The AI Trials Observatory launches

The Observatory’s first release: what randomized trials show so far about how generative AI affects people’s mental health and belief accuracy.
Published

November 10, 2026

Welcome! This is the first announcement from the AI Trials Observatory: a systematic, regularly updated record of what randomized trials show about how generative AI affects people. Each announcement summarizes an update to the evidence record in a short digest.

The Observatory is run by scientists working in academia and nonprofit organizations focused on AI evaluation. Learn more about us.

Generative AI may affect people in many ways. The Observatory launches by tracking two particularly salient outcomes—people’s mental health and the accuracy of their beliefs—and more outcomes are in preparation.

Let’s see what we’ve found so far.

Mental health in brief:

Belief accuracy in brief:

Mental health

On July 27, 2026, the Observatory conducted its first systematic search for randomized trials on how generative AI affects mental health. The search uncovered 62 papers reporting the results of 66 eligible trials in total. Of these, 52 trials reported enough information for us to compute their effects for the dashboard. Here are some of the things we learned.

The evidence is growing fast. As many reports were published in the first 7 months of 2026 (31) as in the previous five years combined (31). If this pace keeps up, the results described below could look very different after the next few updates to the evidence record.

The research frontier currently lags the AI frontier by 7 months or more. The newest AI model tested across the 52 trials is GPT-5.2, which was released in December 2025, 7 months before the date of our evidence search. That trial tested how an AI system built atop GPT-5.2 affected participants’ stress levels. For trials on other outcomes—including anxiety, depression, and wellbeing—the current lag is at least 12 months.

The majority of trials tested “therapeutic” AI: systems purpose-built or prompted to support mental health, rather than general-purpose or companion AI systems.

The average effects of therapeutic AI are broadly beneficial: compared to control groups, people exposed to therapeutic AI reported fewer symptoms of depression (a standardized effect of -0.33 across 11 trials), anxiety (-0.33, 12 trials), and psychological distress (-0.36, 10 trials), and reported greater wellbeing (+0.27, 9 trials) and positive mood (+0.30, 6 trials). These effects are about the size of caffeine’s effect on attention, and roughly two-thirds the size of cognitive behavioral therapy on depression relief.

Interpreting these average effects requires nuance. Despite growing rapidly, the evidence is still thin enough that it is compatible with the average effects being zero in some cases (anxiety and psychological distress). Furthermore, average effects are taken across different contexts. On a context-by-context basis, the evidence is consistent with effects being beneficial in some contexts, while being negligible or even harmful in others. Finally, in trials that compared therapeutic AI to an alternative form of support, such as human interaction or a self-help tool (instead of a control group), the average effects are typically smaller and more uncertain.

When it comes to general-purpose and companion AI systems, there is currently too little evidence to say much either way. The one exception is anxiety: compared to control groups, people exposed to general-purpose AI reported fewer anxiety symptoms (-0.64, 5 trials), though here too the evidence is thin and there remains much uncertainty; the average effect may change a lot in future updates as more evidence accumulates.

Belief accuracy

On October 12, 2026, the Observatory conducted its first systematic search for randomized trials on how generative AI affects the accuracy of people’s beliefs. The search uncovered 42 papers reporting the results of 48 eligible trials in total. Of these 48 trials, 42 reported enough information for us to compute their effects for the dashboard. Here are some of the things we learned.

The evidence is growing fast. Two-thirds of the reports (28) appeared in 2026 alone, twice as many as in the previous three years combined (14). If this pace keeps up, the results described below could look very different after the next few updates to the evidence record.

The research frontier currently lags the AI frontier by 5 months or more. The newest AI model tested across the 42 trials is ChatGPT tested in May 2026, 5 months before the date of our evidence search. That trial tested how access to ChatGPT affected people’s factual misperceptions. For trials on other outcomes—across the broad categories of conspiracy belief, science skepticism, and belief in misinformation—the current lag is 10 months or more.

The majority of trials tested “corrective” AI: systems purpose-built or prompted to improve belief accuracy, rather than general-purpose or companion AI systems. There are scant trials testing AI systems actively steered to mislead people.

Corrective AI improves belief accuracy on average: compared to control groups, exposure to corrective AI decreased people’s factual misperceptions (a standardized effect of -0.24 across 6 trials), as well as their belief in conspiracy theories (-0.45, 9 trials) and misinformation (-0.61, 5 trials, driven mostly by unusually large effects from two trials). These average effects appear similar to or larger than those of non-AI interventions designed to improve belief accuracy, such as fact-checks and media literacy interventions, documented in other research (e.g., see here, here, and here). That comparison is only rough, since it draws on separate studies. But in trials that directly compared corrective AI with an alternative, such as web search or a fact sheet, we found the average effects of AI are slightly larger on some outcomes (e.g., decreasing factual misperceptions: -0.22, 5 trials).

Interpreting these average effects requires nuance. Despite growing rapidly, the evidence is still thin and in most cases there is a lot of uncertainty. Furthermore, average effects are taken across different contexts. On a context-by-context basis, the evidence suggests corrective AI could be highly beneficial for belief accuracy in some contexts, while being much less beneficial in others (e.g., see misinformation belief).

When it comes to general-purpose and companion AI systems, the evidence is thinner and less clear. While there is a reasonably sized average effect of these systems on decreasing factual misperceptions (-0.65, 5 trials), it is mostly driven by unusually large effects from two trials. Average effects on misinformation belief are currently small and compatible with zero. As with the other results, these numbers are likely to change as further evidence accumulates.

Finally, all of the effects highlighted above come from beliefs that were discussed or possibly discussed with AI. However, AI exposure may have a different effect on beliefs that were not discussed. For example, a reliance on AI could lead people to become less discerning, harming accuracy on beliefs that were not discussed. Alternatively, interactions with AI could impart transferable skills and judgment, improving accuracy on such beliefs. Some trials have tested the effect of AI systems on beliefs that were not discussed, but the evidence is currently too thin to say much. In future updates, that may well change.

The above is just a summary. Don’t take our word for it: