For seventy years, a television ad was bought by a person who called another person. An agency negotiated with a network months ahead of a season, locked in a block of spots, and trusted a panel of metered households to say how many people watched. The unit was the gross rating point. The pace was once a year.
That world is ending. In May 2025, streaming passed broadcast and cable combined for the first time in Nielsen's measure of total US television use, taking 44.8 percent of viewing against their 44.2 percent. The screen did not change. The way ads reach it did. Connected TV, the term for any television connected to the internet through a smart TV, a Roku, a Fire TV stick, or a games console, is now bought through the same auction plumbing as a display banner. That is the real story: not that people moved to streaming, but that the biggest screen in the house quietly became a programmatic channel.
Origin: how the living room joined the auction
Streaming with ads is not new. Hulu carried commercials from its launch in 2008, and for years it was nearly the only major service that did. The shift that mattered was not ad-supported streaming itself. It was the moment a streaming ad slot became an item that software could buy in real time.
Early streaming ad sales looked like old TV sales with a new coat of paint. A buyer signed an insertion order, a publisher reserved a fixed number of impressions, and humans handled the rest. Programmatic display had run on real-time auctions since around 2009, but CTV inventory stayed mostly manual through the first half of the 2010s. The hardware was online; the buying was not.
It changed when the demand-side platforms that already bought display learned to bid on a TV impression. The Trade Desk, working with Roku and other early partners, ran some of the first programmatic CTV campaigns in the mid-2010s, and the appeal was immediate: a buyer could apply audience data and frequency control to a television ad, things linear TV could never offer. One often-cited early campaign for Honda pivoted from desktop video into CTV and scaled across an East Coast dealer network on the strength of the results, even as the executives involved asked what a Roku was.
From there the plumbing fell into place. Supply-side platforms built for video, Magnite and FreeWheel chief among them, gave streaming publishers a way to expose inventory to many buyers at once. Ad servers learned to stitch commercials into a stream. By the early 2020s a CTV impression could move through the same exchange, the same auction, the same identity tools as any other programmatic unit. The living room had joined the marketplace. What it still lacked was supply.
Present: three new ad giants and a 44 billion dollar market
The supply problem solved itself in about two years, because the largest subscription services decided they wanted ad money.
Netflix held out the longest and reversed hardest. It launched a cheaper ad-supported plan in late 2022 after years of insisting it never would. Adoption was slow at first and then was not. By May 2025 the company said the ad tier reached 94 million monthly active users, and by its 2026 upfront it claimed more than 250 million monthly active viewers worldwide. Read that number carefully: Netflix moved from counting profiles to counting estimated viewers per household, which inflates the figure against the old metric. The direction is still real. Where the ad plan is offered, it has been drawing the clear majority of new sign-ups.
Amazon did not wait for anyone to opt in. In January 2024 it made ads the default on Prime Video and asked subscribers who wanted them gone to pay extra. Overnight, an enormous audience became ad-supported. Amazon's total ad revenue jumped 24 percent to 11.8 billion dollars in the quarter that followed, and the company later said Prime Video ads reached around 315 million viewers globally. Disney reported 157 million monthly ad-supported viewers across Disney+, Hulu, and ESPN+ in early 2025, with the ad tier pulling more than half of new US Disney+ sign-ups.
Three services that trained a generation to expect no ads now sell them at scale. That is why the money arrived. Global CTV advertising revenue reached about 44 billion dollars in 2025 on Omdia's count. In the US, eMarketer expects the CTV ad market to more than double between 2022 and the end of 2026. The clearest signal of the power shift is the upfront itself: in 2026, US CTV upfront spending of 17.73 billion dollars is set to pass primetime linear at 16.98 billion for the first time. The ritual built for broadcast now favors streaming.
How a CTV ad actually gets bought
A buyer reaching that inventory has roughly three routes, and they trade control for flexibility.
Programmatic guaranteed is the closest to the old way. The advertiser and the publisher agree on a fixed number of impressions at a fixed price, then execute the deal through a demand-side platform such as The Trade Desk, Google Display & Video 360, or Amazon DSP. It is an insertion order run on programmatic rails: reserved inventory, automated delivery, predictable volume. It is how most premium streaming inventory, and almost all live sports, still trades.
A private marketplace, or PMP, is an invitation-only auction. A publisher opens premium inventory to a chosen set of advertisers, who bid against each other but commit to nothing. The buyer gets transparency about where the ad runs and keeps control of price. The open exchange exists for CTV too, but premium streamers expose less of their best inventory there, and it is where the channel's quality problems concentrate.
None of these is the old TV buy. Every one applies audience targeting, applies frequency rules, and clears through software. A media buyer who understands display real-time bidding already understands the machinery; only the screen is different.
The measurement mess: real, and not close to fixed
Here is the part the channel's growth story tends to skip. CTV is bought like programmatic, but it cannot yet be measured like programmatic, and it never had the thing that made linear TV legible: a single agreed currency.
For decades, US television ran on Nielsen. Buyers and sellers did not love it, but they transacted against one number, which made a deal possible. Streaming broke that. The services with the most viewing, Netflix, Prime Video, YouTube, do not share their viewership data on common terms, so there is no shared count of who saw what. Nielsen still anchors national TV measurement, but challengers including Comscore, iSpot, and VideoAmp have taken real ground. By one Advertiser Perceptions survey, roughly 60 percent of marketers used a Nielsen alternative in the prior year, most of them running it alongside Nielsen rather than instead of it. The realistic outcome is not one new winner. It is a multi-currency market, and several analysts expect alternative currencies to stay as add-ons, never the universal yardstick. That sounds tolerant. In practice it means two parties to a deal can each measure the same campaign and disagree on what happened.
Fragmentation is not only a currency problem. It breaks frequency capping, the basic discipline of not showing one person the same ad too many times. CTV is not one platform; it is dozens of apps running on dozens of devices, each with its own identity system. Buy across Roku, Fire TV, and a smart TV brand and there is no native way to know the same household is being hit on all three. The results get genuinely absurd. One CTV ad-quality case study found a single device that logged 711 ad requests in ten days against a cap of three impressions. Every one of those extra impressions was paid for, and every one made the viewer like the brand a little less.
Then there is supply-chain fraud. Because a CTV ad request can be faked by software with no screen behind it, the channel attracts invalid traffic, and verification firms keep finding it. Pixalate estimated that in the first quarter of 2025, 13 percent of global open programmatic CTV ad impressions were sold by unauthorized sellers, meaning the entity selling the inventory was not one the publisher's ads.txt file had cleared to sell it. The honest summary is uncomfortable: the channel grew faster than its plumbing for counting, deduplicating, and verifying. None of that is solved.
Future and impact: attention as a currency, and a tighter grip at the top
Two forces will shape the next few years. One is a genuine attempt to fix measurement. The other is consolidation that may decide who gets measured at all.
The fix gathering the most momentum is attention measurement. Viewability, the metric inherited from display, asks a low-bar question for a TV: was the ad on screen at all. It cannot tell the difference between an ad watched and an ad playing to an empty room while someone makes dinner. Attention vendors aim higher. Adelaide builds an Attention Unit score from signals including time in view, content type, and ad placement, and reports that campaigns optimized to attention delivered meaningful lifts in both upper and lower funnel outcomes. The signals are moving into the buying path itself: OpenX and TVision have built real-time attention targeting for CTV buyers, and the Media Rating Council and IAB finalized attention measurement guidelines in late 2025. Attention is not a perfect currency. Eye-tracking panels are small, and a single score hides a lot. But it answers a question viewability cannot, and a growing share of buyers expect it to become a standard way to value CTV inventory.
The formats are evolving alongside the measurement. Pause ads, which appear when a viewer stops the stream, turn a dead moment into inventory. Shoppable and interactive units let a viewer act on a remote: Amazon has extended remote-clickable shoppable ad technology beyond Prime Video, with Samsung becoming the first external platform to integrate it. The pitch is that CTV can drive a purchase, not only deliver reach, which is the ground performance-minded advertisers care about.
The harder trend is who controls the channel. The same scale that made streaming an ad market is concentrating it. Omdia projects that by 2030, Google, Amazon, and Netflix will hold half of an 81 billion dollar global CTV ad market between them, with Google at 26 percent, Amazon at 13, and Netflix at 9. The fight is not only for ad dollars; it is for the operating system, the device, and the viewer relationship. Netflix has built its own in-house ad stack, the Netflix Ads Suite, to own its ad technology rather than rent it. The Trade Desk, the largest independent demand-side platform, is building Ventura, its own CTV operating system to keep an independent path into the living room. If a handful of platforms own the screen, the data, and the measurement, CTV starts to look less like the open programmatic market it joined and more like another walled garden, where the seller grades its own homework.
For a marketer the practical reading is steady. CTV is now a serious line in the plan, and it should be bought with the rigor of any programmatic channel: insist on transparency about where ads run, push for cross-platform frequency control, treat any single platform's self-reported numbers with doubt, and lean on attention or independent verification rather than viewability alone. The biggest screen in the house is finally addressable. It is not yet honest about what it delivers, and pretending otherwise is how budgets leak.
Council summary
The post argues that the real shift in television is not viewers moving to streaming but the living room becoming a programmatic channel, bought through the same auctions as a display banner. It traces that arc from Hulu's 2008 ad launch and the first Trade Desk and Roku CTV campaigns, through the supply flood when Netflix, Amazon, and Disney all turned on ads, to a present where US CTV upfront spending passes primetime linear. The honest core is the measurement gap: no shared currency, broken cross-platform frequency capping, and real supply-chain fraud, none of it solved. Every statistic was checked against the cited primary sources; the Pixalate fraud claim was made precise at 13 percent and both it and the frequency-capping case study were repointed to their correct sources. The takeaway is to buy CTV with programmatic rigor: demand placement transparency, push for cross-platform frequency control, and trust attention or independent verification over a platform's self-reported numbers.
Comments