Google Demand Gen Ads for Ecommerce: 2026 Guide

Google Demand Gen ads for ecommerce give brands a visual-first channel across YouTube Shorts, Gmail, and Google Discover — reaching up to 3 billion users with intent-rich placements that neither Meta nor Performance Max can replicate alone. Set up correctly, Demand Gen acts as a full-funnel layer that fills the awareness gap between your organic content and your conversion campaigns, delivering 3x higher CTR versus standard display campaigns in ecommerce verticals.

What Are Google Demand Gen Campaigns (and How They Differ from PMax)

Google Demand Gen is a campaign type that runs visual and video ads across YouTube (in-stream, Shorts, in-feed), Gmail, and Google Discover. Google launched it in 2023 as the successor to Discovery Ads — but the upgrade is substantial. Where Discovery Ads were image-only and limited to Gmail and Discover, Demand Gen adds full YouTube inventory and short-form video support.

The distinction that matters for ecommerce operators: Demand Gen is a creative-first channel. You control exactly which creatives run, where they appear, and which audience segments see them. This is fundamentally different from Performance Max, where Google's automation controls placement, audience, and creative mixing across all Google inventory simultaneously.

That reporting transparency is underrated. With Demand Gen, you can see which specific placements, creatives, and audience segments are driving results — data that PMax actively obscures. For ecommerce brands that care about understanding why something works, not just whether it works, that visibility is a meaningful advantage.

Where Your Ads Appear: YouTube Shorts, Gmail, Google Discover

YouTube Shorts is the highest-growth placement in Demand Gen right now. Shorts hit 70 billion daily views globally in 2025, and Google's full Shorts inventory is available to Demand Gen campaigns — meaning your 9:16 vertical video can run between organic Shorts content. For ecommerce brands with product demo content or UGC-style creative, this is your lowest-CPM video inventory on the Google network.

YouTube In-Feed and In-Stream placements give you access to longer video formats. In-feed ads appear in search results and alongside watch pages — high intent, since viewers are actively looking for content. In-stream (skippable) ads interrupt viewing sessions and work better for storytelling-format creative: brand story, problem/solution, customer testimonial.

Google Discover reaches users scrolling their personalized Google feed on mobile — a high-passive-attention surface. Discover users aren't searching; they're browsing. This placement works well for evergreen product discovery content and lifestyle imagery that stops the scroll without triggering ad fatigue the way social feeds do.

Gmail placements appear in the Promotions and Social tabs as collapsed ads that expand when tapped. Gmail converts best for retargeting segments — users who've already visited your store — because the inbox context primes for purchase consideration in a way passive browsing doesn't.

60% of all ecommerce purchases happen on mobile (Digital Commerce 360, 2026), and every Demand Gen placement is mobile-first by design. If your creative isn't optimized for a 6-inch screen, performance will reflect it regardless of how well the campaign is structured.

Campaign Setup: Audiences, Creative Formats, and Bidding

Audience Segments

Start with three audience layers, in priority order:

1. Customer Match list — upload your customer email list (minimum 1,000 matched users). This becomes the seed for Lookalike expansion and signals Google's algorithm about your ideal buyer profile.
2. Lookalike segments — Google builds these from your Customer Match seed. Set expansion to "Narrow" initially to stay close to your existing customer profile; expand to "Broad" once you've validated ROAS at narrow.
3. In-Market + custom intent audiences — layer Google's in-market segments with custom intent audiences built from competitor product page URLs or category-level search queries.

Do not enable Optimized Targeting on a new Demand Gen campaign. Let it run with your defined audiences for at least 2–3 weeks before enabling it — otherwise Google expands reach before you have enough conversion data to guide the algorithm toward quality buyers.

Creative Formats

Each ad group should include all available formats for your target placements:

• 9:16 vertical video (6–60 seconds) — required for Shorts placement; this is your highest-priority format in 2026
• 16:9 horizontal video (skippable or non-skippable) — for in-stream and in-feed YouTube
• Square image (1:1, minimum 1200×1200) — for Discover and Gmail
• Landscape image (1.91:1, minimum 1200×628) — for Discover desktop
• Portrait image (4:5) — mobile-optimized Discover

Google requires a headline (up to 40 characters), description (up to 90 characters), and business name for all formats. Write at least 3 headlines and 3 descriptions per ad group — the algorithm rotates combinations and surfaces the top performers within 2–3 weeks.

Bidding Strategy

For new campaigns: start with Maximize Conversions (no target CPA). This gives the algorithm room to gather data across your audience segments without being constrained by a CPA floor you've potentially set too low based on other channels. After 30–50 conversions in the campaign — typically 2–4 weeks — switch to Target CPA or Target ROAS based on your blended economics.

The most common campaign setup mistake we see: setting tCPA on day one at a number derived from Search or Shopping, where conversion rates are higher because intent is higher. Demand Gen operates at the awareness-to-consideration stage; your initial CPA will naturally be higher until the algorithm finds the right audience mix. Give it room.

Demand Gen vs. Performance Max: How to Run Both Without Overlap

The most common objection: "Won't Demand Gen cannibalize Performance Max?" The short answer is no — if structured correctly. PMax runs across all Google inventory including Search and Shopping, where your highest-intent buyers live. Demand Gen runs exclusively on YouTube, Gmail, and Discover — surfaces where buyers are browsing, not searching. These are genuinely different funnel stages reaching buyers at different moments in their decision process.

The risk of overlap is real but manageable with three structural choices:

Brand exclusions: In your PMax campaigns, add brand terms as negative keywords where the interface allows. This prevents PMax from capturing credit for conversions that Demand Gen drove at the awareness stage — a misattribution that inflates PMax ROAS and makes Demand Gen look underperforming when it isn't.

Audience signal separation: In PMax, use your existing customer list as an "observation" signal (not targeting). In Demand Gen, actively target Lookalike segments built from that list. This keeps acquisition-focused spend in Demand Gen and high-intent/retention spend in PMax, with minimal audience overlap.

View-through conversion window: Set Demand Gen's view-through conversion window to 1 day — down from the default 30 days. A 30-day view-through window inflates Demand Gen's reported contribution significantly and makes budget allocation decisions harder to trust. One day keeps attribution honest.

These ratios assume your performance marketing strategy is optimized at the PMax level before adding Demand Gen — a leaky PMax campaign compounds the problem when you add upper-funnel spend.

Creative That Performs in Demand Gen

The creative format that consistently outperforms in Demand Gen ecommerce campaigns is the product-in-use vertical video — a 15–30 second clip showing a real person using or wearing the product, with on-screen text callouts for key benefits or offers, and a clear end card with CTA. UGC-style production (shot on mobile, natural lighting, creator-format) outperforms polished studio creative on Shorts by a wide margin because it matches the native content aesthetic.

Here's the framework that works:

• 0–3 seconds: Hook — show the product or the outcome immediately. No logos, no brand intros, no scene-setting. Start mid-action.
• 3–12 seconds: Context — who is this for, what problem does it solve, or what does it feel like to own it?
• 12–25 seconds: Proof — a quick result, a review snippet, or a before/after if applicable to your product category.
• 25–30 seconds: CTA — clear text overlay plus voiceover: "Shop now," "Get 20% off today," or the specific offer.

For image-only ads running on Discover: lifestyle imagery outperforms product-only shots in every category we've tested. Shoppers on Discover are in a passive, aspirational mindset — they respond to context and environment, not white-background product photos.

What consistently underperforms: text-heavy graphics, talking-head videos without captions (85% of mobile video is watched without sound), and generic stock photography. The algorithm can optimize placement and audience, but it can only work with what you give it — weak creative limits ceiling regardless of targeting precision.

For a broader look at how we structure Demand Gen campaign architecture, including audience layering in detail, see our prior guide. And for how Demand Gen fits into a full paid media stack alongside Meta, our performance marketing team can walk through a strategy specific to your brand's current channel mix.

FAQ: Google Demand Gen for Ecommerce

What's the minimum budget to run Google Demand Gen ads?

There's no official minimum, but we recommend at least $50–75/day per campaign to give the algorithm enough data to optimize. Campaigns running below $30/day take significantly longer to exit the learning phase, and tCPA-based bidding becomes unreliable at that level. For most ecommerce brands, a $1,500–2,500/month starting budget allows meaningful testing across 2–3 audience segments and creative sets before you have enough data to make informed scaling decisions.

Is Demand Gen better than Meta ads for ecommerce?

They serve different functions, and the comparison isn't straightforward. Meta's advantage is its social graph — it knows relationships, interests, and declared behaviors from direct user input. Google Demand Gen's advantage is intent proximity — users on YouTube and Discover are already in a research or buying mindset. Most ecommerce brands at scale run both: Meta for top-of-funnel acquisition and creative testing, Demand Gen for mid-funnel reinforcement among warm audiences and Lookalikes. The brands that pick one and ignore the other are leaving incremental reach on the table.

How long does it take for Demand Gen campaigns to optimize?

Expect a 2–4 week learning phase before performance stabilizes. Google's algorithm needs a minimum of 30–50 conversions to move out of learning mode when using conversion-based bidding. If your conversion volume is low (fewer than 10 purchases per week from Google), consider optimizing for a micro-conversion — add-to-cart or initiated checkout — rather than purchase. This gives the algorithm more signal to work with and accelerates the learning phase without compromising data quality.

Does Demand Gen work for lower-priced products?

Yes, with adjusted ROAS expectations. Lower AOV products (under $50) typically see higher conversion volumes but lower absolute ROAS from Demand Gen, since the awareness channel primes purchase rather than driving it directly. The real value for lower-AOV brands is driving first purchase at a sustainable CAC, then using email and SMS lifecycle sequences to recover margin through repeat purchases. Demand Gen feeds the top of that funnel; your retention stack closes the economics.

Can I use Demand Gen for retargeting?

Yes — and it's one of the most effective use cases. Upload your site visitor list segmented by behavior (product viewers, cart abandoners, checkout starters) as Customer Match audiences, then serve highly specific creative to each segment. A cart abandoner needs an urgency-forward message that's different from what a first-time visitor should see. Gmail placements work particularly well for retargeting because the inbox context creates a purchase consideration moment that passive browsing surfaces don't reliably trigger.