept climbing”—and likely to stay flat given U.S. lifestyle choices and “constrained…healthcare access.”
• Accounting for catastrophic and existential risk: Some pessimists use their 10th percentile forecasts (elicited for the unconditional and rapid scenarios) to account for catastrophic and existential risk, with a handful even entering zero as a value: “There's more than a 10% chance that humanity goes extinct.” Optimists tend to judge such risk as unlikely—“I find it far more likely that AI will provide far more beneficial outcomes than pose existential risk to mankind that could cause some type of widespread health crisis, war or other activity that would significantly lower average life expectancies”—or, in a few cases, to exclude it on methodological grounds: “Not really pricing in p(doom) since extinction resolves ambiguously.”
• A biological ceiling: Pessimists often see a hard limit to potential lifespans: “Extending life expectancy from 70 to 80 was easier than extending to 90 will be, AI or no AI.”; “~90 years is really pushing the limits of [the] normal human body.” A few optimists express skepticism such a ceiling exists: “Aging itself might be slowed a lot, or maybe even solved. If that happens, some people could live around 200 or longer, and the average could rise much more than most people expect.”

Global Economic Growth

Question. What will be the annualized change, in percent, in the global real gross domestic product (GDP) for the following groups and conditional on the following scenarios between the following years?

Results. Advanced economies grew about 1.9% in 2025, and forecasters expect growth to be higher the faster AI progresses. Experts predict 2050 growth of 2% under slow progress, 3% under moderate progress, and 5% under the rapid AI progress scenario, and 2.7% unconditionally. The public tracks experts closely, but superforecasters stay lower throughout, reaching just 3% even under rapid progress. The groups therefore largely agree for the slower scenarios and separate only as assumed progress quickens. Within-group spread stays modest at the low end — for slow progress in 2050, half of experts predict growth between 1.4% and 3.2%.

Emerging and developing economies grew about 4.4% in 2025, and forecasters expect them to keep growing faster than advanced economies by roughly one to two and a half percentage points, with the gap narrowest under rapid progress. Unconditional medians stay near 4 to 4.5% through 2050 across all groups, and even the rapid AI progress scenario raises the 2050 median only to 6% for experts and the public, with superforecasters lower at 4.7%. The groups mostly agree on these medians, but individual forecasts vary more here than on any other Wave 10 question. Unconditionally in 2050, a quarter of experts forecast below 3.7% and a quarter above 5.7%, and that gap grows wider for faster progress.

Rationale analysis

• The 2030 time horizon: The majority of respondents point to the fact that the first resolution window (between the beginning of 2025 and the beginning of 2030) is already partially settled: “Growth for 2025-2027 is essentially locked in…so a five-year annualized number can only move as far as 2028-2029 can drag it.” The high-growth and low-growth poles referred to below, therefore, mainly apply to 2045–2050.
• Pace of change: High-growth respondents often point to AI as a general purpose technology that is likely to “massively increase economic growth” by 2050. They frequently cite “automation, cheaper services, and faster R&D, especially in rapid progress,” and point to the potential for “larger gains through AI-driven productivity,” particularly if a rapid-progress scenario materializes. One writes, “with rapid progress, we can't even imagine the impact it will have, probably bigger than electrification and industrialization.” Low-growth respondents tend to counter that “productivity growth is not about capabilities. It is about diffusion and adoption, and both are slow.” Several note that growth and productivity gains have been limited to date: “AI has been around for 3 years and the long talked about productivity boost hasn’t happened yet.”
• Demographics: Low-growth respondents often argue that low fertility rates and shrinking workforces will limit growth: “GDP growth is the sum of labor-force growth and productivity growth. The first…is now turning negative across the advanced world and much of the emerging world…That is why 2% is a large number, not a small one.” Some high-growth respondents acknowledge that argument, but reason that AI can substitute for scarce labor: “The world will probably need moderate AI progress to overcome basic demographic drag.”; “AI humanoids may be more important for economic growth by 2050.”
• Emerging Markets and Developing Economies (EMDEs): High-growth respondents typically expect EMDEs to be able to grow faster because they start from a lower base and have the potential to leapfrog legacy systems, with AI cutting the cost of doing so: “I put emerging markets higher because they can still catch up, build more, and copy what already works from richer countries.” Low-growth respondents often argue AI destroys the traditional path EMDEs have used to develop their economies: “AI progress will likely end one of the key drivers of growth of emerging economies: catch-up growth is currently enabled by the fact that lots of people are initially stuck with low-productivity jobs, but you can quickly transfer them to high-productivity jobs after technology transfer. If future jobs are made by AI, this no longer applies.”; “A combination of robotics and AI technology will directly undermine [EMDEs] current advantage and bring manufacturing back to advanced economies that no longer require cheap human labor.”
• Explosive growth: Many high-growth respondents take historically unprecedented growth seriously enough to include in their 90% forecasts, particularly under a rapid-progress scenario: “I do think there is a >10% chance of unprecedented growth—30% seems plausible”; “If our AIs have robots building robot factories, they are limited by very little on the human scales…”; “Growth probably drastically accelerates after the singularity.” Low-growth respondents typically express skepticism that explosive growth is even possible, much less likely: “GDP growth is very ‘sticky’. The “economy” is a large aggregate, and its rate of growth is unlikely to change much even with the proliferation of AI.”; “AI may be our latest and greatest innovation, but such innovation has been the norm and already factored into expectations.”
• Economic shocks: Low-growth respondents frequently invoke the potential for things like war, tariffs, general recessions, and the popping of an AI bubble to counteract AI-driven growth: “Strait of Hormuz will surely hurt a lot of economies for a while.”; “Predicting a market crash in the next 5 years, due to overinvestment in AI.” High-growth respondents generally point to AI investment as an engine of growth—“a historic capital expenditure surge”— that lifts growth even before productivity gains arrive, and imply that any other exogenous shocks will likely be small compared to the scale of the AI-driven transformation: “We will see a boom in GDP that will be similar to the prime years of the industrial revolution, but at a much faster pace.”

Consumer Value of Generative AI

Question. Consider US adults who use generative AI chatbot tools in a given year. If they were offered a cash payment to give up access to all such tools for one month, what would the mean payment they would require — known as their willingness to accept (WTA) — be, in 2026 USD?

Results.⁴ The willingness-to-accept value of a month of generative AI access was $402.60 in June 2026. Experts and superforecasters expect it to climb steeply while the public expects it to stay roughly flat: for 2030 the unconditional median forecast reaches $750 for experts and $875 for superforecasters but only $408 for the public, and under the rapid AI progress scenario experts and superforecasters both forecast $1,500 while the public stays near $500. The public is also the least homogeneous early on — even in the 2026 unconditional forecast, a quarter of the public predict less than $150 while a quarter predict more than $420, a much wider range than experts' $403 to $500.

Rationale analysis

• Level of dependence: The most common dividing line is how dependent people will become on chatbots. Most high-payment respondents argue that dependence is primed to deepen considerably into daily life and work, and that this will raise the value of access: “As we depend more on AI, the costs of giving it up will rise.”; “One month is relatively easy in 2026. I can take a holiday and unplug…By 2030, AI will be more deeply ingrained in our lives and abstinence will be more difficult.” Low-payment respondents tend to see typical use as a casual convenience with ready substitutes. As one writes, “If you use ChatGPT as a better Google search for, say, cooking recipes, you can make do with just plain old Google search.”
• Who moves the mean: High- and low-payment respondents alike often note the mean willingness to accept (WTA) is dominated by a minority of heavy users with an exceptionally high WTA, but they tend to draw opposite conclusions regarding how this dynamic is likely to evolve going forward. High-payment respondents typically forecast that the high-user share will grow: “By 2030, there’s probably going to be a subset of people who are willing to pay extremely high amounts to be able to access models. This shifts the mean up a lot.”; “In moderate and especially rapid…the mechanism that drives the mean is the share of users who simply max out at $5,000.” Many low-payment respondents instead predict that as AI use expands, new users will drag down the mean: “New users will appear. These might be people that get less value out of generative AI (otherwise they would already use it). Hence WTA will be pulled down.”
• Implications of capability progress: High-payment respondents generally expect better models and agents to result in a rising WTA: “As capabilities increase and cost decreases, productivity gains from AI will be worth more…for a significant number of people, necessitating higher stipends to not use AI.”; “In rapid progress, AI agents could become central to productivity and personal management, pushing mean WTA toward $3,000 by 2030.” Several low-payment respondents argue that consumer chatbot value is unlikely to rise regardless of how much they improve: “We are fairly close to saturating the capabilities needed for adult user AI chatbots. The future value will be in more advanced use-cases; teams of AIs solving research problems or AI-native companies, etc; but those won’t turn up in this metric.” Others speculate that chatbots may become less valuable as AI becomes near ubiquitous elsewhere: “As AI gets embedded into operating systems and search, maybe people will not use ‘chatbot tools,’ which could flatten measured WTA even as real value rises.”
• Relationship to salaries: High-payment respondents tend to treat the value of workers’ time as the figure WTA will converge toward as AI become more and more integrated into workflows: “The question in some ways becomes effectively ‘how much would you need to be paid to not work for a month?’” A few low-payment respondents mention salaries, but typically to emphasize that they’re at a level where a lower WTA figure will likely prove to be enticing: “The average net monthly U.S. salary is roughly $4000, and most people’s livelihood doesn’t depend on access to AI chatbots, so they will continue to be likely to renounce the technology for a month in exchange for a sum that is meaningful for them.” Relatedly, one low-payment respondent notes that “by 2030, if AI is that helpful we might not have jobs, so [we] would not need AI at work.”
• Trust in the instrument: High-payment respondents generally accept the extended-ladder baseline and extrapolate its observed growth: “I’ll anchor off the Brynjolfsson et al. increase of 27% over 8 months, for a rise of 3.375% per month, and apply that to the FRI $402.60 baseline.” Low-payment respondents are often skeptical the measure tracks real value: “The methodology literature is clear that hypothetical WTA runs above real-money WTA, because a hypothetical question is not incentive-compatible—nothing turns on the answer—and the effect is larger and harder to correct for WTA than for WTP [willingness to pay].” A few discount the FRI baseline, with one low-payment respondent writing, “My answers assume a more conservative baseline now of 200.”

Self-Reported Happiness

Question. What will be the global average score (0–10) on the Gallup World Poll Cantril ladder — the population-weighted mean across all surveyed countries, as published in Our World in Data’s World series — in the following years and conditional on the following scenarios?

Results. The global Cantril-ladder score was 5.44 in 2025, and forecasters expect it to stay nearly flat. In the near term the groups barely differ: all three give a 2026 median of 5.4, and the middle half of each group falls between 5.4 and 5.5. Gaps open only slightly at longer horizons and under faster progress. Unconditionally the median rises to about 5.5 to 5.6 by 2040, and the main exception is the public under the rapid AI progress scenario, whose 2040 median of 6.1 sits above experts (5.6) and superforecasters (5.7).

Rationale analysis

• The 2026 time horizon: For 2026, the consensus is that the reported level of happiness will change little relative to 2025: “Given that the global survey data for 2024 and 2025 have become established facts, and the indicators are calculated using a 3-year moving average, the final score for 2026 is highly anchored mathematically.” The poles referred to below, therefore, mainly apply to 2030 and 2040.
• Impact on labor: Many flat-to-low-happiness respondents point to AI-driven job losses, particularly under the rapid-progress scenario, as the central consideration: “Unemployment is one of the main drivers of life dissatisfaction and unhappiness, so large-scale job displacement caused by AI could massively reduce life satisfaction.”; “Rapid progress likely means mass unemployment, which would cause serious unhappiness.” High-happiness respondents tend to focus instead on the potential for AI to improve the quality of work and allow for more leisure time. “Elimination of drudge work will be a positive for happiness,” writes one, with another adding that AI could “[lessen] the burden of work and domestic admin tasks and humanoid robots could assist with chores, caring assistance and healthcare.”
• Distribution of the gains: Many flat-to-low-happiness respondents expect that wealth stemming from AI adoption will not be broadly shared: “This will be a period of winner takes most and the rest of the population…will inherit a less happy future”; “Studies show that people are happier when they have about the same as the people around them…[and] I think AI will tend to increase inequality and people’s awareness of inequality.” High-happiness respondents often express or imply optimism that gains will be broadly shared: “AI will cause a huge increase in worldwide happiness due to the abundance it creates. People will have their basic needs met much easier in every aspect of life.”; “AI could raise income[s], improve education and health care access, reduce the cost of services, and make daily work easier.”
• The flat trendline: Flat-to-low-happiness respondents frequently point to the stability of the historical trendline: “The Great Financial Crisis, the COVID-19 pandemic, and the 2022 Ukraine-and-inflation episode collectively moved the global aggregate by no more than 0.10 ladder points.” Several also cite hedonic adaptation and human biology as reasons even transformative gains likely will not register. High-happiness respondents tend to argue that AI-scale change is not comparable to prior events that failed to move the score: “A decade of transformation delivering the end of poverty, collapsed disease burden, and material security [would be] a larger event than anything in that record.”
• Sense of purpose: Flat-to-low-happiness respondents often speculate that AI could “cause a crisis in human purpose” and a “loss of dignity” that could drag the index down: “Rapid progress may reduce human motivation to pursue life goals, as many tasks and aspirations could increasingly be achieved by AI.” High-happiness respondents tend to see AI mastery of formerly human tasks as liberating: “Everyone can use his/her free time to do what is most fun for oneself.”; “Plentiful leisure time resulting from rapid AI progress could significantly increase life satisfaction.”
• The population-weighted factor: Because the global happiness score is population-weighted, India, China, Africa and the wider global south dominate it; many respondents cite that factor but draw different conclusions from it. One high-happiness respondent writes, “In the West we tend to navel gaze, or focus on the latest outrage du jour, but when it comes to world happiness, the fact that millions of citizens of the global south have been slowly but surely climbing out of wretched poverty matters far more. I expect that trend to continue.” Whereas a flat-to-low-happiness respondent writes, “Sub-Saharan Africa’s fast-growing population sits at Cantril ~4.5, so its rising population weight pulls the aggregate down.”

Footnotes

1. FRI's survey included higher bid ladders and other methodological differences. Estimates are not directly comparable but still provide a meaningful frame of reference. ↩

2. In some cases, the "aggregate" refers to the mean; in others, the median is used, depending on which is more appropriate for the distribution of responses. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

3. We occasionally elicit participants' quantile forecasts (estimates of specific percentiles of a continuous outcome) to illustrate the range and uncertainty of their predictions. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

4. Some respondents were unsure what counted as “generative AI chatbot tools” (e.g., genAI embedded in services like Google search, or only standalone assistants like Claude and Gemini); whether work tools (e.g., Claude Code, Harvey) were included; and whether the question covered work or just personal use. These results should be read with more caution than usual. ↩