.png)
If you open an account you have not touched since 2023 and your Discovery campaigns are gone, nobody deleted them. Google renamed the campaign type, moved it into a new build, and upgraded your campaigns whether you asked or not.
Demand Gen replaced Discovery campaigns. Google announced it at Marketing Live in May 2023, opened it to advertisers in beta through late 2023, and auto-upgraded every remaining Discovery campaign by March 2024. There is no Discovery campaign type left to create. That is the whole answer, and it is the part page-one results keep burying under three paragraphs of preamble.
The migration itself was uneventful, which is why so few people noticed it. Campaign IDs carried over. Historical performance data stayed attached to the campaign. Budgets, bid strategies, and existing image assets came along. Smart Bidding did not reset its learning because, as far as the auction was concerned, it was the same campaign with more inventory available.
I have inherited accounts where the only visible trace is a gap in change history and a campaign name that still says “Discovery – Prospecting – 2022.” If you are auditing an old account, that name is your clue, not your problem.
What changed deserves attention. The upgrade quietly widened where your money goes.
Discovery ran on three feeds:
Demand Gen added YouTube Shorts and in-stream placements, made video a first-class creative format alongside images, and introduced lookalike segments seeded from your own converter lists at three tunable widths.
Practically, an old Discovery campaign that migrated untouched can now buy vertical short-form inventory it was never built for. Its static images may be cropped for a horizontal feed. Nothing broke. It simply started spending in a place where its creative was not designed to compete, and the CPA drift appears slowly enough that most people blame seasonality.
Every Google rebrand is sold as an expansion, and this one mostly was. But two things got worse.
First, you cannot choose your surfaces. Discovery gave you a narrow, predictable set of feeds. Demand Gen gives you YouTube, Shorts, in-stream, Discover, and Gmail as a bundle. There is no checkbox for Discover only or for keeping budget out of Shorts. You get channel-level reporting after the fact and account-level placement exclusions. That is your lot.
Second, the creative bar rose. A campaign that once ran fine on three landscape images now competes against advertisers uploading vertical video. Google’s asset requirements push you toward 1.91:1, 1:1, and 4:5 crops, plus video if you want full delivery. Skip vertical assets and you are not excluded from Shorts. You are just badly dressed for it.
The other consolidation people miss is Video Action Campaigns. Google folded VACs into Demand Gen during 2025. If you ran YouTube conversion campaigns separately from Discovery, both line items now live inside one campaign type.
That is genuinely simpler. It also means one campaign type is doing three jobs that used to have three different playbooks, while the default settings can only be tuned for one of them.
Demand Gen exists because Google was losing mid-funnel budget to Meta and TikTok. It was losing it on the buying model, not the inventory.
Advertisers on paid social do not buy placements. They upload creative, define an audience, and let the platform decide where it lands. Discovery was still shaped like a Google campaign: pick feeds, write headlines, hope. Demand Gen is shaped like a social campaign, right down to lookalikes and a creative preview that shows the ad in-feed before launch.
Look across PMax, Demand Gen, and AI Max and the pattern is the same. Google removes a lever you used to pull manually and gives you reach plus reporting in return. Whether that is a good deal depends on whether you were pulling the lever well. Most accounts were not. Some were, and those are the accounts that got quietly worse.
One campaign. One conversion goal. One country. One ad group per audience idea.
The mistake I see most often is one ad group holding a lookalike segment, two custom segments, and a remarketing list. That makes the campaign impossible to read. When CPA moves, you cannot tell whether the audience or the creative caused it. Separate them and you can kill the loser in week two instead of guessing for a quarter.
Start with three ad groups:
Give each ad group its own creative. Ship a fourth only when one of the first three earns more budget.
Demand Gen rewards creative volume more honestly than any other Google campaign type because the ad is doing the persuading rather than the query.
Per ad group, I want at minimum:
That last bullet is where most Demand Gen budgets die. It has nothing to do with campaign settings. You spend real money teaching someone to want a product, they tap, and they land on a general category page that says nothing about the thing they just watched.
Search traffic tolerates that because the visitor arrived with intent. Feed traffic does not. If your Demand Gen conversion rate is half your Search rate, check the landing page before touching the bid strategy.
Demand Gen opens on Maximize clicks or Maximize conversions, depending on how you enter the flow. Both defaults are wrong for most advertisers.
Maximize clicks buys the cheapest possible taps from the cheapest possible inventory. On feed placements, that often means accidental thumbs. Start on Maximize conversions with no target for the first two weeks. Let it gather at least 30 to 50 conversions, then switch to Target CPA at roughly the CPA it actually achieved, not the CPA you wish it achieved.
Set a target 40% below reality and delivery collapses. You will stare at a campaign spending $12 a day and conclude Demand Gen does not work.
As a floor, I will not launch one with a daily budget under ten times my target CPA. Below that, the campaign cannot buy enough conversions per day to learn anything before the month ends.
Two more settings are worth changing before launch:
Then give it 14 days without touching it. That is the hardest part, and the one I still get wrong when a client is watching the dashboard hourly.
The short version, because I have written about this pairing at length in Demand Gen vs Performance Max: Demand Gen creates interest among people who were not looking for you. Performance Max harvests intent that already exists, including Search and Shopping inventory that Demand Gen cannot touch.
If your product solves a problem people know they have, start with PMax. If you are launching a category nobody searches for yet, start with Demand Gen and expect branded search volume to be the earliest signal that it is working.
Running both is normal and often right. Use brand exclusions in PMax so it stops taking credit for demand the other campaign built.
This is the part of the job I am glad to have handed off. At groas, the engine runs those levers continuously rather than on a review cadence. Copy and creative get generated and tested against each other. Budget shifts toward what converts. And dynamic landing pages deploy variants of your existing page matched to what the visitor was actually looking for, which is the single fix Demand Gen traffic needs most.
A senior strategist owns the account, and you get a written report every week of exactly what changed. The shape of the offer matters: below $25k a month in spend, it is fully managed and there is no dashboard to log into. That is either the appeal or the dealbreaker, depending on the person. Above that, you can have software access and run the day-to-day yourself with the engine underneath.
Whatever you use, the operating principle is the same one this rebrand was teaching. Google removed the levers that used to reward manual effort. The levers it left behind reward frequency.
Creative tested weekly beats creative tested quarterly by more than any bid adjustment you will ever make.
So find the campaign in your account with “Discovery” still in the name. Open its creative tab. If there is no vertical video there, you have been paying for Shorts inventory since March 2024 with assets built for a feed that does not exist anymore.
Fix that one thing this week. The rest of this article can wait.
