Lookalike audiences in Google Ads: what exists instead

Only in one corner of the platform. Google’s original lookalike feature, Similar Audiences, was retired in 2023 and no longer exists in Search, Display or Shopping campaigns. What remains is Lookalike segments, limited to Demand Gen and Video brand campaigns, plus automated tools that do the job without showing you a segment.

Lookalike Audiences is a Meta term. Google’s old version was called Similar Audiences, or similar segments, and it worked the same way: you gave it a seed list and it built a new audience of people who resembled them. When Google shut it down, most of that job moved into its automation.

So you can still reach people who resemble your customers. In most campaign types it happens automatically, and you steer it through the data you feed in.

Google removed Similar Audiences in 2023

Similar Audiences was deprecated on a published timeline. Google announced the sunset in November 2022 and set out the dates in its Google Ads Help notice on the similar audiences sunset:

• May 2023: new similar audience segments stopped being generated, and existing ones could no longer be added to campaigns and ad groups.
• August 2023: all remaining similar audience segments were removed from ad groups and campaigns.

Any tutorial that builds a “similar audience” from a remarketing list predates those changes.

Why Google retired them

Google framed the change around privacy. Its notice says advertisers need more durable strategies as observable data becomes less available, and it pointed people towards automated solutions that find relevant audiences without relying on third-party cookies.

Google folded most of the capability into its bidding and targeting automation instead. The modelling still happens, but in most campaign types you no longer see the segment, name it, or exclude it.

That causes real problems in practice. You used to be able to check a modelled segment and turn it off if it underperformed. Today that decision mostly belongs to the algorithm, and your main control is the quality of the data you feed it.

What Google uses instead

Several features now share the work Similar Audiences used to do on its own. None is a one-for-one swap, and which ones you get depends on your campaign type.

Lookalike segments in Demand Gen

This is the nearest thing to a real lookalike left in Google Ads. Google’s Lookalike segments start from a seed list of your own first-party data, such as a Customer Match list, website visitors or YouTube engagers, and find people who share their characteristics. You choose a reach level: narrow (roughly the most similar 2.5% of users in your target location), balanced (about 5%) or broad (about 10%).

They are changing, though. Through 2026 Google is moving Lookalike segments in Demand Gen from a hard targeting constraint to a suggestion mode, where the seed list acts as a signal and the system can reach people outside the chosen similarity band if it expects them to convert. Advertisers who want the original behaviour can keep Lookalike segments as a targeting constraint by opting out through a Google form linked from that help page.

Optimised targeting

Optimised targeting looks for new users likely to convert by modelling against your campaign’s conversion data. It is available in Display, Demand Gen and Video campaigns using the Sales, Leads or Website traffic goals, and it can take your audience segments and customer lists as signals.

Functionally it is the closest automated replacement. You do not create, review or target it directly. It is on by default, so your job is to feed it good conversion data.

Audience expansion

Audience expansion widens reach beyond the segments you selected, and applies to Video campaigns using the Brand awareness and reach or Product and brand consideration goals.

It is blunter than optimised targeting and built for reach rather than conversions, so treat it as a volume lever.

Customer Match as the seed list

Customer Match is where your first-party data enters the system, and it is now the most important input you control. Google’s Customer Match uses your uploaded customer data to reach those people across Search, the Shopping tab, Gmail, YouTube and Display.

Two requirements catch people out. To stay eligible, a list needs at least 100 members added or updated within the last 540 days. Memberships older than that drop out, so a list you uploaded once and forgot will stop working.

Customer Match lists feed Lookalike segments directly, and optimised targeting and Performance Max can use them as signals. That is the modern version of seeding a lookalike: you supply the seed, and the automation does what Similar Audiences used to do visibly.

Match rates are the part nobody warns you about. Google can only match people it can tie to a Google account using the details you uploaded, so the usable list is often much smaller than the file. Uploading email addresses, phone numbers and postal addresses together, rather than email alone, gives Google more ways to find each person.

Custom segments are the closest manual option

Outside Demand Gen, custom segments are the only audience type you still build by hand that behaves anything like prospecting. You define them by entering search terms people use, websites they browse, or apps they use, and Google finds people matching that behaviour.

The important difference is that they are not modelled on your customers. A lookalike starts from your data; a custom segment starts from your assumptions. That makes them useful when you know your market well and weak when you are guessing.

Performance Max audience signals

Performance Max accepts audience signals, which are suggestions rather than targeting. You tell Google which segments and customer lists represent your best buyers, and it uses them as a starting hint. Google states plainly that Performance Max may show ads to audiences outside your signals when they are likely to convert.

Audience signals behave like an old-style lookalike seed, but Performance Max does not create a lookalike segment you can view, report on, export or exclude.

How to build the closest thing to a lookalike today

If you run Demand Gen, build a Lookalike segment from your best seed list. Everywhere else there is no button for this, so the work shifts to feeding the automation properly. In practice that means four things:

• Upload real Customer Match lists, and keep them fresh. Segment them by value: a list of your top 20 percent of spenders teaches the system far more than everyone who ever bought.
• Make sure conversion tracking is accurate first. Optimised targeting models against your conversions, so if you are counting page views or form loads as conversions, it will find you more people who load forms. Fix the tracking before you blame the targeting.
• Give it enough conversion volume. An account producing a handful of conversions a month will not get much from any of these features. If volume is the constraint, count a meaningful mid-funnel action as a secondary conversion so the system has something to learn from.
• Use audience signals in Performance Max deliberately. Put your highest-value Customer Match list in as the signal rather than a broad interest segment, and resist changing it every fortnight.

None of that needs a new campaign type, just correct inputs. The same inputs lift results across the rest of the account, covered in how to get more from Google Ads.

If you are shifting budget across from Meta

The mental model does not transfer cleanly, and assuming it does is a common reason Meta advertisers find Google underwhelming at first.

On Meta you build the audience and the platform serves it. On Google you mostly supply the signals and the platform builds the audience privately. Demand Gen Lookalike segments are the familiar exception. You have less visibility and less manual control, and you make up for it with better data going in.

The other difference is intent. A Meta lookalike reaches people who resemble your customers whether or not they want anything right now.

Google’s search inventory reaches people already looking, so audience modelling matters far less there, and whether search will work for your business depends mostly on existing demand. Modelling matters most in Demand Gen, Display and YouTube, the placements closest to how Meta works. If lookalike-style prospecting is central to your plan, that is social media marketing work, and Google Ads is better used for capturing the demand it creates.

Run both, and let each do the part it is good at.

Frequently asked questions

Can you still see Similar Audiences in your Google Ads account? No. They were removed from all ad groups and campaigns in August 2023. Older reports may still show their past performance.

Is optimised targeting turned on by default? Yes. Google says it is turned on automatically in eligible campaigns, which surprises advertisers who assume they are only reaching their selected segments. You can switch it off in the ad group settings, so check it before you conclude your targeting is tighter than it is.