Google Demand Gen Campaigns: An Ecommerce Media Buying Guide

Your Shopping campaigns have held the same impression share for three months. Raising the budget now buys the same shoppers at a higher cost per click, and brand search volume has stopped growing. The next suggestion on the table is Demand Gen. Two weeks after launch the new campaign reports a ROAS of 9x, and the store's total revenue looks exactly as it did before. Both of those things can be true at the same time, and the gap between them is what this guide is about.

Short version: Demand Gen is Google's visual campaign type for creating demand. One campaign can serve image and video ads across YouTube (in-stream, in-feed and Shorts), the Discover feed, Gmail, Google Maps and, if you choose, the Google Display Network, and Google is now moving Display campaigns into it. Its appeal for ecommerce is control: unlike Performance Max, you choose the audiences, pick the channels for each ad group, and attach your product feed so ads carry live products. Its danger is reporting that flatters itself, because view-through credit and warm audiences make the channel look better than it is. This guide covers where Google Demand Gen campaigns serve, what changed in 2026, how they sit beside PMax, Search and YouTube, the creative your team needs to supply, how to measure the results honestly, and how to tell whether your account is ready.

The framing that keeps expectations sane: Demand Gen creates demand, as the name says. Nobody scrolling Discover or Shorts is searching for your product right now. You are paying to interrupt attention with something visual, which is a different job from capturing intent on Search and Shopping, and it should be judged by different math.

What Google Demand Gen campaigns are and where they serve

Demand Gen replaced Discovery campaigns and inherited their surfaces. Google has since upgraded Video action campaigns into it as well, so Demand Gen now covers most conversion-focused buying on YouTube in addition to the old Discovery inventory. According to Google's channel controls documentation, a single campaign can serve on:

• YouTube in-stream: skippable video ads before and during videos. Videos shorter than 10 seconds are not eligible here.
• YouTube in-feed: the home feed, search results and the watch next list.
• YouTube Shorts: vertical video and images between short clips.
• Discover: the content feed in the Google app.
• Gmail: image and video ads inside the inbox.
• Google Maps: which only serves when a location extension is enabled on the account, so most online-only stores will not appear there.
• Google Display Network: optional, and the home for Display campaigns as Google migrates them into Demand Gen.

That is a large amount of visual inventory from one campaign, and almost all of it is scrolling behaviour rather than search behaviour. The diagram shows the channels and where Demand Gen sits in a store's account.

Channel controls: choosing where each ad group runs

Channel controls are set at the ad group level. You can leave selection automatic, which Google recommends for most advertisers, or choose channels yourself. Choosing matters more than it sounds. An ad group built around vertical video can run on Shorts only. An image-led ad group can run on Discover and Gmail without the Display Network. For a first ecommerce test we usually start on Google's own channels and leave the Display Network out, then test it as a separate ad group once there is a baseline, because display inventory tends to bring cheaper clicks with less intent behind them and can blur the first read of the channel.

Product feeds and shoppable formats

Demand Gen accepts a product feed from Merchant Center, and Google lists product feed formats on YouTube, Gmail, Discover and the Display Network. Your image or video creative is paired with live product cards, prices and links, which turns a piece of brand creative into something closer to a browsable shelf. You choose which products appear through product scopes, so a test can run on best sellers rather than the whole catalogue. This is the feature that separates Demand Gen from a generic visual buy: the ad can end in products, not just a landing page. It also means the feed's titles and images show up in front of people who have never heard of you, which is one more reason to keep the feed clean.

What changed in Demand Gen in 2026

Google now publishes a monthly round of Demand Gen updates, and most months contain something small. A few changes this year matter to how a store should plan.

Display is moving into Demand Gen. Google began a phased rollout of a migration tool in June 2026. Its help documentation says new Display campaigns can only be created within Demand Gen, that remaining eligible Display campaigns will be migrated automatically, and that migration cannot be reversed. In practice, display prospecting and display remarketing will be built inside Demand Gen from here on, with the Display Network as one channel among several.

Lookalike segments are becoming suggestions. Google's documentation says the default behaviour of lookalike segments in Demand Gen will evolve toward a suggestion mode, where the segment guides targeting rather than fencing it. Advertisers who need the old behaviour can opt out through a form and keep lookalikes as a targeting constraint.

One-tap image ads on Shorts and Gmail. The September 2026 update lets people reach your landing page with a single tap from image ads on YouTube Shorts and Gmail. The same update introduced Business Agent for YouTube ads, a conversational assistant that answers product questions beside a video ad, offered through a sign-up form.

A push toward first-party data. The October 2026 update introduced a Data Strength Uplift metric in Google Ads and expanded Data Manager, Google's tool for connecting customer and conversion data. The direction is clear: Demand Gen performs better when it receives clean conversion data and customer lists, and Google is building tools to measure how much of that you send.

Google Demand Gen campaigns vs PMax vs YouTube video campaigns

The three campaign types overlap on surfaces, which causes most of the confusion, so the distinction is worth stating cleanly.

Against Performance Max, the difference is control. PMax has become less opaque this year: Google added first-party list exclusions, a budget report and placement reporting segmented by network. It still decides the channels, the audiences and the creative combinations. Demand Gen hands those levers back. You pick the audience segments, you see where ads served, you decide which creative runs where, and you can switch off what is not working. The trade is that control means responsibility: PMax will find something that works on its own, while Demand Gen performs exactly as well as your audience and creative decisions. The full comparison, including when each earns budget, is in our PMax vs Shopping vs Demand Gen breakdown.

Against dedicated YouTube video campaigns, the line has moved. Since Video action campaigns were folded into Demand Gen, conversion-focused video buying on YouTube runs through Demand Gen, in-stream included. Video campaigns with reach, view or awareness goals remain the tool when the job is getting a message seen by as many of the right people as possible at a low cost per thousand. If your plan is built around video storytelling and watch-time economics, our YouTube media buying guide covers that discipline. If your plan is visual product prospecting judged on sales, that is Demand Gen.

Audiences and lookalike segments in Demand Gen

Demand Gen's targeting toolkit is the strongest argument for it. You bring your own data: website visitors, app users, people who engaged with your YouTube channel, and customer lists built from your buyer file. Google adds its own segments: custom segments built from search terms and URLs, in-market, affinity and life events. On top of those sit the automated options: optimized targeting, a new customer acquisition goal, and lookalike segments, which Google builds from a seed list at narrow, balanced or broad reach.

Lookalikes are becoming signals, not fences

Lookalikes were retired almost everywhere else in Google Ads when similar audiences ended, so their presence in Demand Gen mattered: it was the closest Google came to the seed-audience prospecting that made paid social work. The move to suggestion mode changes how you read them. A lookalike ad group in suggestion mode can reach people outside the modelled segment, which may improve volume but makes the segment's results less pure. Treat lookalike performance as "seed plus Google's expansion" unless you have opted out, and judge it on new customers rather than on how tightly it matched the seed.

Seed quality, and keeping warm and cold apart

The quality of a lookalike is the quality of its seed. A seed list of your best customers, meaning recent, high-value and repeat buyers, produces a very different segment from a seed of everyone who ever bought once. Start lookalikes from your strongest list, not your longest one. Google itself recommends putting lookalike segments in their own ad groups, and the same logic applies to remarketing and customer lists: keep each in a separate ad group so you can read warm and cold performance separately rather than blended into one flattering number. Our guide to Google remarketing for ecommerce covers how to split those warm lists by intent and window.

Creative requirements: what your team needs to supply

We run media, not production, so this section is a supply list rather than a service description. Demand Gen performs in direct proportion to the assets behind it, and the campaign cannot be built well around gaps. These are the shapes and limits Google's asset specifications list for Demand Gen.

For images, your team should supply landscape, square and portrait versions of each concept, at the recommended sizes rather than the minimums, because different placements pull different shapes and a missing shape simply forfeits that inventory. Lifestyle imagery showing the product in use usually does better than flat product-on-white in feed environments, where the ad competes with organic content rather than with other ads. For video, horizontal and vertical cuts are both needed, with vertical essential for Shorts. Short lengths carry most of the load, but remember that anything under 10 seconds cannot serve in-stream. The rules that govern video elsewhere apply here too: the product visible early, understandable with the sound off, and more than one variation, because feed creative tires quickly and a single asset running for months decays in plain sight.

Brands with an active social presence usually have most of this already. Content made for Instagram and TikTok transfers well, because Demand Gen placements reward the same native, unpolished style. Brands with nothing beyond catalogue photography should plan creative production with their own team or partners before expecting much from the channel, because no amount of media buying makes up for assets people scroll past.

Is your account ready for Google Demand Gen campaigns?

Readiness comes down to a handful of things you can check in ten minutes: whether capture is already profitable, whether tracking is trustworthy, how many purchases the bidding can learn from, whether the budget fits your cost per purchase, and whether the creative library covers the shapes the channel uses. The checker below scores those eight questions and lists the fixes that matter most for your answers.

Two of the questions work as gates rather than points. Broken purchase tracking and unprofitable Search and Shopping campaigns each mean "not yet" regardless of the total, because a strong creative library cannot make up for a campaign optimizing toward the wrong data, or for demand that nothing in the account captures.

When to add Demand Gen, and the budget math

Demand Gen is a later-stage channel. The sequencing rule is the same one we apply to all demand creation: capture existing demand profitably before paying to create new demand. That means Search and Shopping running at acceptable returns, conversion tracking trusted, and creative assets already in hand. A brand that has not met those conditions will get more from fixing its foundations than from any prospecting channel, because demand created on top of a leaky funnel warms up competitors' customers at your expense.

On budget, the constraint is learning data. Google's guidance for target CPA bidding in Demand Gen is a daily budget of at least 10 times the target CPA. Many stores cannot fund that at launch, which is why a first test usually runs on Maximize conversions without a target, with spend concentrated rather than scattered: one campaign, two or three audience ad groups, and a daily budget sized so the campaign gathers purchases weekly rather than monthly.

As a hypothetical shape of the math: a store converting Demand Gen traffic at around one percent with a one dollar average CPC pays roughly 100 dollars per purchase. A 20 dollar daily test budget would collect a purchase every five days and never learn anything. The same store at 200 dollars a day, twice its cost per purchase, can gather around 60 purchases in a month if the cost holds, which is enough to make a real decision. Google's 10 times guidance would point to 1,000 dollars a day before adding a CPA target, which is a reason to set the target later, once there is a baseline. Run your own numbers through the ROAS calculator before committing, and size the test to reach a verdict, not to feel affordable.

A sensible allocation for a first test is 10 to 20 percent of total account spend, starting with remarketing and customer list audiences before opening lookalikes, and expanding only when the measured, click-based numbers earn it.

How to launch a first Demand Gen test

Once the checker and the budget math say go, this is the order we set a first test up in.

1. Confirm the foundations. Purchase tracking reconciled against store orders, Search and Shopping profitable, and the Merchant Center feed linked with few disapprovals.
2. Build one campaign on the purchase goal. Start on Maximize conversions without a target. Consider the new customer acquisition goal if winning new buyers is the point of the test.
3. Create two or three ad groups by audience. One for warm lists (site visitors and your customer list), one lookalike seeded from your best customers, and optionally one custom segment built from competitor or category searches.
4. Set channels deliberately. Start on Google's own channels, keep the Display Network for a later test, and give vertical video its own Shorts-focused ad group if you have enough of it.
5. Load the full creative set. Every image shape, horizontal and vertical video, your logo, and a product scope limited to best sellers so the feed shows products you can afford to sell at this price.
6. Write down the measurement plan before launch. Which columns you will read, the blended baseline you will compare against, how long the test runs (four to six weeks is typical), and the result that ends it.
7. Leave it alone while it learns. Avoid large budget, audience or creative changes for the first two weeks so the bidding works from stable data.

Measuring Demand Gen honestly

Demand Gen reporting flatters itself, and knowing this in advance is the difference between scaling a channel and scaling a reporting artefact. A large share of reported conversions can be view-through: someone saw the ad, did not click, and later bought. Video placements can also count engaged views, where someone watched part of the ad without clicking. Some of those people were influenced; many were going to buy anyway, especially the ones already on your remarketing lists. A Demand Gen ROAS that blends generous view-through credit with warm audiences can look spectacular while contributing very little extra revenue.

Then judge it the honest way. Read warm and cold audiences separately, because remarketing credit and genuine prospecting are different achievements. Watch blended MER, total revenue over total ad spend, in the weeks after Demand Gen budget arrives, because a channel that creates demand should eventually move the whole account, not just its own column. Watch branded search volume and new customer counts too, since those are where created demand shows up. Where the budget justifies it, a geographic holdout is the cleanest incrementality test available. Our piece on what a good ROAS actually is covers why the blended view outranks any single channel's self-reported return.

Common mistakes with Demand Gen

Running it before Search and Shopping work. Demand creation on top of weak capture is the most expensive sequencing error in ecommerce accounts.

Believing the blended view-through number. Split view-through and engaged views from clicks before drawing any conclusion, or the channel will grade its own homework.

One image, every placement. Missing shapes forfeit inventory, and a single creative tires within weeks in feed environments.

Assuming lookalikes still behave as hard targeting. In suggestion mode they can reach beyond the seed. Check for the signal tag before reading lookalike results as a clean audience test.

Seeding lookalikes from the whole customer file. The segment inherits the seed's quality; use best customers, not all customers.

Spreading a small budget across many audiences. Five ad groups sharing a thin budget all fail slowly; one audience funded properly reaches a verdict.

Treating it as a PMax replacement. They are different tools: PMax converts across everything with feed-driven automation, Demand Gen prospects with controlled audiences. Most accounts that run one still need the other.

Frequently asked questions

What is the difference between Demand Gen and Performance Max?

Control and job. PMax is a conversion-focused automation layer across all Google surfaces where Google picks audiences and channels, with your lists acting as signals. Demand Gen is a campaign for visual and video surfaces where you pick the audiences, choose channels per ad group, and control the creative pairing. PMax captures and converts; Demand Gen introduces. Accounts commonly run both with different expectations for each.

Where do Demand Gen ads show?

On YouTube (in-stream, in-feed and Shorts), the Discover feed, Gmail, Google Maps when a location extension is enabled, and optionally the Google Display Network. Channel controls let you choose these per ad group, or leave selection to Google. As Display campaigns migrate into Demand Gen, the Display Network becomes one of its channels rather than a separate campaign type.

Do Demand Gen campaigns work for ecommerce?

They work for the job they are built for: putting visual product creative in front of new audiences and warming them toward Search, Shopping and remarketing. With a Merchant Center feed attached, ads carry live product cards, which suits categories where people buy what they see. They do not work as a last-click ROAS machine, and brands that judge them that way will conclude they failed.

What budget do Demand Gen campaigns need?

Enough to gather purchase data at a useful pace, which depends on your cost per purchase rather than a universal figure. Google recommends a daily budget of at least 10 times your target CPA for target CPA bidding. Without a target, a practical floor is around twice your cost per purchase per day, in one campaign with few ad groups. A budget that produces one purchase every week or two never exits learning and never produces a verdict.

Do you create the images and videos for Demand Gen?

No. We are a media buying agency: campaign structure, audiences, feed integration, bidding and measurement. Your team supplies the creative, and content already made for social usually transfers well because the placements reward the same native style. We advise on the formats and shapes the buy needs, and we tell you plainly when the asset library is not ready for the channel.

Bringing it together

Demand Gen earns a place in ecommerce accounts that have outgrown pure capture: Search and Shopping profitable, tracking trusted, and a creative library people will stop scrolling for. Its advantages over PMax are visibility and control, its advantage over a plain display buy is the product feed and the audience toolkit, and with Display moving into it, it is becoming the default home for visual prospecting on Google. Its biggest risk is still its own reporting, which is why view-through discipline and blended MER decide the verdict rather than the campaign's own column. The comparison piece on PMax vs Shopping vs Demand Gen and the YouTube media buying guide cover the neighbouring decisions. If you want an outside read on whether your account is ready for demand creation, request a free audit and we will look at the foundations first.

About the author

This guide is written by Vasant Chaudhary, founder of Improtics, a Google Ads agency focused on ecommerce. He has more than five years of experience managing over 50 ecommerce and lead generation accounts across the US, UK and India. He focuses on media buying, audience strategy and measurement, the levers that decide whether demand creation adds revenue or just impressions. Get in touch or start with a free audit call.