LLMs are disrupting the world. In my Summer 2025 #FutureOfTV.Live TV survey I decided to field a complex, open-ended question and see if Claude, my preferred LLM, could help me zip through the tedious, difficult task of coding, validating, and analyzing 600 US-census adjusted, unstructured human responses.
This originally appeared on StreamTV Insider.
What would users of The Roku Channel write-in if I asked them what their favorite program was?
Go to slide eight in the above report and you’ll see the result. The slide elucidates a key insight about viewer habits in one beautiful, impactful word cloud.
That key insight?
- Viewers aren’t a monolith. Their approach to TV discovery falls into one of several patterns. Show titles are key, of course. But a huge number of viewers are genre-focused; others dive in with specific channel names in mind; some channel-surf; and a certain percentage are confused as to whether what they’re watching is an app, a channel, a VOD service or something else.
One thing that was also clear to me was that this one technique had expanded the insight I’d normally get from each question.
So I dedicated five of my ten-question Fall 2025 survey to exploring the technique further. Two were dedicated to sports and music TV fans. Two were dedicated to Samsung TV Plus and Pluto TV users. And the final was dedicated to the question of TV discovery using ChatGPT.
That question: Have you ever used ChatGPT or a similar LLM (e.g., Claude, Gemini, Perplexity) to find sports, news or other shows to watch on TV? Describe the experience.
How’d it go?
First off, as far as the usefulness of the technique overall, you’ll have to see it and judge for yourself. The latest report is available for free, here. I’m satisfied with the results.
It was not easy. Maybe I was pushing Claude to its limit. But I found myself frustrated with it, correcting it, trying to coax the right diagrams out of it, and checking everything multiple times. It felt like January 2023 all over again.
But mostly? It worked. It pulled five great insights for me that would likely not have made it through a more traditional survey technique and analysis.
1. Only 16% Have Used, But Their High Praise Stands Out
In a typical survey, I’d specifically ask a user if they had used ChatGPT. Answers would be constrained to a Yes, No, or something in between.
Using an LLM-dependent open-ended technique, I was able to arrive at a similar place, ending up with a specific number, with a Yes, No, or Maybe, as determined by Claude. 16% of the answers essentially represented a Yes, with 65% representing a No.
chatgpt for TV viewers graph
But with the open-ended, LLM centric analysis I was able to learn so much more. Here are four additional insights from the one question.
-
Sensing Positive GPT Vibes. While 16% of the respondent base is small, when you read the comments, it’s striking how positive they are. “Super helpful to use,” “Super easy and works great,” were just a couple of the actual responses.
-
**TV Discovery Trifecta: Easy, Fast, Personal. **Respondents consistently praised the experience using words like easy, fast, and personal. “Makes it easy to find the episodes I’m looking for”, “very simple and easy”, and “it was really easy + makes it so much easier” were a few responses
-
**Sports TV’s Top Challenge. **But the most striking comments and insights to me were those of sports fans that had already discovered how good it was for helping fans find when and where the game would be on and televised. “A full explanation of days and times of games,” wrote one respondent. “I used it to find new way to engage in sports content,” wrote another.
-
**ChatGPT’s Game to Lose. And finally, our **visualization showed that ChatGPT received 24 mentions to Gemini’s 7, with Perplexity and Copilot both at 3. That makes this opportunity ChatGPT’s to lose. (Which, they might well, given the stiff competition.)
TV discovery is widely stated as one of the most difficult challenges today in our business. The solution has to be fast, easy, and personalized. In the world of sports TV, where TV schedules for big games are not always published sufficiently in advance, this problem becomes even more pressing.
And here again, ChatGPT is absolutely crushing it. Below is a visualization of this question that you can find in the report. But I’ve pulled some of the choicest responses for easier reading below.
- Yes, I’ve used ChatGPT and similar AI tools to find sports and news shows. It’s quick and helps me discover programs I like.
- I used Gemini to find out what channels that was playing Monday night football
- Yes i used it to help find where’s the football on Amazon Prime video!
- Yes I used it to find a college football game
- I have used ChatGPT to find shows or sports to watch, The experience was fast and convenient
- Yes I have I alwyss use chatgpt to make sure I find all the sports im looking for definitely yes. It is a great new way to engage in sports content.
- Yes, it gave quick, personalized TV and sports recommendations
- Yes Gemini to find college football games
- it was easier to find football listings
- Yes, I use it to find my team that I like and it makes it so much easier
- Yes, I have used Microsoft CoPilot to find out what games were on TV and when they were playing
- Yes. I use it to find sport
For more, read the table below and check out the full report. LLMs can add real value in the TV discovery domain. I’d expect to see rising innovation on this front soon. What do you think? Comment, share, like.
chatgpt results chart
Brian Ring is a streaming technology expert with deep experience in full-stack GTM, Product Marketing and Revenue Growth motions. He’s worked with Amagi, Synamedia, MediaKind, Mux, Ateme and many others. His expertise includes all segments of video tech infrastructure including content creation, management, distribution, and monetization. Current consulting projects are focused on LLM-integrated media workflows; the CTV programmatic stack; the next evolution of FAST.