Lookalike Audiences in 2026: How They Work and When They Still Beat Broad

A lookalike audience is a Meta audience of people who resemble a source group you provide, such as your customers, ranked by how similar they are within the country or region you choose. In 2026 lookalikes still work, but in many Sales campaigns Meta treats them as a suggestion rather than a boundary, and they no longer reliably beat broad targeting. Build them from your best buyers and test them against broad.

How do lookalike audiences work?

You give Meta a seed: a custom audience such as a customer list, people who bought on your website, or people who engaged with your Instagram account. Meta looks for patterns among those people and finds others in your chosen location who share them.

Three settings define a lookalike:

1. Source. The seed audience, or a pixel event, catalogue or Page you choose.
2. Location. One or more countries or regions where Meta should look for similar people.
3. Size. In Ads Manager, a percentage from 1% to 10% of the population in that location. 1% is the most similar slice, 10% the broadest.

Meta's documentation states that a seed needs at least 100 people, and Ads Manager asks for at least 100 of them to be in a single country. It also notes that a new lookalike can take from about an hour to several hours to be ready, and that lookalikes used in active ads are refreshed periodically, so they track changes in the seed over time.

What has changed?

Lookalikes were once the main targeting tool for e-commerce. Two things changed that.

• Signal loss. Since Apple's App Tracking Transparency prompt, Meta sees fewer cross-app signals, which weakened some of what made lookalikes precise. See App Tracking Transparency.
• Automation. Meta's delivery system got better at finding buyers from broad audiences on its own, and Meta built that into Advantage+ audience and Advantage+ sales campaigns.

Today, if you add a lookalike to an ad set with Advantage+ audience on, Meta uses it as an audience suggestion: a starting point it prioritises before searching more widely. Even in some manual setups, a feature Meta calls Advantage lookalike (formerly lookalike expansion) can deliver beyond your chosen percentage when Meta expects better results, and in some accounts it can't be switched off for conversion goals. Check the audience section of your ad set to see which behaviour applies.

How do you create a lookalike audience?

1. Build the seed first. In Audiences, create a custom audience: a customer list, website purchasers over a set number of days, or another source you trust.
2. Choose Create audience, then Lookalike audience.
3. Select the source you just built, or a pixel event, catalogue or Page if those are offered.
4. Choose the location. Pick the country or region you sell to. Meta needs enough seed members in a single country to build the audience.
5. Choose the size. Start with 1%, or create several sizes at once (for example 1%, 2% and 3%) to test later.
6. Create the audience and wait for it to be ready. Its status shows in the Audiences list.
7. Add it to an ad set, either as the audience or, with Advantage+ audience on, as a suggestion.

Name each lookalike with its seed, date and size, such as "LAL 1% purchasers 180d IN Oct 2026". Six months later, you'll know exactly what it contains.

What makes a good seed audience?

A lookalike can only be as good as the people it copies. Meta finds people like your seed, so a seed of bargain hunters returns more bargain hunters.

Size versus quality

Meta's minimum is 100 people, but a seed that small is a thin pattern to learn from. A seed of a few thousand purchasers gives Meta far more to work with. Equally, don't dilute a strong seed to make it bigger: 2,000 real customers beat 50,000 page engagers when the goal is sales.

Keep seeds fresh

Customer lists you upload once go stale. Pixel-based seeds, such as purchasers in the last 180 days, update automatically. If you use a customer list, re-upload it monthly or connect your store so it syncs.

What is a value-based lookalike?

A value-based lookalike uses a seed that includes how much each customer is worth, so Meta looks harder for people like your highest-value customers rather than treating every buyer equally.

To build one, upload a customer list that includes a value column, such as total spend or predicted lifetime value, and choose that column when you create the audience. Some accounts also offer value-based sources built from pixel purchase values.

When value-based lookalikes help:

• Your customers vary a lot in value. If some buy once at a small order value and others buy every month, value weighting can matter.
• You have enough customers. Value-based seeds need a reasonable spread of buyers to show the difference between high and low value.
• You care about customer lifetime value, not just the first order.

If all your customers buy one product once at the same price, a value-based lookalike adds little over a standard purchaser lookalike.

Which lookalike size should you use?

The percentage trades similarity for reach. In a country with a large population, even 1% is several hundred thousand people or more.

You can also build a range, such as 2% to 5%, which excludes the closest 2%. That is useful when you already run a 1–2% lookalike and want a new slice without overlap.

In small countries or tightly defined regions, a 1% audience can be too small to deliver efficiently. Watch frequency and CPM: if both rise quickly while results fall, the audience is too narrow for the budget.

Do lookalikes still beat broad targeting?

Sometimes, and you only find out by testing. Broad targeting means setting only basic limits such as location and age, then letting Meta use your creative and conversion data to find buyers. In many accounts with good tracking and enough purchase volume, broad now performs as well as or better than lookalikes, because Meta can search the whole population rather than a slice of it.

Lookalikes still tend to help when:

• The account is new and has little conversion history, so a strong seed gives Meta a head start.
• The product is niche, and the people who buy it are hard to find from creative alone.
• Budgets are small, and you can't afford a long period of Meta exploring the wide audience.
• Tracking is weak, so Meta's conversion signal is thin and a seed fills the gap.

Broad tends to win when the pixel and Conversions API are reliable, the account has steady purchase volume and the creative clearly signals who the product is for. Our guide to broad vs interest vs lookalike targeting goes deeper on the trade-offs.

How to test lookalike against broad

1. Create two ad sets in the same campaign on ad set budgets: one with a 1–3% purchaser lookalike, one broad.
2. Use identical ads in both, so audience is the only difference.
3. Give each the same daily budget, enough for each to spend at least several times your target CPA over the test.
4. Run for at least a week, so weekday patterns and attribution delays even out.
5. Judge on cost per purchase and new-customer share, not CTR. A lookalike can look efficient because it reaches many existing customers.
6. Repeat the test every few months. The winner can change as your account and Meta's systems change.

How should you use lookalikes in your campaign structure?

A practical setup for most e-commerce accounts in 2026:

• Main scaling campaign: broad or an Advantage+ sales campaign, with your customer data connected so Meta can learn from it.
• Lookalike as a suggestion: inside Advantage+ audience, add your best purchaser or value-based lookalike as a suggestion rather than a hard limit.
• Dedicated lookalike ad set: only if testing shows it beats broad in your account, kept in a separate ad set so you can measure it.
• Retargeting: handled separately with custom audiences of recent visitors and cart abandoners. Our retargeting guide covers it.

Avoid stacking many overlapping lookalikes (1%, 2%, 3%, 1% of a different seed) in separate ad sets on a small budget. They overlap heavily, compete with each other and split your conversion data so that none of them learns well. See CBO vs ABO for how many ad sets a budget can feed.

Common lookalike mistakes

1. Seeding from engagement for a sales goal. People who like posts aren't the same as people who buy.
2. Never refreshing customer lists. A seed from two years ago describes customers you had, not the ones you want.
3. Judging on CTR or CPM. Lookalikes can have cheap clicks and expensive purchases. Decide on cost per purchase.
4. Ignoring overlap with existing customers. Without exclusions, lookalikes can soak up budget reaching people who would have bought anyway.
5. Assuming the percentage is a hard fence. In many setups Meta can go beyond it. Read the audience settings.
6. Too many lookalike ad sets. Consolidate so each ad set gets enough purchases to learn.

The bottom line

Lookalikes still have a place, but as a tool, not a strategy. Build them from your best and most recent customers, use value-based seeds when your customers vary in worth, start around 1–3%, and treat them as a suggestion to Meta's automation rather than a fence. Above all, test them against broad on identical creatives, judge on cost per purchase from new customers, and let your own account decide.