YouTube Isn’t TV. AI Isn’t Art. Depending on Who Gets Disrupted

A musician can say AI-generated music is killing human creativity, then use generative AI to make a flyer for an upcoming show. An indie artist can tell fans Spotify’s economics are terrible for musicians, push everyone toward Bandcamp, then open Spotify to listen to someone else. An ad buyer can insist YouTube isn’t TV while buying YouTube because that’s where the audience went.

Calling all of that hypocrisy is satisfying, but also incomplete.

People sit in multiple places in the value chain. They buy labor and sell labor. They’re customers and suppliers. They employ people and work for employers. They own IP and consume somebody else’s.

Technology gets a very different reception depending on which side of the transaction it hits.

When technology lowers the cost of something someone buys, efficiency feels pretty fucking great. When it lowers the economic value of something that same person sells, the conversation suddenly gets much more philosophical.

That doesn’t make the philosophy fake. People can be sincere and economically self-interested at the same time. Copyright, consent, attribution, artistic intent and labor displacement are legitimate issues.

They’re also sitting on top of a much simpler economic question that shows up every time technology makes something cheaper or faster to produce.

Who gets the productivity dividend?

Everybody Loves Automation When Somebody Else Gets Automated

Take the musician who needs a music video. A small production team might cost $5,000. An AI video tool can get them something usable for a couple hundred bucks. The musician just cut thousands of dollars from the cost of promoting the song. The director, editor, animator, or crew who might’ve gotten that work just watched the same “efficiency” erase a paycheck.

The ethical line around automation tends to get easier to move when the labor being automated belongs to somebody else. That doesn’t mean every artist works this way. Plenty reject generative AI across their entire creative process. Others separate assistive tools from fully generated work. Some care primarily about whether the model was trained with permission.

An artist can prefer Bandcamp when it’s their music being sold, then open Spotify when they want to listen to everybody else’s. In one transaction, they care about how much value flows back to the creator. In the other, convenience wins.

The contradiction gets sharper when someone demands that every consumer absorb higher cost or more friction to protect the economics of the creator, then declines to absorb that same friction when consuming somebody else’s work.

Technology doesn’t distribute its gains evenly. Cost savings often show up somewhere else as lost revenue, less labor, compressed margins or weaker bargaining power.

YouTube Gets Less “TV” the Closer You Sit to the TV Budget

The fight over whether YouTube counts as TV has always been funnier once you follow the money.

Consumers aren’t sitting around parsing the ontological boundaries of TV. They turn on the biggest screen in the house and watch video. YouTube sits next to Netflix, Disney+, Peacock and every other app on a CTV interface. Advertisers are chasing the same finite pool of attention across all of them.

People who say YouTube is TV are talking about attention. People who say YouTube isn’t TV are talking about money.

Put YouTube inside the TV category and a bunch of uncomfortable questions arrive with it. Which agency team owns the spend? Does YouTube take money from traditional TV budgets? Which measurement systems get used? What happens to expertise built around linear buying? Which sellers, intermediaries and planning organizations lose influence when creator video competes directly with professionally produced programming?

“TV” becomes an economic border.

Nobody has to be lying for that border to get defended. An ad exec can sincerely believe there are meaningful differences between a CBS drama and a MrBeast video. There sure as shit are.

But those differences don’t change the fact that both can compete for the same viewer, the same hour and eventually the same ad dollar.

Category definitions get a lot more emotionally charged when collapsing the category threatens someone’s economics.

AI’s “Is It Art?” Fight Is Already Turning Into a Licensing Business

AI is producing another round of category enforcement.

AI isn’t art. AI isn’t music. AI isn’t creativity.

There’s serious shit underneath those arguments. Models can be trained on copyrighted work without permission. Artists can have their likenesses or voices imitated. Fraud, attribution, royalty dilution, consent and disclosure are all live issues. Workers can lose income even when the output still requires some human supervision.

Those problems don’t get solved by winning an argument over whether the resulting file qualifies as “music.”

Markets eventually need rules granular enough to handle how the technology actually gets used.

Deezer said fully AI-generated music crossed 50% of its new daily uploads at the June peak, with roughly 90,000 synthetic tracks arriving per day. Those tracks still represented only about 1% to 3% of listening. Production capacity exploded. Demand didn’t follow at remotely the same rate.

Spotify’s taking a similarly practical approach. Its new AI Persona badge addresses whether an artist profile represents an actual person. AI Credits separately disclose how AI may have been used in making the music. Spotify also plans to keep AI Personas out of editorial and algorithmic recommendations by default unless a listener intentionally engages with them.

A human musician using AI during production, a fully synthetic artist, an unauthorized voice clone and a licensed AI remix create different commercial problems. Throwing all four into one “AI music” bucket becomes useless pretty fast.

The market’ s already moving past the binary question and into participation rules, disclosure, licensing and money.

Management Calls It Productivity. Payroll Calls It Exposure

Accedo sent me its new Off the Record report ahead of IBC, built around candid conversations with streaming execs about what they’re actually doing with AI.

It’s qualitative, the interviews are anonymous, and Accedo doesn’t disclose a respondent count. So treat it as a window into how these incentives are showing up inside streaming companies, not a census of the business.

One line jumped off the page: “employees don’t always have much incentive to automate their own jobs.”

Yeah. No shit.

Management can look at AI and see more output from the same payroll. An employee can look at exactly the same productivity gain and wonder how long it takes before somebody in finance decides the company can produce the same output with a smaller payroll.

A transformation director at a large media group expects a conventional eight-person software team to eventually shrink to roughly four people, built around a hybrid product and UX role plus AI-native engineering roles. Getting there would require rewriting job descriptions and rebuilding how business, product and technology teams interact.

Another executive described AI adoption as an 80/20 problem, with a relatively small group enthusiastically changing how they work while a much larger group lagged behind. Management may interpret that as resistance to productivity. Some employees are making a perfectly rational calculation about what happens if they become extremely good at proving that the company needs fewer people doing their job.

The company and the worker can agree that the software makes somebody more productive. They don’t automatically agree on who should benefit from that productivity.

When Everybody Gets Faster, Yesterday’s Price Gets Harder to Defend

Accedo’s report gets even better when the same incentive fight moves outside the company.

One product exec said customers don’t particularly care that a vendor calls its product “AI-powered.” They want to know “what they no longer have to pay for.”

Say a vendor used to need 12 weeks to complete a project. AI cuts the work to six.

That doesn’t automatically mean the client deserves a 50% discount. Companies have always invested in tools, IP, processes and expertise that let them do work faster, and keeping some of that efficiency as margin is part of the reward for getting better at the job. The old Picasso logic still applies: the buyer is paying for the result, not simply the amount of time somebody spent producing it.

AI complicates that bargain because the productivity gain is becoming available across the market at once. If one vendor can turn 12 weeks into six, that’s an advantage. If every capable vendor can do it, six weeks starts becoming the new competitive baseline.

The customer doesn’t suddenly own the vendor’s productivity gain. But the vendor also can’t assume yesterday’s labor economics will support tomorrow’s pricing forever. Competitors can quote faster timelines, lower prices, or both, and clients now know the tools exist.

AI doesn’t dictate who gets the savings. Competition, pricing power and differentiation decide how much stays with the vendor and how much eventually flows back to the customer.

Accedo found another version of the same problem inside company P&Ls. Execs described tangible productivity gains they couldn’t confidently convert into financial value. One COO questioned what a supposed 25% engineering productivity gain had actually bought when the engineering budget hadn’t fallen and the roadmap wasn’t moving 25% faster.

If this is the part of the AI conversation you actually give a shit about, the TSW Guide to AI & The Modern Media Workflow goes much deeper on where AI is actually changing media operations, workflow, distribution and monetization.

Calling something more productive is easy. Eventually the gain has to land somewhere measurable: lower cost, greater output, more revenue, faster delivery or higher margin.

If nobody can find it, there’s a decent chance somebody else captured it.

The Streaming Wars Take

Rights, pricing and headcount are where AI stops being a debate-club topic and starts getting booked into the P&L.

Suno launched its new v6 models after developing them with Warner Music Group, BMG and Believe. Warner had sued Suno over copyright infringement, then settled and moved into a licensing relationship. Its terms require licensed models, economic terms reflecting the value of music, and artist or songwriter opt-in for uses involving names, images, likenesses, voices and compositions. BMG’s agreement similarly provides participation and compensation for artists and songwriters who opt in.

Warner didn’t need the philosophical argument around AI music to disappear before creating a business model around it.

“Is AI music legitimate?” quickly breaks into more useful questions: Are we talking about a fully synthetic artist making fully generated music, or a human artist using AI somewhere in the creative process? Was an artist’s voice or likeness involved? Which catalog can train the model? Who gave permission? What can the customer generate? What has to be disclosed? Which uses trigger payment? How does the artist get a cut?

That delineation is going to matter more as AI becomes normal creative infrastructure. Calling everything touched by AI “AI music” makes about as much sense as calling every movie with CGI a cartoon. The useful line isn’t simply whether AI was involved. It’s what AI actually did, who directed it, whose work or identity went into it, and what the listener is being told.

“Is YouTube TV?” becomes: Which budget buys it? Who owns the account? What gets measured against it?

“Is AI making employees more productive?” becomes: How many people does the company need, what happens to compensation, and who keeps the savings?

One exec in the Accedo report expects streaming’s AI conversation to move toward subscribers per employee, engineering cost per service and operating cost per subscriber. Those numbers will expose where the productivity dividend actually went.

People can sincerely care about creativity, quality, ethics, identity and the meaning of art. An ad buyer can sincerely believe professionally produced TV differs from creator video. An employee can sincerely worry that AI-generated work lacks judgment.

People experience technological change according to where they sit in the value chain, and whether that change makes something they buy cheaper or something they sell worth less. Once the money starts moving, permission, pricing, headcount and bargaining power decide who actually captures the productivity dividend.