Shoppable CTV in Demand Gen 2026: Selling From the Living Room Screen
The television became a clickable channel again — and in 2026 it is no longer a brand-budget experiment but a setting inside a Demand Gen campaign. Shoppable CTV in Demand Gen renders a product carousel from your Merchant Center feed on top of the video, and the viewer scrolls it with a remote. This guide covers what actually shipped in 2026, what you must fix before launch, how to measure a channel that produces almost no clicks, and what budget makes the test meaningful instead of decorative.
What shoppable CTV in Demand Gen is, and how it differs from ordinary video
CTV — connected TV — means impressions on smart TVs, streaming sticks (Roku, Fire TV, Apple TV), game consoles and the YouTube app on the living room screen. Historically that inventory sat at the top of the funnel: reach, frequency, awareness. Attribution was weak for a boring reason — there is nothing to click on a television. A remote is not a mouse.
Shoppable CTV changes the unit of advertising. Alongside the video, the ad shows products pulled from a connected feed: image, title, price. The viewer moves between cards with the remote, opens an item, sends it to their phone or saves it for later. Most purchases finish on a second device, which is exactly why measurement here is built on “TV impression → visit from another device”, not on a click inside the ad.
The mechanic is different from a standard video campaign in one important way: the inventory is the same, but the ad becomes a catalogue. It does not replace search performance and it does not make Performance Max work redundant. It captures the slice of demand that previously watched your ad and walked away.
How this differs from buying TV media directly
Traditional TV is sold in packages: you buy airtime, pay against an audience forecast, and receive ratings weeks later. CTV inside Google Ads keeps the logic of an ad account — auction, daily budget, audience signals, frequency caps, same-day reporting, and an off switch. The entry price is different too: not a five-figure package but a daily budget you can set this afternoon.
What you give up is control over context. You do not pick the show or guarantee prime time; you work with audience signals and exclusions instead. For a performance team that trade is usually acceptable — TV inventory becomes one more testable source governed by the same rules as everything else, including placement exclusions if your brand is sensitive about adjacency.
What actually shipped in Demand Gen in 2026
Three capabilities moved from beta to general availability in January 2026, and together they change how a video budget gets planned.
1. Shoppable CTV
Connect a product feed to a Demand Gen campaign and the carousel renders on television. Google reports roughly 7% incremental conversions at flat return from adding the TV placement. Treat that as a platform benchmark, not a promise: your result depends on average order value, purchase cycle length and feed quality. You still need your own experiment.
2. Attributed branded searches
This metric counts people who searched for your brand after being exposed to a Demand Gen ad. The window is 30 days, view-through, data is aggregated rather than user-level, and it surfaces as a conversion action in reporting. Activation typically runs through a Google representative and requires brand mapping — you define which queries count as branded.
It is a direct answer to the oldest upper-funnel argument in paid media. Previously you proved the halo by hand, watching branded volume drift in the search terms report and Search Console. Now the lift shows up inside the account.
3. Travel feeds
Hotel Center feeds connect to Demand Gen, and the system assembles video ads from property photos, names, prices, ratings and availability. Rates change in the feed, the creative updates itself — no re-export every time a room price moves.
| Capability | Prerequisite | Best fit |
|---|---|---|
| Shoppable CTV | Merchant Center product feed plus video assets | Retail and e-commerce with a decent average order value |
| Attributed branded searches | Rep activation and brand mapping | Anyone funding upper funnel who cannot defend it |
| Travel feeds | Hotel Center feed | Hotels, OTAs, vacation rentals |
What to fix before you launch
- The feed. The carousel shows image, title and price, which means every weakness in your feed is about to appear on a 55-inch screen. Titles must read without page context, images must be free of collages and watermarks, prices must match the site. If product disapprovals in Merchant Center are a recurring problem for you, fix the feed first and launch TV second.
- Feed segmentation. Do not hand the whole catalogue to television. Build a set of items with real margin and stable stock using custom labels in your product feed — the on-screen carousel is short and every card has to earn its slot.
- Video assets. Horizontal 16:9, 1080p or better, large legible type, and sound treated as a real channel of meaning: unlike feeds, TV audio is almost always on. Fifteen to thirty seconds works well because the viewer cannot flick past you the way they do in a feed.
- Conversion tracking. Purchases land on a different device, so you need a server-side setup or at minimum enhanced conversions. Without a solid signal the campaign optimizes blind.
- Frequency expectations. TV inventory over-serves a household faster than you expect. Set caps and watch them — the mechanics are in the guide to frequency capping and Target Frequency.
Targeting: who should even see you on a television
Demand Gen runs on audience signals rather than keywords, and on TV that difference bites. You are buying attention, not intent. Three consequences follow.
One. The most predictable results come from audiences that already know you: site visitors from the last 30–90 days, cart abandoners, customer lists via Customer Match. That is closer to “reminder on a big screen” than to conquest, and it is usually where the test first pays for itself. Building those segments without strangling reach is covered in the piece on custom segments and intent audiences.
Two. A household is not a person. One television serves a family, so demographic precision drops and frequency counts per device. Do not be surprised when a large share of impressions lands in “unknown” demographics.
Three. Optimized targeting expands beyond your lists by default. On TV that can eat budget before you have judged quality, so decide on the setting before launch — see controlling audience expansion.
Building a 15–30 second spot that works on a couch
A TV viewer generally is not scrolling and cannot skip you the way a feed user can. That buys you real seconds of attention and punishes an empty opening: three seconds of a logo on white is three seconds of paid screensaver.
| Seconds | Job | What is on screen |
|---|---|---|
| 0–3 | Name the subject | The product in use, or the problem scene. Not your founding year. |
| 3–10 | Give a reason | One benefit in the customer’s own words, product in close-up |
| 10–20 | Remove the objection | Price, guarantee, delivery time — whatever stalls the decision |
| 20–30 | Point somewhere | Brand plus a short prompt, timed with the product carousel |
Practical rules people keep forgetting: shoot 16:9 with no pillar-boxing; keep type at least one twelfth of frame height; check contrast under evening light; write audio as a standalone message rather than background music. Fine-print disclaimers are unreadable at three metres — if they are mandatory, put them in the voiceover. Testing several cuts and killing the weak ones is covered in the guide to video ads built for performance.
Measuring a channel with almost no clicks
This is the reason TV budgets die inside performance teams. Break measurement into layers.
Layer 1: what the account now counts by itself
Attributed branded searches close part of the gap — you can see that exposure was followed by brand search. Remember it is view-through: strong as a directional indicator, weak as a basis for ROI math.
Layer 2: lift studies
Brand Lift and Search Lift compare exposed and control groups, which moves you closer to causality. Setup and minimum thresholds are in the article on Brand Lift and Search Lift on YouTube.
Layer 3: geo experiments
The honest answer to “what if we had not run TV at all”. Split regions into test and control, run CTV only in test, and compare total conversions across all channels rather than the campaign’s own numbers. The method is in the breakdown of incrementality and geo experiments. For CTV this layer is effectively mandatory.
The fewer clicks a channel produces, the more your conclusion depends on experiment design. On CTV you cannot read ROAS in the interface and call it a result — half the effect leaves through other channels and other devices.
Layer 4: reconciling with real revenue
If your sales close in a CRM, wire up offline conversion import. Otherwise the algorithm optimizes toward form fills while you are trying to buy revenue.
Budget and a six-week test plan
The platform’s own guidance for a meaningful test is around $50–100 per day per campaign over four to six weeks, roughly $1,500–5,000 in total. Treat that as a floor for data volume, not as advice: with a $30 average order value and a 1% conversion rate, $50 a day buys statistics slower than your patience lasts. Work back from your allowable CPA — the method is in the piece on media buying unit economics.
- Week 0. Clean the feed, isolate 50–300 products, produce three or four video cuts, verify conversion tracking end to end.
- Week 1. Launch Demand Gen with TV inventory and the feed attached. Set a frequency cap and then leave the campaign alone for seven days — edits during learning reset it.
- Weeks 2–3. Check how impressions split between TV and other placements, prune obvious waste, review view rates by asset. Do not optimize toward in-platform CPA yet.
- Weeks 3–6. Hold the geo experiment. Watch branded search volume in parallel — rising brand queries are the earliest sign the channel is doing something.
- After the test. Compute incremental CPA from the test-versus-control delta, not from reported conversions. If it clears your threshold, scale budget in 20–30% weekly steps.
The five reports to look at every week
- Device and placement split. You need to know what share of spend actually reached televisions. If TV took 5% of impressions, there is no “CTV effect” to discuss.
- Frequency per unique device. The comfortable range on TV tends to be lower than in feeds; past roughly four to six weekly exposures the awareness curve flattens while annoyance does not. Directional number — read your own curve.
- View rates by asset. Creative differences show up on TV faster than conversion differences.
- Branded search and direct traffic. Look beyond attributed branded searches to overall branded query volume; a lightweight dashboard beats manual exports, as described in the guide to Google Ads data in BigQuery.
- Test versus control summary. The only table that settles the argument: total conversions and revenue across all channels, by region group.
One rule about period comparisons: never benchmark a TV test against last month if budgets, seasonality or promotions moved at the same time. Naive before-and-after analysis misleads more often on CTV than anywhere else, because the channel works with delay by design.
Mistakes that kill a CTV test
- Waiting for clicks. CTR on TV is low and that is normal. If your stakeholder grades the channel on CTR, the test dies in week two.
- Shipping the whole catalogue. The carousel is short. Zero-margin or out-of-stock items simply steal slots.
- Mixing everything into one campaign. TV-heavy Demand Gen and feed-heavy Demand Gen answer different questions. Separate them or you will never know what worked.
- Ignoring cannibalization. Rising branded search after TV has to be read together with your brand campaigns — part of those “new” conversions is you buying yourself. The arithmetic is in brand defense in search.
- Reading Merchant Center numbers as if nothing changed. Since late August 2026 feed performance is counted differently and traffic moved between rows — see Merchant Center cross-platform reporting. Compare periods without accounting for that and you will credit TV with someone else’s lift.
- Weak audio and tiny type. People watch from three metres. Anything that only reads on a laptop mockup does not read from a sofa.
When CTV is simply too early
The format is not universal. It will almost certainly lose money if you have no product feed, if search demand is still unexhausted, if conversion tracking is unreliable, or if you cannot leave a test running for six weeks without interference. Fix the bottom of the funnel first: clean search terms, correct conversions, a landing page that holds. TV amplifies a working machine; it does not build one.
Teams assembling the infrastructure for launches like this can look at the PPC Rebels Google Ads agency accounts, and the running library of account-level breakdowns lives in the PPC Rebels paid traffic blog. If trademark rules are part of your creative review — and on TV they usually are — the companion piece on Google Ads trademark policy covers what you may and may not say about other brands.
FAQ: shoppable CTV and Demand Gen on television
How is shoppable CTV different from a regular YouTube video campaign?
The inventory overlaps, but the ad carries a product carousel from your Merchant Center feed that the viewer navigates with a remote. A standard video campaign offers a clip and a link that is awkward to click on a TV.
Do I need Merchant Center for shoppable CTV?
Yes. The product cards come from a connected feed. Without it you are running an ordinary video format with no catalogue.
Which devices show the format?
Smart TVs, streaming devices such as Roku and Fire TV, and game consoles — anywhere the YouTube TV app runs.
What is the minimum budget for a real test?
The platform benchmark is $50–100 per day for four to six weeks. Your actual number depends on allowable CPA and conversion frequency; at a low order value that budget will not produce statistics in time.
Can I trust attributed branded searches?
Use it as a directional signal. It counts branded queries within 30 days of a view, which makes it correlational — solid for trend reading, not for ROI claims without an experiment.
How do I separate TV’s effect from other channels?
Run a geo experiment: enable CTV only in test regions and compare total conversions with control regions. Everything else is circumstantial.
How many products should I expose to CTV?
Usually a narrow set — items with margin, stable inventory and strong photography. Dozens to a couple hundred SKUs is typical; beyond that the carousel never gets to them.
Does CTV work for services or B2B?
The product carousel does not, since it needs a feed. Plain CTV awareness can work, but a long B2B cycle has to be measured with lift and incrementality rather than in-platform CPA.
Do I need TV-specific creative?
Strongly recommended. Vertical feed cuts look wrong on a large screen, small type disappears, and TV audio is almost always on — the message has to survive being heard rather than read.
Where do travel feeds fit?
They are the second feed type for Demand Gen: instead of products, the ad assembles properties from Hotel Center with price, rating and availability, updating without manual creative rebuilds.
What comes first, CTV or Performance Max?
If search demand is not yet exhausted, search and PMax come first. CTV earns its place when the lower funnel is saturated and you need to create demand.
How do I know a test failed rather than “still learning”?
If after four to six weeks test regions show no lift in total conversions or branded search versus control, the channel did not work with this offer and this creative. Change the offer and the spot before you touch the bid.
