Topics
May 6, 2026
May 6, 2026
From December 2025 to February 2026, 2,000 people – everywhere from Honolulu to Newcastle, from ages 18 to 75, including medical professionals, entertainers, students, and farmers – were living their lives when something brought them to either ChatGPT or Gemini.
We wanted to see how people are actually using these tools. Not hypothetically or after-the-fact. So, using Dscout's Intercept tool, we caught them just as they arrived at either site and asked if we could follow along. And those 2,000 people showed us a little bit of everything.
From folks coping with OCD symptoms to planning their dream life—we saw it all in real-time, in the moment, stream of consciousness. They told us how they feel about AI, why they use it, and what it has replaced in their lives.
With so much data to unpack, we’re rolling out a report of what we found (coming soon!). But to give you a sneak peek, here are five things we learned about how folks are using and thinking about AI right now.
Most use cases in our study fell within the broad categories of learning and work, and both had high success rates. But at the top of the success charts were major decisions and personal advice, two categories with lower overall volume but even stronger results.
What made these sessions work was rarely the information itself. Most people arrived already knowing what they needed to do.
“I already had a bias toward what I wanted to see because I had built a history with ChatGPT.” - Quentien, 1/31/26 10:09 AM
That's AI working as a confidence tool. People used it to de-risk decisions they'd already mostly made, treating the response as a second set of eyes before committing. In major decisions, that looked like five-year financial plans and health logistics. In personal advice, it was often the right way to word a difficult text or navigate a hard conversation.
Whether AI as a validation engine is a feature or a risk is a question we’ll dig into in the full report.
Work tasks were the second most common use case for our participants, and they’re doing a lot of writing at work. AI eases the burden of the blank page for them.
Just as we saw with major decisions and personal advice, AI is functioning as a confidence tool here, too. Confidence at work is about tone. Workers often know what they want to say but can't find the right professional register to say it.
The quality of the output is often more important than pure velocity. Often this means the AI just needs to provide a solid starting point that can be refined. It doesn’t necessarily have to be perfect on the first turn.
Workers described this relationship as more collaborative than transactional, jumpstarting momentum and providing reassurance of professional appropriateness.
“I was going to basically provide a draft of my initial email template and then I wanted Gemini to actually fully complete the email and make a complete professional email that I could actually send.” - Hana 1/20/26 11:46 AM
Creative projects were our third most common category (and the most polarizing). When it works well, AI is a mechanism for creative flow.
But creative work also produced the highest failure rate of any category we measured, and it came from two directions. When asked to originate ideas rather than refine existing ones, AI defaults to the generic.
The more interesting failure, though, was the interpretation gap. Creative tasks rely on the user's ability to convey their vision, and when AI falls back on generic patterns, it breaks flow far more severely than it would in a utility task. A generic answer to a factual question is still useful. A generic rendering of your creative vision is a dead end, and it's frustrating to not feel understood.
“I need help with ideas for a marketing campaign. I tried to refine the prompt and it still didn't really understand what I wanted, and kept recycling the same ideas.” - Amy 12/19/25 2:03 PM
It’s much easier for AI to be useful than it is for it to be in tune with your mind's eye. In the full report, we dig into what separates the prompts that land from the ones that don't.
When we asked participants how they would have accomplished the same task without AI, more than anything else, they said they would have turned to a search engine. For learning, major decisions, and shopping, displacement rates topped 70%.
People are still searching though! During the 3 months of this study, we observed 12,500 hours of AI usage and a side-by-side trail of 700,000 Google searches. AI isn’t replacing search, it’s supplementing it.
Search provides raw materials, while AI does much of the assembly for you. Synthesis-heavy tasks favor AI, but many are still using search to verify information.
“I would rather ChatGPT do some of the legwork for me up front, and I can double check that work.” - Natasha, 12/16/25 4:08 PM
By the numbers…
When we asked participants to describe their relationship with AI in three words, the most common were helpful, reliable, and fun. When we asked who AI would be if it were a person, the two most common answers were assistant and friend.
“Probably a friend who I can just call up or text to be like ‘hey what do you think about this’ and feel trusting about their response.” - Kristen, 2/5/26 7:03 AM
When you explain a problem to a receptive listener, even an artificial one, you organize your own thinking in the process. AI reflects your thoughts back, and that's often enough to shift from stuck to moving.
From investigating a spiritual awakening to fixing broken glasses after a night of partying, what people largely used these tools for was something to think with. For a deeper dive on all of the above, and a more critical perspective on what it all means, stay tuned for our full report on the moments these 2,000 AI users shared with us.
You may also like…
