How similar audiences worked
Google took an existing list of yours — site visitors, a customer list upload, app users — and looked for other people whose behaviour resembled the members of that list. The result was a second, larger segment called a similar audience, or in later interfaces a similar segment. It sat beside the original in the audience manager, refreshed itself as the seed changed, and could be added to campaigns like any other audience.
The mechanism was never visible. Google did not say which traits it matched on, and there was no dial to make the resemblance tighter or looser. A seed list that was small, stale or vaguely defined produced a similar audience that inherited every one of those faults, without ever saying so.
Why similar audiences mattered
For several years they were the standard answer inside Google Ads to the question every advertiser eventually asks: how do I find more people like the ones who already buy from me. They were the nearest equivalent to Meta’s lookalike audiences, and almost every account of any size used them somewhere.
They matter now for a different reason. Google has retired them. The capability was folded into systems that decide reach for themselves rather than handing you a list: Optimized targeting on Display and Demand Gen, audience expansion on video campaigns, and Smart Bidding on Search, where your own data now acts as a signal rather than a boundary. There is nothing left to switch on.
Common mistakes with similar audiences
The live mistake is repeating advice that no longer applies. Articles, agency proposals and training courses still list “build similar audiences” as a step. If you meet that instruction in a proposal, it tells you something useful about the proposal: it was assembled from older material rather than written for your account.
The second is mixing up the platforms. Meta’s lookalike audiences still exist and still work broadly the way similar audiences did, so the same sentence is correct on one platform and impossible on the other. That is precisely why the confusion has lasted.
The third is quieter. Many accounts still carry segments, exclusions and naming conventions built around the retired feature. They serve nobody, but they crowd the audience manager and make it harder to see what the account is really targeting.
What to do instead
Give the automated systems the seed you would have given the old feature. A customer list, a converters list, or visitors who reached a high-intent page, attached to the campaign as your own data, is the same starting point expressed differently. On Display and Demand Gen the expansion happens through Optimized targeting; on Search it happens through bidding, which means the quality of your conversion tracking now does the work the seed list used to do.
Then tidy up. Open the audience manager, remove segments that no longer serve, and check that no campaign is relying on a name that implies a targeting behaviour it no longer has. Where you genuinely want a lookalike you can control and size yourself, that tool lives on Meta, not on Google, and the two should be planned separately rather than described as one tactic.
