AI can improve content discovery, but not if people don’t trust it

The modern TV landscape has a math problem. People are paying for more services, but they’re struggling to find things to watch. This has clear financial implications, but it’s also souring consumers on the overall TV experience.

Amid the overload of content, generative AI (GenAI) has the power to better connect people with what they’re looking for. Right now, however, consumers don’t trust the results that GenAI provides. This presents a key barrier that publishers will need to overcome to ensure high-quality, next-gen content experiences.

Today, 53% of Americans pay for four or more streaming services, but just as many say the amount of available content and services has become overwhelming¹. On paper, large language models (LLMs), the engines that power GenAI, seem like the ideal guide for consumers trying to navigate growing TV fragmentation. But with 75% of Americans fact-checking the results that AI chatbots like ChatGPT provide, LLMs have some work to do to prove their worth—especially when they’re incorporated into the TV services they’re paying for.

The high stakes of hallucinations

In a casual (and free) search for information, incorrect and incomplete results from a chatbot are a minor annoyance, and most consumers know that AI makes up answers when it doesn’t know the truth (i.e., ‘hallucinations’). Their expectations within a paid TV or streaming service, however, will be much higher, and poor experiences can lead to service-ending consequences as consumers weigh the value of the individual services they pay for.

The content conundrum is real. In fact, only 15% of people definitively know what they want to watch when they turn their TVs on¹. As a result, the time spent trying to find something to watch is rising: 14 minutes, on average. 

When we look at the growing expanse of content available, it’s easy to understand why discovery times are rising. In addition to the wealth of programming across traditional television channels (MVPDs and vMVPDs), Gracenote has a record of:

• More than 1.8 million program titles across nearly 350 SVOD catalogs, and
• Almost 210,000 program titles across nearly 2,100 individual FAST channels

And streaming catalogs are only going to get bigger. Over the past year, for example, the five streaming services tracked in the Gracenote Data Hub grew their catalogs a collective 20%. Similarly, global FAST channels grew their catalogs by nearly 15%.

AI and the future of entertainment

Amid the fragmentation, AI can help—even among untrusting consumers. Concerns about accuracy notwithstanding, nearly six in 10 people say AI has the potential to become their favorite source of entertainment information or they already are. And on average, 66% believe AI will be very or extremely important in providing good entertainment experiences.

In the world of entertainment, publishers and content providers know their customers won’t want to stop and verify every search query they launch. They also know that cancellation risk is real. Broadband TV News reported that average churn rates across major streamers were at 5.5% last year, up from 2% back in 2019.

Bridging the trust gap with consumers is a key value proposition, and LLMs that are grounded with industry standard data will ensure timely and accurate experiences that are also hyper personalized—a key advantage over traditional search infrastructures.

The adoption of GenAI is an inevitability. The advantages of a natural language interface are too great to ignore as the TV landscape continues to fragment. Through this lens, it’s very possible that the winners in the streaming wars aren’t the ones with the most content; they’ll be the ones who ensure that the right viewers can find it.

For more insights, download our TV search and discovery in the AI era report.

Note

1. Gracenote 2025 streaming user survey