CTV vs OTT: One Is a Screen, the Other Is a Delivery Method

OTT is how the video reaches the viewer: over the internet, bypassing cable, satellite and terrestrial broadcast. CTV is what they watch it on: a television screen connected to the internet, either built in or through a stick, box or console. They are not two options to choose between. One names a delivery method and the other names a device, which is why every CTV impression is also OTT, while a great deal of OTT viewing is not CTV at all because it happens on a phone, a tablet or a laptop.

That is the answer. What follows is why the distinction costs money when a media plan gets it wrong.

Why the two terms got tangled in the first place

The confusion is not your fault. It is inherited from the industry’s own definitions, which changed underneath everyone.

IAB Europe’s current framework separates the two cleanly, putting delivery method in one column (streaming, linear, cable, satellite, OTT) and device in another (CTV, mobile, desktop, traditional TV). IAB UK is equally explicit that CTV is a subset of OTT, and that OTT covers all devices and environments rather than being exclusive to the television screen (IAB UK, key CTV definitions).

The older American usage went the other way. The IAB’s 2018 digital video glossary described OTT device-first, as hardware that connects to a television to deliver internet-based video, which put OTT and CTV in the same conceptual box and left a decade of media kits using the words interchangeably. Both readings are still in circulation, which is why two vendors can quote you “OTT inventory” and mean genuinely different things.

The practical habit worth building: when a seller says OTT, ask which devices are included and in what proportion. When a seller says CTV, that should mean the big screen and nothing else. If they cannot answer in device percentages, they are reselling something they have not inspected.

The screen changes the ad, not just the price

A television is a lean-back, full-screen, often shared environment, and the ad formats reflect that. CTV spots are typically 15 or 30 seconds, run full screen, and are non-skippable. There is also no click, because a person holding a remote has no cursor. The only direct response mechanisms available are a QR code on screen and the phone already in the viewer’s hand.

Streaming on a phone or laptop is a different medium wearing the same content. It is a single-viewer session, the ad is frequently skippable, and it can carry a real clickthrough to a landing page. Everything you know about direct response applies there and almost none of it applies on the television.

Which means one creative cannot serve both, and the difference is not resolution. A vertical social cut, with burned-in captions and a logo in the corner, looks like a mistake on a 65-inch screen at four meters, and text sized for a phone is unreadable at that distance. Television creative needs to survive being watched by three people at once with the lights on, which is a production standard rather than an export setting. If the current asset library was built for feeds, budget for work made for the large screen before buying the media.

Blending the two is a pricing decision somebody else makes for you

Television-screen inventory carries a substantially higher CPM than the same content streamed to a phone, and the gap is wide enough to matter to any budget.

So a line item labeled OTT that quietly contains both is a problem. Whoever is optimizing it toward efficiency will drift toward the cheaper end, and you will receive a blended CPM that reads as a good CTV price while a meaningful share of the delivery happened on handsets. The reverse also happens: you pay for reach on the big screen and it lands on a laptop in a browser tab.

Two requirements fix most of it. Insist on device-type breakouts in reporting, so you can see what share of impressions actually reached a TV screen. And ask for app-level transparency, because a large amount of connected TV inventory is sold without naming the app it ran in. A supply chain you cannot inspect is one where you have to take the CPM on faith, and the tools for inspecting it, app-ads.txt and sellers.json, exist precisely because the problem is common.

Identity and frequency work differently on a television

There are no cookies on a TV. Targeting runs on the household IP address and on device-level advertising identifiers issued by the platform, which means the addressable unit is a household rather than a person. That is fine for a category bought by households and misleading for anything else.

Co-viewing compounds it. When three people watch one impression, the impression count and the human count diverge, and some platforms apply co-viewing multipliers to report a reach figure larger than the impressions delivered. That is a defensible modeling choice and it is still a model. Ask whether a reach number is measured or modeled, and if it is modeled, on what.

The bigger operational failure is frequency. Each app or publisher is effectively its own supply path, and a cap set in one place does not travel to another. The predictable result is a household that sees the identical spot far more often than any plan intended, on three different services, in one evening. Buy through fewer paths, ask specifically about cross-publisher frequency management, and check delivered frequency distributions rather than averages, because the average hides the households absorbing the damage.

You cannot last-click a television

Since there is no click, attribution on CTV is inferential by construction. What vendors offer instead is usually household matching plus a long view-through window, and that combination will happily claim credit for a purchase from anyone whose home saw the ad within thirty days. Treat those numbers as directional at absolute best. They are not lies, but they are not causal evidence either.

What does work is designed before launch rather than reported afterwards. Hold out matched geographic regions and compare total conversions, not campaign-attributed ones. Watch branded search volume in exposed versus unexposed markets, because the most reliable signature of a television campaign working is people going and looking you up, which is the connection between the living room and the search bar. Track direct and organic sessions in exposed geos over the same window.

The vendors know this is the weak point, and the interesting products are the ones attacking it directly rather than widening the view-through window, which is what the streaming platforms are now building toward. Judge any measurement offer by whether it could ever produce a negative result. If the methodology cannot show that a campaign did nothing, it is not measurement.

Where to start if this is your first buy

Decide which problem you are solving before choosing a label. If the goal is reach, demand creation and the credibility that comes with appearing on a television, you want CTV specifically, you should expect to pay for it, and you should judge it on branded search and incrementality rather than on any click. If the goal is cheap video impressions with a working clickthrough, you want streaming inventory across mobile and desktop, and calling it OTT is accurate.

Then set the measurement design before the campaign launches, hold back the control geos, insist on device and app breakouts in the contract rather than in a later email, and run long enough for a lagging signal to appear. A four-week test against a lower-funnel benchmark will conclude that television does not work, which is a conclusion about the test rather than the channel. The general case for the medium, and the mechanics of programmatic buying behind it, sit in the broader primer on connected TV advertising.

Everything above is a buying decision rather than a definitional one, and it is the reason the distinction matters. Where the plan already exists and the question is how it fits with everything else you run, that belongs in the paid media program rather than in a separate television budget nobody measures.

Frequently Asked Questions

Is CTV part of OTT, or the other way around?

CTV is a subset of OTT. OTT describes video delivered over the internet rather than through cable, satellite or terrestrial broadcast, on any device. CTV describes one device category within that: the internet-connected television screen, whether the connection is built into the set or supplied by a streaming stick, box or games console. Every CTV impression is an OTT impression. Not every OTT impression is CTV, because phones, tablets and computers account for a large share of streaming.

Does a YouTube ad watched on a television count as CTV?

Yes. The classification follows the screen, not the platform. YouTube viewed through a smart TV app or a connected device is CTV inventory, and the same YouTube ad served to a phone is not. This matters when comparing costs, because a single YouTube line item can span both and report one blended number. Ask for the device split before drawing conclusions about what your video is costing on television.

Can a viewer click a CTV ad?

Not in the way they can on any other digital channel, because a remote control has no cursor. The workarounds are QR codes displayed during the spot, which a viewer scans with the phone already beside them, and second-screen behavior where they simply search for you. Both are real and neither produces the clean click-to-conversion path that other channels report, which is why CTV measurement has to be built on incrementality rather than attribution.

Why are CTV CPMs higher than other streaming inventory?

Full-screen, non-skippable placement on the largest screen in the home, in premium content, watched by more than one person at a time and generally with attention that other formats do not command. Supply is also more constrained than on the open web. Whether the premium is worth paying depends entirely on whether you can measure the effect, which is why the measurement design matters more than the rate negotiation for a first buy.

How do we measure CTV without clicks?

Design a test rather than reading a report. Hold out matched geographic markets, run the campaign everywhere else, and compare total conversions, branded search volume and direct traffic between exposed and held-out regions over the same period. That gives you a causal read on what the spend produced. View-through conversions based on household IP matching over a long window should be treated as a directional signal, never as the primary evidence.