Demand Gen Google Ads 2026: the successor to Discovery, and how to run it

Ever since Google finally moved all Discovery campaigns over to Demand Gen, I get the same question almost every week in meetings: "Do we have to rebuild our campaigns? And does this TikTok-style visual thing actually work?". Short answer: yes, you need to get on it, and run well it's one of the best 2026 levers to create demand before a prospect ever types your brand into Google.

The trouble is that many advertisers treat Demand Gen like an old Display campaign in new clothes. It isn't. It's a discovery format, a hybrid between paid social and Search, that demands its own logic for creatives, audiences and measurement. Here's what I've taken away from my latest accounts.

Performance Max goes and finds the conversion wherever it sits. Demand Gen creates the desire where the prospect wasn't looking for you yet. Two missions, two ways of steering.

What Demand Gen actually changes versus Discovery

Discovery already ran on the YouTube feed, the Discover tab and Gmail. Demand Gen keeps that inventory and adds three things that change everything: video (in-stream and especially YouTube Shorts), finer audience controls (lookalike segments, exclusions, first-party audiences), and action-oriented bidding with clear goals (conversions, value, or clicks/views depending on the funnel stage).

In other words: you move from "image + text pushed to people probably interested" to a true engine for visual and video demand creation, able to play across the whole top and middle of the funnel. For a brand that wants to stop only capturing existing demand (Search), this is exactly the brick that was missing from a complete Google Ads strategy.

Demand Gen or Performance Max: don't confuse the two

It's the most common mix-up, and an expensive one. Here's how I decide with clients:

In practice the two don't compete: Performance Max harvests mature demand, Demand Gen fills the reservoir upstream. On a healthy e-commerce account I often run both, watching that Demand Gen doesn't just retarget people PMax would have converted anyway. For the broader debate between acquisition and capture, I laid out the frame in my article SEO vs SEA: combining both for maximum ROI.

Structuring a Demand Gen campaign that performs

1. Start from audiences, not keywords

Demand Gen has no queries: you don't target a typed intent, you target people. I always structure around three circles: first-party audiences (customers, visitors, subscribers: your data is gold here), lookalike segments (built on your best customers), and finally interest and in-market audiences to broaden. One circle per ad group, never all mixed together: that's the only way to read afterwards what actually works.

2. Think "social-first" creative

The classic error: recycling Display banners. Demand Gen lives on YouTube and Shorts, next to native content. Your creatives must look like that feed, not like an institutional ad. Concretely: short vertical 9:16 videos (6 to 15 seconds), a hook in the first second, captions (sound is often off), and several variants per angle. I aim for at least 3 to 5 video creatives plus a set of images per ad group, to let the algorithm arbitrate.

3. Pick the right bidding goal for the stage

Demand Gen serves several funnel stages, so the bidding goal must follow. For pure awareness, optimize for views or clicks. As soon as you want performance, switch to Maximize conversions then to a target CPA or target ROAS once enough data has accumulated. The trap: setting a target ROAS too early, before the campaign has learned. The algorithm throttles and delivery collapses.

Tracking: the real arbiter

Demand Gen is a discovery format: a large part of its impact is indirect (views, assisted visits, delayed conversions). If your measurement stops at last click, you'll undervalue the campaign and cut it too soon. Two non-negotiable prerequisites:

• A robust conversion setup, ideally server-side, so you don't lose signal with the death of third-party cookies and Consent Mode. I explain the why and how in my article on server-side tracking and the Conversions API.
• A reading in assisted conversions and engaged views, not just last click, to credit Demand Gen for its real role as a demand creator.

Without that analytics base, you're flying blind. That's the whole point of my Analytics & tracking approach: clean up conversion data before scaling budgets.

The mistakes that sink a Demand Gen campaign

• Migrating Discovery without rethinking creatives: you keep old static images and miss out on video, which is the heart of the format.
• Stacking audiences in a single ad group: you then can't tell who converts. One circle = one ad group.
• Judging on last click: you cut a campaign that was in fact assisting a large share of Search and brand conversions.
• Over-segmenting a small budget: below the learning threshold, nothing stabilizes.
• Forgetting exclusions: without excluding recent converters, Demand Gen retargets existing customers and artificially inflates ROAS.

Many of these traps overlap with the classics of Google media; I detailed them on the budget side in the 5 mistakes that sink a Google Ads budget, and on the automation side in my Performance Max guide.

Where to start, concretely

If you're coming from Discovery: don't just let Google migrate. Take one campaign, add 3 to 5 vertical video creatives, split your first-party audiences from your lookalikes, and wire up clean tracking before you scale. If you're starting from scratch: begin with a single campaign, a clear goal (often lead generation or product awareness), and let it learn two to three weeks before judging. In its day-to-day steering Demand Gen looks a lot like Meta Ads, and the Meta vs Google comparison helps you think through how the two ecosystems complement each other.

Run well, Demand Gen finally fills the gap between awareness and conversion inside Google Ads. It's not a gimmick: it's the piece that lets you stop only capturing existing demand, and start manufacturing it.