Demand Gen campaigns: when to use them (and when not to).

Demand Gen is the campaign type advertisers adopt for the wrong reason and abandon for the wrong reason. It gets switched on because a Google rep suggested it, judged on last-click CPA against Search after two weeks, and switched off as "expensive." That test was rigged from the start — not because Demand Gen is bad, but because it was asked to do a job it was never designed for. It is a mid-funnel machine: its job is to create and capture demand among people who were not searching for you yet.

What it actually is

Demand Gen (the successor to Discovery campaigns, upgraded with video) serves visually-led ads across YouTube — including Shorts, in-feed, and in-stream — plus the Discover feed and Gmail, from a single campaign with shared optimization. It supports image, carousel, and video assets, product feeds for retail, and, distinctively, lookalike segments built from your first-party lists — a mechanic familiar to anyone who ran social prospecting, and one that standard Google campaign types do not offer. Think of it as Google's social-style prospecting surface, pointed at three billion feed-scrolling users.

Demand Gen vs PMax vs standalone video

• →PMax is goal-first: it spans everything including Search and Shopping, chases the cheapest conversion wherever it lives, and gives limited surface control. Run it as your conversion workhorse.
• →Demand Gen is audience-first: you choose the surfaces and the audiences, and it optimizes within them. Run it to reach defined cold and warm audiences on visual inventory — with more control over where and to whom than PMax will ever give you.
• →Standalone video campaigns are format-first: maximum control over YouTube-specific settings, placements, and frequency. Run them when video isolation and placement-level reporting matter more than reach.
• →The three coexist cleanly: Search and Shopping capture demand, Demand Gen nudges and warms it, PMax arbitrages conversions across everything. Overlap is managed by exclusions and honest measurement.

When it earns its budget

The setups where we see Demand Gen perform: brands with strong visual creative and a real first-party list to seed lookalikes from; retail accounts pairing the product feed with lifestyle creative for a social-shopping experience on Google surfaces; and advertisers whose Meta prospecting works and who want the same motion on a second ecosystem at often-cheaper CPMs. The prerequisite in every case is creative — image and video assets in all the ratios (landscape, square, vertical), built to work without sound, refreshed on a cadence. Demand Gen with two static images is a rounding error, not a strategy.

How to judge it honestly

Give it its own success criteria before launch. Mid-funnel spend shows up as cheaper blended acquisition over 4-8 weeks — rising branded search, fuller remarketing pools, better Search conversion rates — not as day-seven last-click wins. Track cost per engaged new visitor and downstream conversion of Demand Gen-touched audiences, compare account-level CAC in weeks with and without the campaign live, and use the same incrementality thinking you would apply to any upper-funnel spend. And skip it entirely if you have not yet maxed out demand capture — if your Search impression share on money terms is 40%, buy that first; it is cheaper than manufacturing new demand.

Demand Gen is a good tool with a specific job. Give it real creative, real audiences, and mid-funnel success criteria, and it becomes the bridge between social-style prospecting and Google's intent machine. The performance-side YouTube playbook it pairs with is in our YouTube direct-response piece.