What is Demand Gen in Google Ads?

Demand Gen fills a specific gap in the account. Search campaigns respond to an intent the person has already expressed by typing something, and that starting point defines everything else: keywords, match types, copy that echoes the language of the query. Demand Gen starts from the opposite end. There is no query to interpret here, only an audience shown something while it is doing something else, whether that is watching a Short, going through the inbox or scrolling the Discover feed.

Hence the name. The campaign does not collect existing demand, it triggers it. That puts it closer to advertising on social networks than to the rest of the Google ecosystem, and it was framed that way from the start, as the natural destination for the visual assets a brand already produces for social platforms.

The campaign type replaced the former video action campaigns, which could no longer be created from 2025 onwards and were gradually moved into Demand Gen. With that move the inventory stopped being limited to YouTube and its video partners, and image formats entered territory that until then belonged to video alone.

Ads run on YouTube, including Shorts, the home feed, the watch next section and search within the platform, plus Discover, Gmail, Maps and the Google Display Network. Channel controls let you choose which of those surfaces the campaign appears on, so a brand can stay with YouTube and Discover if the inbox does not suit its message.

The assets are visual by definition. The campaign accepts images in landscape, square, portrait and logo ratios, video in various orientations, and a block of text with headlines, descriptions, business name, call to action and final URL. On top of that come carousel formats with several image cards and the option to connect a product feed from Merchant Center.

Targeting rests on audiences rather than queries. You can use Google affinity or in-market segments, your own customer lists, and lookalike segments built from a source list such as past buyers or site visitors. The available bidding strategies work on that basis, ranging from maximise clicks to target CPA and conversion value. The difference from Performance Max lies in how control is shared: Performance Max runs across the whole Google inventory including Search and decides on its own where to spend, whereas Demand Gen leaves out the search text ad and hands the advertiser back the choice of channel, creative and audience.

The point that separates Demand Gen from the rest has direct consequences for the results report. When someone searches for a product and clicks, the path from ad to sale is short and the last click explains almost everything. When someone watches a video on Shorts, does not react, and types the brand name into search three days later, that same model credits the sale to brand search and leaves the campaign that started it looking like a loss.

That is why the associated measurement tools are different ones. Google offers brand lift, search lift and conversion lift studies here, alongside A/B creative experiments and data-driven attribution. These answer a different question from the usual one, not how much this click converted but how far the behaviour of the exposed group shifted compared with the group that was not exposed.

For a small team the practical consequence is uncomfortable. Without a measurement plan settled before the campaign is switched on, Demand Gen gets judged by the yardstick of search, comes off badly in the comparison and is turned off after three weeks. The budget then returns to brand campaigns, which keep harvesting efficiently a demand that fewer and fewer people are creating.

• Upload image and video into the same campaign, with portrait, square and landscape ratios covered, so every surface finds a suitable format.
• Build lookalike segments on high-quality first-party lists, such as recent buyers, rather than on broad unfiltered visitor lists.
• Decide before launch which signal will validate the campaign, whether a lift study or the movement in brand searches, and write it down.
• Use channel controls to exclude surfaces that do not suit the creative, instead of accepting the default spread and later blaming the result on it.
• Treat the creative as the main variable and plan it into A/B experiments, because in this campaign type the message weighs more than bid fine-tuning.
• Give the campaign a learning window and a budget consistent with the target CPA, rather than adjusting it every couple of days.

• Measuring the campaign with last-click attribution and concluding it does not work, when the effect shows up later in brand searches and direct traffic.
• Reusing banners from an old display campaign unchanged, with small type and ratios that are useless in portrait.
• Confusing it with Performance Max and running both over the same audience with no criterion for splitting the budget.
• Aiming it at too narrow an audience, which removes the reason for choosing this campaign type in the first place, namely reaching people who do not know the brand yet.
• Sending the traffic to a landing page written for someone who is already searching for the product, without first explaining what it is.