SIMILAR AUDIENCES ARE GONE: WHAT REPLACED THEM AND HOW THE NEW SYSTEM WORKS

For years, similar audiences (also called similar segments) were the default prospecting move: you took a remarketing list, Google built a segment of people who resembled it, and you targeted that segment to find new customers. So asking "how do I use similar audiences in 2026?" feels like a normal audience-targeting question. It isn't. The honest answer reframes it into a different one entirely, similar audiences are gone: what replaced them and how the new system works, because the feature no longer exists in any account.

Google stopped generating similar audiences on May 1, 2023, and removed them from every ad group and campaign by August 1, 2023. Campaigns that had relied on them were not left empty; they were automatically transitioned to optimized targeting or audience expansion. In other words, in 2026 the practical question is not "should I switch similar audiences on?" but "what is the system that quietly took their place, and how do I steer it?" This article is built around what I actually observed in accounts before, during, and after that transition, and around the inference you can draw from it.

What similar audiences were, and why people leaned on them

Similar audiences were Google-generated segments of users who resembled the members of one of your existing data lists, usually a remarketing list of website visitors or a customer list. You fed Google a seed audience, Google profiled the shared characteristics of that list, and it assembled a separate, targetable audience segment of "similar" people. You then added that segment to a campaign the same way you would add an in-market or affinity audience. The deprecation of this behaviour is documented directly by Google in its notice on changes to audience targeting.

The appeal: a tangible, controllable segment

From operational experience, the reason advertisers liked similar audiences was rarely raw performance. It was tangibility. The segment was a thing you could see in the audience manager, add, remove, and report on at the line-item level. You knew which list it was derived from, you could read the approximate size, and you could compare its performance against your other audiences. For prospecting on Display and video, that felt like a controllable extension of your best first-party data: people "like" your converters, an approximation of your ideal customer, packaged into a row you could manage and bolt onto your existing user base.

The trade-off was that a similar audience was essentially static. It resembled the seed list, not your live conversions. If your real buyers drifted away from the profile of that old remarketing list, the segment kept chasing the list anyway. It optimized to resemblance, not to outcome.

What they never were

It's worth being precise, because the language around similar audiences oversold them. They never optimized to a conversion goal in real time. They were a snapshot of "who looks like this list", refreshed periodically, and then targeted manually. The moment your account had enough conversion volume to let Google model your actual converters, a list-resemblance segment was already the weaker tool, a fact Google's own direction hinted at long before the sunset by pushing advertisers toward conversion-based automation.

The 2023 deprecation: what actually changed

The single most important fact for 2026 is blunt: similar audiences no longer exist anywhere in Google Ads. Google's timeline was explicit, generation stopped on May 1, 2023, the segments were removed from all campaigns and ad groups by August 1, 2023, and the official reason given was the shift away from third-party signals toward privacy-durable, first-party data solutions. The label lingered in some reports for a short while, which created confusion, but the underlying segments were gone.

What "automatically transitioned" means in practice

From operational experience: the accounts I watched through the transition did not get a dramatic crash, because Google did not simply delete the targeting and walk away. Display, Demand Gen and Video action campaigns that had used similar audiences were opted into optimized targeting; awareness and consideration video campaigns were opted into audience expansion. The seed lists you had originally used as the basis for similar segments kept working, but now as signals feeding a conversion model rather than as the recipe for a static lookalike segment.

The inference here matters more than the event. Many accounts had never really managed their prospecting audiences; they had let similar audiences do the resembling and then left it alone. When the transition flipped them onto optimized targeting, the system started expanding well beyond the old segment boundaries, and accounts that had treated similar audiences as a tidy, fixed line item suddenly saw their reach and their audience reporting change shape. That, not the loss of the feature itself, was the real source of the post-migration wobble.

What Google tells you to use instead

The official guidance is unambiguous about the replacements. For Display, Demand Gen and Video action campaigns, Google points to optimized targeting with first-party audiences. For awareness and consideration video, it points to audience expansion. For Search and Shopping, it points to Smart Bidding, which leans on your first-party data automatically. The mechanics of the main replacement are laid out in Google's reference on About optimized targeting.

What replaced similar audiences in 2026

Similar audiences were not replaced by a single feature but by three overlapping mechanisms, plus Smart Bidding absorbing part of the job. Understanding which one took over which campaign type is the key to steering the new system instead of fighting it.

Optimized targeting: the conversion-driven heir

Optimized targeting is the closest functional replacement for prospecting on Display, Demand Gen and Video action campaigns. Instead of building a segment that resembles a list, it profiles people who resemble your actual converters, using real-time conversion data, and then expands beyond your selected audiences to reach them. Your chosen audiences and keywords become a starting signal, not a fence; if the system finds better-performing users outside them, it goes there. This is why, in my experience, it tends to find higher-intent users than a static similar segment ever did, the model is chasing the outcome, not the resemblance. It is conversion-driven rather than purely audience based, which is what lets it surface potential customers you would never have reached by targeting a fixed list.

Audience expansion: the reach-focused heir

For awareness and consideration video campaigns, the role passed to audience expansion. The distinction matters and is constantly confused: audience expansion broadens your existing audience to similar users while staying tethered to that audience framework, optimizing for reach and engagement. Optimized targeting, by contrast, is free to leave your audience framework entirely in pursuit of conversions. One expands the audience; the other expands the conversions.

Lookalike segments: the most direct descendant

The feature that most resembles old similar audiences conceptually is Lookalike segments, available in Demand Gen campaigns. You provide a first-party seed list as the source audience, and Google builds a lookalike audience of users who share characteristics with it, with a reach slider for similarity. Through 2026, though, Google is changing how rigid that is: Lookalike segments are moving from a hard similarity constraint toward a suggestion the system can exceed when it predicts conversions. That phased change is documented in Google's page on Use Lookalike segments to grow your audience.

How the new system actually works

The mental model that helps most is this: similar audiences answered "who looks like this list?", while the system that replaced them answers "who behaves like the people converting right now?". That shift, from list-resemblance to outcome-resemblance, explains almost every difference you feel in the account.

The inference from the transition

From operational experience: the accounts that came through the 2023 transition best were not the ones that tried to recreate a tidy lookalike segment they could micromanage. They were the ones that leaned into the conversion signal, fed the system clean first-party data through customer lists, and let optimized targeting expand against a well-defined conversion. Once an account had a strong, accurate conversion action and a few weeks of data, optimized targeting consistently surfaced higher-intent users than the old similar audiences ever did, because it was profiling converters, not a stale remarketing list. The conversion rates on the expanded reach held up, and the ad spend that used to leak into list-resemblance lookalikes was redirected toward people actually likely to convert.

The accounts that struggled were those that treated the new system like the old one, expecting a fixed segment they could inspect and bid on as a discrete row. In practice the audience reporting is now directional, not the crisp line item it used to be, and trying to "lock" the targeting back down fought the very mechanism that made it work. The control was, again, an illusion of activity rather than a source of performance.

Where the seed list still matters

Crucially, your first-party data did not lose its value, it changed role. The customer lists and remarketing lists that used to seed similar audiences are now the strongest signals you can hand the new system. On Lookalike segments they define the seed; on optimized targeting they sharpen the profile of who is worth expanding toward. The better and fresher your list, the better the model behaves. In a privacy-first Google Ads, the seed list is more important than ever, even though you no longer target a visible "similar" segment built from it.

The bottom line

In 2026, similar audiences are a feature you read about rather than use, which is exactly why framing it as similar audiences are gone: what replaced them and how the new system works is more useful than asking how to switch them on. For prospecting, the answer is optimized targeting; for reach video, audience expansion; for a direct lookalike in Demand Gen, Lookalike segments. Stop trying to rebuild the old static segment, feed the new system strong first-party data, define your conversions cleanly, and let it expand toward outcomes. The question is not how to bring similar audiences back, it's whether you've adapted to the conversion-driven system that replaced them.