The Challenge
Understanding how a tool used by hundreds of millions of people each week shapes emotional well-being isn't a question you can answer with a one-time survey. Early in 2025, researchers at OpenAI and MIT designed a randomized controlled trial built for the long haul: roughly 1,000 participants, each completing 28 daily touchpoints, 4 weekly surveys, and 2 comprehensive assessments over a full month. That's 34 total interactions and more than 200 minutes of engagement per participant, a level of sustained commitment that most studies never attempt, let alone achieve.
Designs this demanding are where online studies usually break down, and this one faced three practical challenges at once:
- How to find participants willing to complete daily tasks for multiple weeks
- How to communicate with participants directly in a streamlined way, to remind them of daily sessions and weekly surveys
- How to have complete confidence that data from more than 32,000 collection points would be trustworthy enough to answer the research question
Most participant pools cannot sustain a month of daily participation without retention collapsing, and every dropout erodes the statistical power the trial was designed to deliver.
The Solution
OpenAI and MIT chose CloudResearch Connect because the platform offers everything a study of this complexity needs: longitudinal projects with retention rates of 80% or more, vetted participants with a proven track record of engagement, the technological infrastructure required to communicate with participants and track their progress, and a reputation for high-quality data that helps answer pressing research questions.
How the Study Worked
The trial randomized participants across multiple interaction modes (voice-to-text or text-only conversations with ChatGPT) with some participants given specific topics to explore. Over 28 days, the participant journey looked like this:
- Day 0Baseline assessment
- Days 1–28Daily ChatGPT sessions plus a post-session evaluation
- Days 7, 14, 21, 28A weekly well-being survey
- Day 28A comprehensive post-study assessment
Managing that level of daily engagement across roughly 1,000 people is a significant logistical challenge. Four features of Connect, aligned with best practices for retaining participants in longitudinal studies, made this possible.
Vetted, High-Quality Participants
Connect maintains its reputation for data quality through careful participant vetting, screening participants to ensure they provide reliable, thoughtful responses. For a 28-day study requiring daily engagement, that vetting was crucial: OpenAI could trust that participants who enrolled would show up day after day and provide quality responses.
Targeting the Right Sample
The study required specific types of participants, including considerations about prior ChatGPT usage. Connect's participant database and screening tools let the research team recruit exactly the sample their research questions required, all before a single session was fielded.
Built-In Participant Communication
The hardest part of any longitudinal study is keeping participants engaged over time. Connect's messaging infrastructure let the team send daily reminders and maintain consistent communication throughout the 28-day period, the kind of sustained engagement that is difficult to achieve without a direct channel to participants.
Fair Compensation
Each participant who completed the full protocol earned $100. That investment kept people motivated through more than 200 minutes of study activities, signaled that the researchers valued participants' efforts, and likely improved data quality, too.
Results
981 participants completed more than 32,000 surveys (as well as roughly 1,500 hours of audio data from the voice-mode conversations) across the 28-day randomized controlled trial, with retention holding above 80% throughout. The findings were published by OpenAI in its affective use study report.
For a deeper look at how the study was structured and fielded, read our full breakdown of OpenAI's 28-day study on Connect.
What's Next
As AI becomes part of daily life, understanding how people interact with it over extended periods only grows more critical. This study demonstrates that ambitious longitudinal research on human–AI interaction can be conducted online with rigor and reliability, and it adds to a portfolio that includes the most intense longitudinal study ever conducted online, run in collaboration with IARPA.
The lessons from those projects are built directly into Connect Waves, which cuts the time required to set up and manage longitudinal studies. Researchers planning their own multi-wave designs can start with Connect's longitudinal study tools.
About OpenAI
OpenAI is an AI research and deployment company and the maker of ChatGPT, used by more than 400 million people every week. Alongside building frontier AI systems, OpenAI studies how those systems affect the people who use them, including large-scale behavioral research into the emotional and social dimensions of human–AI interaction.