Morgan Stanley published their new Intern Survey this year, and per usual, it's one of the best ways to see how people (roughly 20-23) use different platforms and services.
Thread with more:
Thread with more:
ChatGPT absolutely dominates usage. Of the respondents that use it weekly, 72% use it!
The next is GitHub Copilot at 18%, with Gemini and Claude at 22% and 21% respectfully. Perplexity is at 18%, and while there are others.... it's meaningless usage (sorry Meta/Grok)
The next is GitHub Copilot at 18%, with Gemini and Claude at 22% and 21% respectfully. Perplexity is at 18%, and while there are others.... it's meaningless usage (sorry Meta/Grok)
Now what's super interesting is the growth of these...
Year‑over‑year comparisons reveal dramatic gains for some platforms. In 2024, only 6 % of interns used Claude at least occasionally; that figure jumped to 25 % in 2025, and daily use rose from 1 % to 3 %. Gemini’s at‑least‑occasional user base doubled from 12 % to 25 %, while GitHub Copilot rose from 13 % to 28 %. ChatGPT remains dominant—occasional use climbed from 90 % to 95 %, and daily use from 13 % to 30 %
Year‑over‑year comparisons reveal dramatic gains for some platforms. In 2024, only 6 % of interns used Claude at least occasionally; that figure jumped to 25 % in 2025, and daily use rose from 1 % to 3 %. Gemini’s at‑least‑occasional user base doubled from 12 % to 25 %, while GitHub Copilot rose from 13 % to 28 %. ChatGPT remains dominant—occasional use climbed from 90 % to 95 %, and daily use from 13 % to 30 %
As you'd expect, it's mostly used for working with text and generating images. ChatGPT is, once again, used way more.
The other thing... almost 50% of ChatGPT users here pay for ChatGPT! That % of paying users is WAY lower for everyone else.
The other thing... almost 50% of ChatGPT users here pay for ChatGPT! That % of paying users is WAY lower for everyone else.
When asked what about AI concerned users, most of it was around misinformation and accuracy. Almost 85% of the interns were concerned about misinformation and accuracy.
So they're power users, but also concerned about the accuracy of that usage. That's inherently interesting.
So they're power users, but also concerned about the accuracy of that usage. That's inherently interesting.
The few things I think are interesting here are:
1. ChatGPT dominates. Microsoft wins because of this. Microsoft Copilot itself isn't mentioned anywhere, it was never brought up by the interns. The thing is, MSFT is all the infrastructure for Azure with ChatGPT. All the ChatGPT agent computer instances, most of the GPU capacity, the hosting, etc are all on Azure. Even if Microsoft doesn't get direct consumer usage, they get the majority of the revenue via data centers.
2. Consumer coding and prototyping is helpful. Claude grew to match Gemini within this pool of interns in a year, likely due to how insanely good it is at coding. If you noticed, GitHub Copilot was absolutely killing it. This makes me think Vibe coding is huge (I hate that term) and Anthropic is growing fast with a consumer base because of this.
3. Perplexity is still niche. They just are, we see a lot in online circles about Perplexity, about Comet, and all the cool things Perplexity is doing. The reality is they are a niche brand. That's ok! They aren't the OpenAI/Anthropic/Meta/Google lab it feels like they are.
4. Gemini feels pigeon holed into Google and Google services vs. a consumer chatbot we see with ChatGPT. ChatGPT seems trusted, used, and even places where Gemini is better, ChatGPT is still used. There's a different level of trust there. Given Anthropic catching up to Google in this demographic, I'd wager OpenAI is still winning even with worse models. Anecdotally, I find the ChatGPT product to be much better.
1. ChatGPT dominates. Microsoft wins because of this. Microsoft Copilot itself isn't mentioned anywhere, it was never brought up by the interns. The thing is, MSFT is all the infrastructure for Azure with ChatGPT. All the ChatGPT agent computer instances, most of the GPU capacity, the hosting, etc are all on Azure. Even if Microsoft doesn't get direct consumer usage, they get the majority of the revenue via data centers.
2. Consumer coding and prototyping is helpful. Claude grew to match Gemini within this pool of interns in a year, likely due to how insanely good it is at coding. If you noticed, GitHub Copilot was absolutely killing it. This makes me think Vibe coding is huge (I hate that term) and Anthropic is growing fast with a consumer base because of this.
3. Perplexity is still niche. They just are, we see a lot in online circles about Perplexity, about Comet, and all the cool things Perplexity is doing. The reality is they are a niche brand. That's ok! They aren't the OpenAI/Anthropic/Meta/Google lab it feels like they are.
4. Gemini feels pigeon holed into Google and Google services vs. a consumer chatbot we see with ChatGPT. ChatGPT seems trusted, used, and even places where Gemini is better, ChatGPT is still used. There's a different level of trust there. Given Anthropic catching up to Google in this demographic, I'd wager OpenAI is still winning even with worse models. Anecdotally, I find the ChatGPT product to be much better.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
