Custom Audiences vs Lookalike Audiences: Seed and Extension
In short: These work together more than they compete - a custom audience is often the seed that makes a lookalike possible in the first place. Custom Audiences are built from data you already hold or activity people generated directly with your business, so membership is the exact people who did something specific. Lookalike Audiences extend that reach to strangers, using a custom audience as a seed and finding people elsewhere in a chosen country who resemble it on the signals Meta holds. The practical distinction is who's actually in the audience: a custom audience is people you can name reasons for, a lookalike is people you can't name at all, just people the model judged similar. Use a custom audience to retarget or exclude people you already have a relationship with, and use a lookalike once that custom audience is solid enough to model, when you want to prospect for more people like them.
By the AdFlint research team · Fact-checked against current Google and Meta platform behavior · Last reviewed July 2026
Custom Audiences
Meta audiences built from data you already own or activity people generated on Meta, such as site visitors, customer lists, app users, or video viewers.
You create them in Ads Manager from a source - pixel or Conversions API events, an uploaded list, app activity, or on-platform engagement - then target or exclude them in any ad set. They power both retargeting and lookalike seeds. Most accounts under-use exclusions, retargeting existing customers alongside prospects, and forget that website-based sources expire on whatever retention window was set.
Full definitionLookalike Audiences
Meta audiences modelled on a seed list, matching users who resemble your source on the signals Meta holds, sized as a percentage of a chosen country.
You pick a source - purchasers, high-value customers, engaged visitors - and a percentage; smaller percentages mean closer resemblance and less reach, larger ones the reverse. A value-based source lets Meta weight by revenue. Their edge has narrowed as broad targeting and Advantage+ audience improved. The persistent error is seeding from a large low-quality list, which models the wrong person very precisely.
Full definitionSide by side.
The differences that actually change what happens in your account.
| Custom Audiences | Lookalike Audiences | |
|---|---|---|
| Who's actually in it | The specific people who match your source data or on-platform activity. | Strangers who resemble your seed on signals Meta holds, not necessarily anyone in the seed. |
| Depends on the other | No - stands alone as retargeting or exclusion. | Yes - cannot exist without a custom audience or similar list as its seed. |
| Sizing control | Determined by your source and retention window, not directly adjustable. | A percentage-of-country slider you set directly. |
| Value weighting | Possible if the source has a value field, like a customer list with revenue. | Possible - a value-based seed lets the model weight toward high-revenue resemblance. |
| Funnel role | Retargeting, exclusion, and serving as lookalike seed material. | Cold prospecting extended from a warm base. |
| Failure mode | Under-using exclusions, so existing customers get retargeted like prospects. | Seeding from a large, low-quality list, which models the wrong person precisely. |
| Freshness | Varies by source - website audiences expire on retention window, customer lists go stale until re-uploaded. | Recalculated against the current seed each time the lookalike refreshes, but only as good as the seed feeding it. |
What actually separates them.
A custom audience is made of specific, identifiable people tied to your data or activity, while a lookalike is made of strangers who merely resemble that data, not the same people at all.
A lookalike cannot be built without a custom audience or comparable source feeding it as a seed, while a custom audience needs nothing beyond your own data or activity to exist.
Custom audience size is a byproduct of your source and retention settings, while lookalike audience size is a percentage you choose directly, trading resemblance for reach.
Both support value-based weighting, but a custom audience needs a value field in its source data to enable it, while a lookalike needs a value-based custom audience specifically as its seed to inherit that weighting.
A custom audience's biggest common failure is forgetting to exclude it from prospecting, while a lookalike's biggest common failure is seeding it from a large but low-quality list, which produces a confident, well-targeted model of the wrong person.
Which one should you use?
Use Custom Audiences when
- You want to retarget site visitors, purchasers, or people who engaged with your content.
- You need to exclude existing customers from a prospecting campaign.
- You're building the raw material that a future lookalike will need as a seed.
- You want precise control over who's included based on real data rather than modeled resemblance.
Use Lookalike Audiences when
- You already have a solid custom audience, like purchasers, to seed from.
- You want to extend reach to strangers who resemble your best customers.
- You want a value-based seed to weight the lookalike toward high-revenue resemblance.
- Broad or Advantage+ audience prospecting isn't finding enough of your best-fit buyers on its own.
Common questions.
What's the smallest custom audience I can use to seed a lookalike?
Meta recommends a seed in at least the low thousands to model reliably, though a smaller custom audience can technically be used. Below that, there's too little signal for the model to generalize well, and lookalike quality tends to suffer.
If I exclude a custom audience from a lookalike campaign, does that also affect the seed?
No, excluding a custom audience from the ad set so existing customers don't see the ad is separate from the seed used to build the lookalike, and doesn't change or shrink the seed itself. You can exclude your customer list from delivery while still using that same list as the lookalike's seed.
How fresh does my custom audience seed need to be for a lookalike to stay accurate?
A lookalike is generally rebuilt periodically against its current seed, so if the seed audience is a static customer list that never gets refreshed, the lookalike gradually models an increasingly outdated version of your customer base. Refreshing the seed source keeps the lookalike relevant.
Can I build a lookalike from a website custom audience instead of a customer list?
Yes, any custom audience can serve as a lookalike seed, including one built from pixel or Conversions API events on your website, not just an uploaded list. A high-intent website audience, like purchasers, often makes as good or better a seed than a broad customer list.
Or stop choosing between them.
AdFlint picks the setting, writes the ads, and keeps optimizing inside the Google and Meta accounts you already own.
Related comparisons
- Advantage+ Audience vs Broad Targeting
- Broad Targeting vs Detailed Targeting
- Broad Targeting vs Interest Targeting
- Broad Targeting vs Custom Audiences
- Broad Targeting vs Lookalike Audiences
- Broad Targeting vs Website Custom Audiences
- Broad Targeting vs Customer List Custom Audiences
- Broad Targeting vs Engagement Custom Audiences