Skip to content
Anderson Collaborative

Home › Knowledge Base ›Lookalike Audiences: Seeds, Setup, and Measurement

Lookalike Audiences: Seeds, Setup, and Measurement

A lookalike audience helps an advertising platform find new prospects who resemble a customer or visitor group you already know. Strong results start with a seed that represents the outcome you want more of.

Updated September 20, 2026· 8 min read

A lookalike audience is a prospecting list built from patterns in a seed audience you already have. The seed might be purchasers, qualified leads, high-value customers, or another first-party list. The platform then looks for new people with similar signals.

Lookalike work sits next to audience segmentation and paid social planning. A clear conversion event and a well-defined source give the model a useful outcome to learn from.

Seed, lookalike, retargeting, and expansion

ObjectWho sees the adsTypical job
Seed / sourceUsually excluded from the lookalikeThe people whose patterns the model should copy
LookalikeNew people similar to the seedProspecting that stays tied to a named source
RetargetingPeople already on a custom or remarketing listReturn visits, abandoned carts, known customers
Automated expansionPeople inside and, often, outside the selected listDelivery that uses the list as a hint, not a fence

Meta’s create a lookalike audience Help is explicit: people in the source audience are excluded from the lookalike. That is one reason a lookalike is not retargeting.

Choose the seed for the job you want the campaign to do:

SeedWhen it is usefulMain check
PurchasersFinding more buyersEnough recent purchases and reliable order tracking
Qualified leadsGrowing a lead pipelineQualification status is passed back consistently
High-value customersFavoring customer value over raw volumeValue definition and sample size are stable
Site visitorsBuilding a larger early-stage seedExclude low-intent or accidental visits where possible

A larger visitor seed may give the model more reach, while a smaller purchaser or qualified-lead seed may better represent the outcome. The better choice depends on the campaign goal and usable matched size.

Meta Advantage+ custom audience can deliver beyond the selected custom audience when the system thinks that will improve performance. The same Help page states that Advantage+ lookalike is selected automatically for some objectives, and that selecting a custom audience plus a lookalike can surface an Advantage+ audience checkbox. Treat that as expansion, not as a closed lookalike.

Google optimized targeting looks beyond manually selected segments for people likely to convert. Manually selected lists are starting signals. They are not a hard fence unless you turn optimized targeting off and use exclusions.

Meta lookalikes: eligibility and controls

Current Meta create-lookalike Help (accessed September 20, 2026):

  • Source of at least 100 people from one country. Meta generally recommends 1,000 to 5,000 people.
  • Up to 500 lookalikes from one source.
  • Source people are excluded from the lookalike.
  • Saved audiences cannot be used as a source.
  • You need admin or advertiser permission on the ad account for a custom-audience source, admin permission on the Page or Meta Pixel for those sources, or admin/developer status on an app for an app source.
  • You no longer select location while creating a new lookalike. Estimated size appears when you use it on an ad set, using that ad set’s location. Existing lookalikes are unaffected.
  • Status can be Populating (up to 3 days), Ready, or Audience not created (often a source that was too small). You can run ads while it is populating; size will still grow.
  • Lookalikes are limited or unavailable for some ads about financial products and services, employment, or housing opportunities.
  • Targeting options are limited for audiences that include people under 18 (or older in some countries).

Meta’s About lookalike audiences Help still describes a percentage range: smaller percentages match the source more closely and larger percentages reach more people. That slider is not a quality score. Meta’s own note is that a source made from your best customers can outperform a source that includes everyone, depending on the goal. Do not treat “1%” as a universal rule.

Do not put sensitive health or financial labels on the source if Meta’s audience terms forbid them. Meta’s Marketing API audiences overview, accessed September 20, 2026, describes a rollout beginning September 2, 2025 that flags custom and lookalike audiences whose names or setup suggest disallowed health or financial status.

Google: Similar Audiences are gone

Google’s November 1, 2022 audience-strategy Help, Building more durable and effective audience strategies, set the sunset:

  • May 2023: new Similar Audiences (similar segments) stopped being generated and could no longer be added on Google Ads and Display & Video 360. Existing attachments kept working for a time.
  • August 2023: Similar Audiences were removed from ad groups and campaigns. Historical reporting remained.

Older checklists that tell advertisers to add Google Similar Audiences at fixed percentages no longer describe the current product.

What Google documents instead

Campaign situationCurrent controlWhat it is
Display, Demand Gen, and Video campaigns with Sales, Leads, or Website traffic goalsOptimized targetingOn by default. Uses landing page, creative, and optional first-party segments as signals, then can serve beyond those signals to people likely to convert.
Awareness and Reach or Consideration Video, where you previously used similar segmentsAudience expansion plus first-party data segmentsGoogle’s optimized-targeting Help points former similar-segment Video reach/consideration users here. Audience expansion looks for people similar to the selected segment.
Demand GenLookalike segmentsBuilt from first-party seed lists (Customer Match, website, app, Google Analytics, YouTube). Demand Gen only.

Demand Gen Lookalike segments, when used as a targeting constraint, use a reach slider Google documents as aiming for 2.5% (narrow), 5% (balanced), or 10% (broad) of users in the ad’s target location. Seeds refresh every 1 to 2 days. The constraint Help still requires the sum of submitted seed lists to be over 100 active matched people. Google recommends creating the campaign and segment 2 to 3 days before you need it. Lookalike segments cannot be used outside Demand Gen; a mixed audience that includes one will show as Eligible - Limited on other campaign types.

Google’s Use Lookalike segments to grow your audience Help, accessed September 20, 2026, states that Demand Gen Lookalike segments are moving to suggestion mode through a phased approach in 2026, with availability beginning in March 2026. In suggestion mode the reach setting becomes a signal rather than a hard 2.5%, 5%, or 10% boundary, and Google says the former 100-user minimum for creating a segment no longer applies. Advertisers can opt out through the form linked on that Help page. Check the live ad-group control to see which mode applies.

Optimized targeting and Lookalike-as-a-suggestion can both be on. Google says performance can then show up on the Lookalike row while optimized targeting may still reach people beyond that suggestion.

Seed readiness checklist

Use this checklist before you build or refresh a lookalike:

  1. Name the outcome the seed should copy. Purchasers, qualified leads, or high-value customers are different jobs.
  2. Use recent, consented records. Stale emails and unmatched rows shrink the real seed.
  3. Check matched size against the live product rule. Meta: at least 100 people from one country, with 1,000 to 5,000 generally recommended. Demand Gen constraint mode: more than 100 active matched people across seed lists. Google says the 100-user creation minimum no longer applies in suggestion mode.
  4. Exclude the seed from prospecting where the product does not already do it, so you are not paying to re-hit known customers unless that is the plan.
  5. Do not use a saved interest stack as a Meta lookalike source. Meta does not allow saved audiences as the source.
  6. Keep Special Ad Category and under-18 limits in view. Some lookalike controls will not appear.
  7. Wait for Ready / complete status before you judge volume. Populating lists are still growing.
  8. Hold creative and the conversion action when you compare lookalike prospecting with another prospecting method.

IAB’s State of Data 2024, dated March 14, 2024, surveyed more than 500 advertising and data experts and reported that 71% of brands, agencies, and publishers were growing or planning to grow first-party datasets, versus 41% two years earlier. That is dataset-investment context for seed quality. It is not a lookalike performance result.

Measurement example

Example measurement plan:

A shop has 4,200 purchasers from the last 180 days. Meta matches 3,100. That sits in Meta’s recommended 1,000 to 5,000 band and above the 100-person floor. The team builds a lookalike, excludes the purchaser seed, and runs it as prospecting.

If the platform offers an experiment, split comparable new people between:

  • Cell A: the lookalike
  • Cell B: an interest or optimized-targeting setup that does not use that lookalike

Keep the same creative, offer, landing page, conversion action, budget method, and calendar window. Predefine the primary metric as qualified purchases or cost per qualified purchase, not audience size.

Read, for the same window:

  • Spend
  • Unique people reached (with the platform’s identity rules)
  • Qualified conversions and conversion value
  • Cost per qualified conversion
  • How much delivery sat inside versus outside the selected list, if expansion was on

If Cell A has a lower cost per qualified purchase but too little volume to hit the sales plan, the team has a scale-versus-efficiency tradeoff. If both cells miss the qualified-purchase goal, inspect the seed definition, offer, and tracking before widening the audience or enabling more expansion.

A useful lookalike starts with a seed that represents the desired customer action. Expansion can add volume beyond that pattern, so report it separately when the platform exposes that split.

Frequently Asked Questions

What is a lookalike audience?

A lookalike audience is a prospecting group built by an advertising platform from patterns in a seed audience, such as purchasers or qualified leads. The ads go to new people the platform treats as similar. People in the Meta source audience are excluded from that lookalike.

How is a lookalike different from retargeting and automated expansion?

The seed is people you already know. Retargeting advertises back to those people. A lookalike advertises to new people similar to the seed. Automated expansion, including Meta Advantage+ audience suggestions and Google optimized targeting, can deliver beyond the selected list when the system predicts a better result.

What makes a Meta lookalike seed eligible?

Meta’s create-lookalike Help requires at least 100 people from one country in the source and generally recommends 1,000 to 5,000 people. Saved audiences cannot be the source. Location is set on the ad set for new lookalikes. Lookalikes are limited or unavailable for some housing, employment, and financial ads.

Did Google replace Similar Audiences with the same lookalike tool?

No. Google stopped generating Similar Audiences in May 2023 and removed them from campaigns in August 2023. Current replacements depend on campaign type: optimized targeting, audience expansion for some Video campaigns, and Lookalike segments in Demand Gen, which Google is shifting toward suggestion mode during 2026.

Sources

PUT THIS KNOWLEDGE TO WORK

NEED MORE HELP?

Talk with our team about applying Lookalike Audiences to your marketing.

Get a free marketing audit call