Demand Gen Campaigns: The Complete 2026 Google Ads Guide

Demand gen campaigns are Google's visual-first campaign type, showing image and video ads across YouTube (including Shorts), Discover and Gmail. They replaced Discovery ads in 2024 and are the closest thing Google offers to social media advertising. This guide covers what they are, how they differ from Discovery, how lookalike segments work, which businesses they suit, and how to position them next to Performance Max and Search. Let's set expectations straight from the start: Demand Gen is not a search campaign; there is no keyword bidding, and success is not measured with the same ruler as Search metrics.

What is a Demand Gen campaign?

Demand Gen is a Google Ads campaign type that reaches people who are not actively searching but are likely to be interested in your product, across Google's discovery surfaces. The goal is not to capture existing demand but to create new demand: you enter the user's feed before they ever search for you. According to platform documentation, these surfaces reach more than 3 billion monthly users. Teams familiar with the discovery-feed logic of Meta and TikTok will find the transition natural: targeting is built on audiences and creative, not keyword lists.

• YouTube: in-stream, in-feed, Home and the Shorts feed (vertical 9:16 video).
• Discover: the personalized content feed in the Google app.
• Gmail: the Social and Promotions tabs.
• Google Display Network: optional extra reach.
• Maps: promoted pins rolling out gradually as of 2026.

Ads are distributed automatically across YouTube, Discover and Gmail, with the system shifting weight based on performance. Industry benchmarks show click costs vary clearly by placement: YouTube is typically the most expensive but most engaged surface, Discover is cheaper thanks to its huge inventory, and Gmail delivers low volume but strong conversion rates, especially on remarketing lists. These values swing in a wide band by country, industry and creative quality; in your own account, the first two weeks of placement reports will show which surface delivers cheap conversions, not just cheap clicks.

On the creative side, vertical 9:16 video is mandatory for the Shorts placement, and 15-30 second cuts typically perform best in practice. Testing at least 5 different assets per audience (a mix of image, carousel and video) is the platform's official recommendation. For vertical video craft, see our short-form video ads creative guide.

How is it different from Discovery ads?

Demand Gen is the direct successor to Discovery ads; all Discovery campaigns were automatically upgraded in March 2024. The change is more than a rename: it added video and Shorts placements, lookalike segments, placement controls and A/B testing, none of which Discovery had. Discovery only served images; Demand Gen combines image, video and carousel in a single campaign.

• Video support: Discovery was image-only; Demand Gen also serves video on YouTube and Shorts.
• Lookalike segments: audience expansion built from your own lists exists only in Demand Gen.
• Placement control: since 2025 you can pick channels and exclude individual YouTube channels and URLs.
• Bidding options: clicks, conversions and conversion value; in 2025 the primary strategy shifted from tCPA to tROAS.
• Reporting: placement-level and asset-level breakdowns, noticeably more transparent than PMax.

The migration itself was painless, but carrying over old Discovery habits is a mistake: accounts that settled for a single image in the Discovery era leave the strongest surface, YouTube inventory, unused unless they add video. The first job after the upgrade should be adding vertical video assets to existing campaigns.

How do lookalike segments work?

Lookalike segments find new users who resemble a seed list you provide (customer list, site visitors, YouTube engagers). You need a seed list with at least 100 active users, and reach comes in three tiers: narrow (2.5%), balanced (5%) and broad (10%). This feature exists only in Demand Gen; you cannot use it in Search, PMax or Display. The fresher the list and the more valuable the behavior behind it, the more accurate the similarity signal.

The quality of a lookalike is capped by the quality of its seed list. The most valuable seeds are not broad pools like 'all site visitors' but lists with proven value: purchasers, high-basket orders or registered users. In practice, starting with the narrow (2.5%) tier and widening once performance settles gives more controlled results than opening broad from day one.

How do you set up a Demand Gen campaign?

The skeleton of the setup fits in four steps: define the audience signal, prepare the creative set, choose the bid target, wire up measurement. The campaign needs enough data flow to exit learning quickly; in industry practice, starting budgets in the 75-150 USD per day range usually let the algorithm leave learning mode within 2-3 weeks. The most common setup mistake is starting with a single image and a single audience; that leaves the system no variation to optimize.

1. Connect conversion tracking and your product feed (if any); send conversion values for value-based bidding.
2. Build lookalike segments from your seed lists; support them with interest and custom segments.
3. Upload at least 5 assets per audience: a mix of vertical video, landscape and square images, and carousels.
4. Pick tCPA if you target conversion volume, tROAS if you target value; keep targets loose in the first weeks.
5. Review placement controls: exclude irrelevant YouTube channels and URLs.
6. Avoid major changes for 2-3 weeks; do not tighten targets before learning ends.

Which businesses is it right for?

Demand Gen works best for products that sell visually, categories open to discovery, and brands hunting for new customers. Products with low active search volume but high visual appeal (fashion, home decor, travel, beauty) are natural candidates. In practice three profiles stand out: The question is not only 'do I have the budget' but 'can I produce creative regularly'; Demand Gen is sensitive to creative fatigue much like social platforms.

• Ecommerce and D2C brands: product feeds enable catalog-based serving and tROAS targets.
• Brands saturating existing channels: when Search and PMax growth slows, it opens a new audience source.
• Long consideration cycles: travel, education and finance benefit from early-stage mindshare.

Conversely, it may not be the first priority for very narrow B2B niches, brands that cannot produce visual assets, or accounts whose daily budget cannot feed the learning period. In those cases, capturing existing intent with Search first and expanding to discovery surfaces later is the safer sequence. When deciding, look at your own funnel: if your Search impression share is already high and growth has stalled, Demand Gen's priority as a new customer source rises.

How does it fit next to PMax and Search?

The healthiest structure is a three-way division of labor: Search captures existing intent, Performance Max harvests conversions from high-signal users, and Demand Gen creates demand in audiences that are not searching for you yet. According to Google data, advertisers adding Demand Gen next to Search or PMax see 14% more conversions on average; industry observations report ROAS improvements in the 20-30% band for those running both types together. On the budget side, a typical starting point is allocating around 20% of total budget to the discovery layer, keeping the rest in the intent and conversion layers.

To avoid overlap, draw clear role boundaries: see our Performance Max optimization guide for controls that open PMax's black box, and our remarketing strategy guide to win back the audience Demand Gen warms up. Ads Sensor makes this picture manageable: it unifies your Google Ads data with Meta, TikTok and GA4 in one panel, flags channel-level risks with AI analysis, and after you join the beta it automatically tracks the before/after outcome of every recommendation you apply.

The relationship is not one-way either: users who have seen your Demand Gen ads typically convert at a higher rate when they later meet your Search ads. After launching Demand Gen, separately watch the trend of brand queries and conversion rate on the Search side; part of the growth is Demand Gen's invisible contribution.

How do you measure Demand Gen results?

Judging Demand Gen by last-click CPA alone is the most common mistake; a large share of the campaign's value happens without a click. Conversions from users who watched the ad and visited later, engaged-view conversions and lifts in brand search volume have to be read together. Keep the evaluation window wide too: decision cycles on discovery surfaces are longer than on search; a clear gap between the first 7 days' CPA and the 30-day window's CPA is normal.

• Engaged-view conversions (EVC): users who watched at least 10 seconds of the video and converted within a set window.
• View-through conversions (VTC): users who saw without clicking and returned later; in ecommerce this is often larger than click conversions.
• Brand search lift: compare brand query volume in campaign regions against the prior period.
• Billing note: from July 15, 2026, view-optimized campaigns on Discover are billed by CPM instead of CPC; reversible with a single setting.
• Comparison discipline: pick pre- and post-campaign periods of equal length and account for seasonality.

In practice the cleanest approach is to treat the 4-6 weeks before launching Demand Gen as your baseline, then track total conversions, brand search and overall efficiency together. Ads Sensor's AI campaign analysis makes that comparison easy: it automatically compares pre- and post-change periods, monitors anomalies 24/7 and presents the result in a client-ready monthly report mode. Look for answers to three questions in the report: did total conversions grow, did brand search rise, did blended efficiency (total revenue / total spend) hold? If all three are positive, the discovery layer is doing its job.