A favorite series ends, and the search for the next obsession begins. For most viewers, that hunt has become a chore. Fans feel it first. They aren’t just searching for what to watch next – they’re the engine driving word-of-mouth for everyone else.

Despite years of investment in recommendation engines, streaming content discovery still falls short. In Gracenote’s 2025 State of Play survey, viewers across six countries spend 14 minutes on average finding something to watch – and according to a Bango survey of 2,500 US consumers, 30% say choosing what to watch often takes longer than watching, rising to 46% among Gen Z and millennials. 

That frustration carries a commercial risk. In the same survey, 49% of viewers say they’re willing to cancel a service when it’s hard to find what to watch.

RECOMMENDATIONS ARE ONLY HALF THE BATTLE

For years, the industry has tried to fix this with smarter recommendation technology that uses viewing behavior to serve up the next title. 

However, the logic here is only partially sound. In Gracenote’s six-country survey, 27% of viewers (32% in the UK, 31% in France) say recommendations aren’t useful for finding new content, and only 28% choose what to watch based on service recommendations. And in the US, 51% say it’s getting harder to find what they want because there are too many services. 

A recommendation engine can put a title in front of a viewer. The copy around that title, like a synopsis, makes the case for pressing play. Viewers point to the sheer number of services, and that fragmentation sits outside any one platform’s control. The quality and cultural fit of the creative on each title sits inside it, and that creative is what tells a viewer why they’d enjoy watching. Fans choose and pass on titles for reasons a viewing history doesn’t record, such as a character they love or a reference only their community gets.

A title’s structured record (IDs, tags, genre) makes it findable. The descriptive wording –  synopses, titles, taglines, and culturally attuned copy – makes it recommendable. Recommendation technology has had years of investment. The behind-the-scenes marketing that carries a viewer from discovery to decision deserves the same attention.

FANS ARE DISCOVERING SHOWS OFF THE PLATFORM

Much of that discovery now happens off the platform. In Gracenote’s survey, slightly more viewers use internet search to find programs (55%) than browse within streaming platforms (54%). Social feeds and AI chatbots are two of the places it goes. Off the platform, a title competes through the creative that travels with it.

On TikTok, an average of 6.5 million film and TV posts a day went up in 2025, and 46% of TikTok users say they discover older titles on the platform that they wouldn’t otherwise have known about (Ipsos, commissioned by TikTok, September 2025). Independent data points the same way: 63% of 13–24-year-olds say they discover shows through clips on social platforms (Hub Entertainment Research, 2025, reported by CTAM). The on-service algorithm is competing with off-service creative it does not control.

Fans also do the recommending themselves. In a study of anime viewers across 10 markets plus Japan, commissioned by Crunchyroll and run by National Research Group in 2026, 77% say they have recommended anime to someone who had never watched it. That kind of recommendation reaches people a viewing history can’t, because they haven’t watched anything like it yet.

That changes the creative mandate for streaming platforms. A logline that nails the emotional hook, or a social asset tailored to a specific subculture, gives fans cultural currency they can actually trade. That level of contextual creative does far more discovery work than another generic row on a homepage. Fandoms don’t rally around algorithms – they rally around shared meaning, and they only amplify what carries it.

Disney+ is one platform bringing that fan creativity onto the service itself. In August 2026, Disney and TikTok announced a US pilot that will put fan-made TikTok videos about Disney’s franchises into Verts, the short-form video feed on Disney+, with creators given access to Disney’s content library through a new Creator Ambassador Program. The pilot puts fan creativity in the same app as the titles fans are talking about.

AI WON’T SAVE A THIN DESCRIPTION LAYER

The “what should I watch” question is also starting to move to AI assistants. 52% of US AI chatbot users say chatbots could become their preferred way to get information on why, where, and when to watch, and 5% say they already are (Gracenote, 2026, survey of 4,003 US chatbot users aged 13–79). 

An assistant can only recommend a title as persuasively as the material it finds about it, so that copy shapes whether viewers who search by chatbot ever hear about a title. Assistants can also get the basics wrong. In a June 2026 Gracenote test, an ungrounded LLM answering from training data alone fabricated all measured metadata for 506 of 2,600 titles, nearly one in five, across 13 countries. Gracenote sells the data used to ground those answers, so it has a stake in the result. Accurate facts tell a viewer what a title is. The copy a platform publishes in each market gives them a reason to choose it, and better AI makes that human layer more valuable.

This is a multi-market problem. Where AI assistants draw on the copy a platform publishes, a synopsis written for one domestic audience becomes the pitch every market hears. Thin, domestic-first copy caps how persuasively any AI can recommend a title, in any market. The same applies when a fan asks an assistant what to watch after a show they loved. If the copy never says what made that show matter, the assistant has less to go on.

TRANSLATED, BUT NOT RECOMMENDABLE

Search times also vary sharply by market. In Gracenote’s survey, French viewers spend 26 minutes searching, more than double the 12 minutes in the US, and 31% say recommendations aren’t useful for finding new content. 

The survey doesn’t give a cause. One explanation worth considering: a title’s metadata gets translated, but its reason to watch doesn’t always travel with it. What a fandom in São Paulo or Berlin finds recommendable is culturally specific. 

A direct translation of a title or synopsis can carry the basic meaning from one market to the next. It won’t necessarily carry what makes the show worth watching there, such as the hook or the reference that lands with local viewers. When it doesn’t, viewers can scroll past a title that a recommendation engine has already put in front of them.

FINAL THOUGHT

Discovery is a creative and cultural challenge misfiled as a purely technical one. Metadata makes a title findable, whereas cultural resonance makes it recommendable. For platforms looking to reduce churn and lift engagement, regional cultural IQ belongs in the growth budget right alongside the recommendation engine. Ultimately, discovery scales only when the narrative hook arrives in each territory intact.