Metadata has finally found a publicist. Unfortunately, it’s a lawsuit.
Gracenote sued OpenAI in March, alleging the company used its proprietary entertainment metadata and the framework connecting that information without authorization. Then in June, music company Jamendo sued Nvidia, alleging that hundreds of thousands of audio files and related metadata were used to train AI systems. Both cases remain unresolved. The underlying business fight is already clear: AI companies need structured media knowledge, and the owners of that knowledge would like to get paid.
That should get the streaming industry’s attention.
Metadata spent most of the last decade buried beneath shinier conversations about content, subscriber growth, bundling, advertising, and product design. Nobody announces a new title identifier at an upfront. No analyst upgrades a stock because the season and episode hierarchy looks especially clean.
The viewer never sees the data layer. The viewer sees whether it works.
Search for a title and the correct version appears. Open a series and the episodes are in order. Ask where a game is streaming and get the right answer for the right territory. Click on an actor and see the correct filmography. Receive a recommendation that makes sense instead of one connected by the technical definition of “also contains people.”
Metadata makes all of that possible. And now AI wants the same thing.
The Catalog Has to Know What It Owns
Every streaming service presents its catalog as a clean collection of shows, movies, sports, and channels.
Inside the company, the same catalog can look more like a group project where nobody agreed on the file names.
One database uses the production title. Another uses the consumer-facing title. A distributor assigns its own identifier. Finance tracks the asset under whatever name appeared on the first royalty statement. The rights system knows the window closes Friday. The storefront thinks it closes next month.
The viewer sees one movie. The business sees four records and an email thread.
That mess stays manageable until the catalog grows, the company enters more territories, or another library arrives through an acquisition. Then every inconsistency starts multiplying across partners, versions, languages, rights windows, artwork packages, and financial reports.
The problem rarely announces itself as “metadata failure.”
It arrives as a rejected delivery. A missed launch. A rights violation. A report that doesn’t reconcile. A title that disappears from search. A premium piece of inventory packaged under the broad and inspiring label of “entertainment.”
Somebody fixes it manually. The business keeps moving. Leadership assumes the system worked.
It didn’t. A person worked around it.
AI Can Speak Before It Knows Anything
Everybody loves the conversational demo.
Ask the service for “a tense political drama under eight episodes” and watch the answer appear instantly. Ask which actor appeared in a franchise before joining another franchise. Ask where tonight’s game is available. Ask for the Christmas episode featuring the guest star from that medical show.
The language is the easy part.
A reliable answer requires the system to understand titles, episodes, seasons, performers, characters, franchises, genres, themes, rights, territories, schedules, and availability. It needs to know how those things connect and which record is correct.
That structure is the catalog graph.
Large language models can generate a polished sentence from incomplete information. They’re less impressive when the answer needs to match the actual catalog, the actual rights window, and the actual service carrying the event.
AI writes the response. Metadata keeps the response attached to reality.
Gracenote’s own business illustrates why this data carries value. Its products combine content identifiers and descriptive metadata used for search, discovery, distribution, and advertising across television, movies, music, and sports. Nielsen has also positioned Gracenote’s data for AI use, including agreements supporting AI-driven entertainment experiences.
The lawsuits will test what parts of those databases and descriptions receive legal protection. The Copyright Office says individual facts aren’t protected on their own, while original content and sufficiently creative selection, coordination, or arrangement may qualify. Courts will decide how those principles apply to the alleged uses in these cases.
The Boring Layer Controls the Expensive Layer
Streaming companies are preparing to spend heavily on AI search, personalization, advertising, automation, and customer support.
Those investments will inherit the data underneath them.
A service with clean identifiers and strong title relationships can build a useful conversational experience. A service with fragmented records can buy the same model and produce confident nonsense.
The same rule applies to advertising.
A seller may know an impression ran inside a playoff game, a premium drama, or a specific children’s program. The ad market may receive a generic category because the richer program information never made it through the supply chain.
Metadata gives advertisers more context about what they’re buying. It helps sellers package programming more precisely. It helps measurement companies recognize the same content across different systems. It helps rights teams understand where a title can run. It helps finance connect revenue to the correct asset.
The data layer touches the revenue layer whether leadership pays attention or not.
That’s why metadata can’t remain only a technology issue. Technology may manage the systems. Editorial may create descriptions. Legal may control rights. Ad sales may define packaging. Finance may own reporting.
Everybody touches the catalog. But somebody still needs to own the answer when those teams disagree.
Owning the Catalog Doesn’t Mean Building Everything
Media companies should continue licensing specialist data.
Vendors can deliver global identifiers, schedules, credits, images, sports information, language support, and enrichment at a scale most individual companies shouldn’t recreate. Buying that expertise often makes more sense than building it.
The risk appears when the company can’t explain its own catalog without calling the vendor.
Every media business needs control of the core rules that define what it owns and how that content moves. Which record represents the title? How do seasons and episodes connect? Where do rights live? Which version belongs in which territory? Which information reaches ad systems, partners, search, and finance?
Vendors can improve that foundation. They shouldn’t become the only place where the catalog makes sense.
AI makes this question more urgent because the company’s structured data will increasingly shape how its content gets discovered, recommended, packaged, licensed, and described by machines.
The company that controls the graph controls more of those outcomes.
The TSW Guide to Metadata
The TSW Guide to Metadata examines how weak metadata creates profit leaks across operations, finance, rights, advertising, partner delivery, and discovery.
The Guide moves past the AI headlines and into the operating reality. It explains why spreadsheets quietly become critical infrastructure, why adding people rarely fixes broken handoffs, and why another software purchase won’t solve a problem nobody clearly owns.
It also addresses the strategic question AI is forcing media companies to answer: Which parts of the data layer can they afford to license, and which parts do they need to control?
The Guide features insight from Rebecca Avery, a senior streaming operations executive and subject matter expert whose work spans content metadata, media supply chains, rights and avails, partner delivery, catalog onboarding, and platform integration.
Metadata isn’t the sexiest subject in streaming.
That’s probably why so much money has been allowed to leak through it.
Download The TSW Guide to Metadata.
The Streaming Wars Take
Metadata now sits directly in the path of revenue, rights, delivery, and AI. Companies with clean catalogs can cut manual work, reduce rights mistakes, improve partner delivery, give ad sales better context, and build AI products that know what they’re talking about.
The companies that don’t will keep putting expensive interfaces on top of cheap foundations.
So while the AI demo gets the applause, it’s the catalog decides whether it works.
The Streaming Wars is intentionally ad-free
We don’t run display ads. Not because we can’t, but because we don’t believe in them.
They interrupt the reading experience. They cheapen the work. And they burn advertisers’ money on impressions nobody actually wants.
So we chose a different model.
We say the things people in this industry are already thinking but don’t say out loud. We connect the dots beyond the headline and focus on explaining why things matter to the people working in this business.
If you believe industry coverage can exist without clutter and interruption, you can support it here → SUPPORT TSW.
Support is optional. But it directly funds research and continued coverage — and helps prove this model can work.
Support TSW →
