Key Takeaways
- Start with 1% lookalikes for direct-response campaigns. Testing shows they deliver leads at $3.75 vs $6.36 for 10% lookalikes.
- Source audience quality beats size every time. Purchase-based seeds outperform interest targeting by 26% on CPA.
- Use lookalikes as Advantage+ suggestions in 2026, not standalone targeting. Early testing shows 23% better conversion rates with this hybrid approach.
- Refresh your source audiences quarterly at minimum. Stale data is the most common mistake we see in account audits.
In This Article
- What Are Meta Lookalike Audiences?
- Lookalike Audiences vs. Advantage+ Audiences
- How to Create a Lookalike Audience in Meta
- Choosing the Right Source Audience
- Percentage Sizing: 1% vs. 5% vs. 10%
- Super Lookalikes: 11-20% via the Meta API
- Common Mistakes (and How to Fix Them)
- Are Lookalike Audiences Still Worth It in 2026?
- Frequently Asked Questions
Meta lookalike audiences let you reach new people who share behavioral and demographic traits with your existing customers. You upload a source audience (buyers, email subscribers, high-value website visitors), and Meta’s algorithm finds the top 1-10% of users in your target country who look most like them. In 2026, lookalikes remain one of the most reliable prospecting tools in paid social, even as Advantage+ audience targeting absorbs more of the manual work advertisers used to do themselves.
What Are Meta Lookalike Audiences?
A lookalike audience is a targeting layer built from a seed list you provide to Meta. That seed, called the “source audience,” can be a customer list, a website custom audience based on pixel data, or engagement data from your Instagram or Facebook page.
Meta analyzes hundreds of signals across its user base: purchase behavior, app activity, content engagement, device usage, and more. It then ranks every user in your chosen country by how closely they resemble your source audience and groups the top percentage into a targetable segment.
The practical outcome: instead of guessing which interests or demographics define your ideal buyer, you let Meta reverse-engineer it from real conversion data.
How It Works in Practice
- Source quality matters more than source size. A list of 500 repeat purchasers will outperform a list of 50,000 email opens. Meta recommends between 1,000 and 5,000 people in your source, but we have seen strong results from smaller, high-intent lists.
- Lookalikes are country-specific. A 1% lookalike in the US covers roughly 2.1 million people. The same 1% in a smaller market like the Netherlands reaches far fewer.
- They update dynamically. Meta refreshes lookalike audiences every 3-7 days, so as your source audience changes, the lookalike shifts with it.
Lookalike Audiences vs. Advantage+ Audiences
This is the question we get most often from clients right now. Meta rolled out Advantage+ audience as the default for new campaigns, and it can feel like lookalikes are being pushed aside. They are not the same thing, and understanding the difference matters for how you structure your account.
In testing by Jon Loomer, Advantage+ audience and detailed targeting performed comparably, but both significantly outperformed standalone lookalikes in certain campaign types. The caveat: this depends heavily on account maturity, pixel history, and creative volume (Jon Loomer Digital).
What We Recommend at Jetfuel
Feed lookalike audiences as “audience suggestions” inside Advantage+ campaigns. This gives Meta’s AI a strong starting signal based on your best customers while letting it expand beyond the lookalike when it finds better-performing segments. You get the precision of a curated seed with the scale of algorithmic targeting.
How to Create a Lookalike Audience in Meta (Step by Step)
This takes about five minutes once you have your source audience ready. If you do not have a source audience yet, skip to the next section first.
Open Meta Ads Manager and Navigate to Audiences
Go to Meta Ads Manager. In the left navigation, click All tools, then Audiences under the Assets section.
Click “Create Audience” and Select “Lookalike Audience”
Click the blue Create audience dropdown and select Lookalike audience from the three options (Custom, Lookalike, Saved).
Choose Your Source Audience
This is the most important decision in the entire process. Select an existing custom audience from the dropdown. Best sources in order of typical performance: purchase-based custom audiences, high-value customer lists (top 25% by LTV), add-to-cart audiences, engaged email subscribers, and page/profile engagers.
Select Your Target Country
Choose the country where you want to find similar people. Meta will only find lookalikes within the country you select. For multi-country targeting, create separate lookalikes for each.
Set Your Audience Size Percentage
Use the slider to select between 1% and 10%. Start with 1% for direct-response campaigns. See the percentage sizing section below for when to go wider.
Create the Audience
Click Create audience. Meta will begin populating the lookalike. This typically takes 6-24 hours, though you can select it for campaigns immediately.
Apply It to Your Campaign
In your ad set targeting, choose the lookalike audience you just created. Layer on exclusions (existing customers, recent purchasers) to avoid wasting spend on people who have already converted.
Pro Tip
Always exclude your source audience from the lookalike campaign. Otherwise you are paying prospecting CPMs to reach people you already have in a custom audience.
Choosing the Right Source Audience
Your source audience is the single biggest lever in lookalike performance. We have seen the same percentage, same budget, same creative produce wildly different results based solely on the quality of the seed.
Start with conversion events, not vanity metrics. A source audience built on “people who purchased in the last 90 days” will almost always outperform one built on “people who visited the website.” The closer the source is to actual revenue, the better Meta can model what a paying customer looks like.
Segment by value, not just by action. Upload your top 25% of customers by lifetime value as a separate source audience. The people who bought once during a Black Friday sale are not the same signal as your repeat buyers who come back every month.
Keep the recency window tight. For most brands, 30-90 day purchase windows outperform 180-day or all-time lists. Recent buyers reflect your current customer profile, not who was buying from you two years ago when your product line might have been different.
Minimum size considerations. Meta requires at least 100 people from a single country in your source audience, but that is a floor, not a target. We recommend at least 1,000 people for consistent results, and 2,000-5,000 for optimal modeling. If your list is under 1,000, focus on growing your customer base or use a broader event (like add-to-cart) until you have enough volume.
Percentage Sizing: 1% vs. 5% vs. 10%
The percentage determines how closely Meta matches to your source and how large the resulting audience is. This is not a “bigger is better” situation.
AdEspresso ran a $1,500 experiment testing 1%, 5%, and 10% lookalikes head to head. The 1% lookalike delivered a $3.75 cost per lead. The 5% came in at $4.16. And the 10% lookalike cost $6.36 per lead, nearly 70% more expensive than the 1% (AdEspresso).
A layering strategy that works well: Create separate ad sets for 1%, 1-3%, and 3-5% lookalikes. Run them simultaneously with separate budgets. This lets you compare performance at each tier and shift budget toward whatever is working without mixing the data.
Super Lookalikes: 11-20% Audiences via the Meta API
Most advertisers don't know this exists. The Meta Ads Manager UI caps lookalike audiences at 10%. But the Marketing API lets you create lookalikes up to 20%, and in some cases beyond. These aren't available through the normal interface. You need API access or a tool built on top of it.
We call these "super lookalikes" internally, and they fill a gap that nothing else in Meta's targeting stack covers well.
How It Works
The Meta Marketing API accepts a ratio parameter between 0.01 and 0.20 when creating lookalike audiences. Set ratio: 0.15 and you get a 15% lookalike, roughly 31 million people in the US. The same source audience quality rules apply. The algorithm just draws a wider circle around your seed.
When to Use 11-20% Super Lookalikes
These are not for every campaign. But in the right scenario, they outperform both standard 10% lookalikes and pure broad targeting.
Top-of-funnel awareness at scale. When you are spending $500+/day on prospecting and your 10% lookalike is tapped out (frequency climbing, CPMs rising), a 15% lookalike gives you fresh reach without abandoning the seed signal entirely. You are still anchored to your customer data. Broad targeting throws that away.
Testing in markets where 10% is too small. In countries with smaller Meta user bases (Australia, Netherlands, Nordics), a 10% lookalike might only give you 500K-1M people. That is not enough for Meta's algorithm to optimize efficiently on a meaningful budget. Expanding to 15-20% gives the system room to work.
Advantage+ audience seeding for big budgets. If you are feeding lookalikes as audience suggestions into Advantage+ campaigns, a wider seed (15-20%) gives the algorithm a broader starting signal while still anchoring to your customer profile. We have tested this on accounts spending $50K+/month and seen it reduce CPAs versus pure broad with no suggestion at all.
Layered exclusion testing. Create a 20% lookalike, then exclude your 10% lookalike from it. The resulting audience is everyone in the 11-20% band: people who are somewhat similar to your customers but not close matches. This is a useful testing audience. If it converts well, your product has broader appeal than you thought. If it bombs, you know your sweet spot is in the tighter percentages.
Build Super Lookalikes for Free
We built a free tool that lets you create 11-20% lookalike audiences through the Meta API without writing a single line of code. Connect your ad account, pick your source audience and percentage, and the tool generates the lookalike directly in your Ads Manager.
Common Mistakes (and How to Fix Them)
We audit a lot of Meta accounts. These are the lookalike mistakes we see most often.
Using your entire email list as the source. Your full email list includes unengaged subscribers, one-time buyers from three years ago, and people who signed up for a giveaway and never came back. That is noise, not signal. Fix: segment to engaged buyers from the last 90-180 days, or upload only your top LTV segment.
Never refreshing the source audience. If your custom audience is based on a static CSV uploaded 18 months ago, your lookalike is modeling against stale data. Fix: re-upload customer lists quarterly. For pixel-based audiences, make sure your conversion events are firing correctly. Check your Meta ads attribution settings to make sure conversions are being captured accurately.
Stacking too many targeting layers on top of the lookalike. Adding interest targeting on top of a 1% lookalike audience shrinks the pool dramatically and constrains Meta’s optimization. Fix: use the lookalike as your sole targeting layer, and only add exclusions (not inclusions).
Ignoring audience overlap. Running a 1% lookalike and a 1-3% lookalike in separate ad sets means you are bidding against yourself for the users in that 1% slice. Fix: use the “Audience Overlap” tool in Ads Manager to check, and exclude the smaller lookalike from the larger one.
Starting with 10% because “more reach is better.” Broader is not better when your budget is limited. A $50/day budget spread across a 21-million-person 10% lookalike gives Meta almost nothing to optimize with. Fix: start narrow (1%), prove the funnel works, then scale the percentage up.
Not excluding existing customers. If your source audience is your customer list and you do not exclude that same list from the campaign, you are paying to reach people who already bought. Fix: add your customer list and recent converters (via pixel) as exclusions in every prospecting campaign.
Are Lookalike Audiences Still Worth It in 2026?
Short answer: yes, but the way you use them has changed.
Meta’s Advantage+ ecosystem has absorbed a lot of the manual targeting work that lookalikes used to handle exclusively. WordStream’s 2026 data shows Meta is consolidating detailed targeting categories and merging specific interests into broader groups, giving advertisers less granular control overall (WordStream).
When lookalikes outperform broad/Advantage+ in 2026:
- New ad accounts with limited pixel history
- Niche products where Meta’s broad algorithm takes too long to learn
- High-AOV products where the cost of incorrect targeting is expensive
- Campaigns where you need to isolate and measure a specific audience segment
When broad/Advantage+ likely wins:
- Mature accounts with 100+ conversions per week
- Large creative libraries that give the algorithm room to test
- Brand awareness campaigns optimizing for reach
- Accounts with deep conversion event history
For most of the brands we work with at Jetfuel, the answer is not either/or. We run lookalikes as prospecting seeds alongside Advantage+ campaigns.
Frequently Asked Questions
What percentage should I use for Meta lookalike audiences?
Start with 1% for any direct-response campaign. AdEspresso’s testing showed 1% lookalikes delivered leads at $3.75 versus $6.36 for 10% lookalikes, a 70% cost difference. Only expand to 3-5% once your 1% is performing well and you need more scale. Use 6-10% primarily for awareness campaigns where reach matters more than per-conversion efficiency.
Are lookalike audiences still effective in 2026?
Yes. Lookalikes remain effective, especially when paired with Advantage+ as audience suggestions rather than used as standalone targeting. They are particularly valuable for newer ad accounts without deep pixel history, niche products, and high-AOV items where incorrect targeting is costly. The brands seeing the best results in 2026 treat lookalikes as a signal input to Meta’s AI, not as a rigid audience boundary.
How often should I refresh my source audience?
For CSV-based customer lists, re-upload quarterly at minimum. For pixel-based custom audiences (website visitors, purchasers), they refresh automatically as long as your pixel and conversion events are firing correctly. Review your source audiences monthly to check that the underlying data is current and that the audience size is growing. A source audience that has not changed in six months is a red flag.
Lookalike audiences vs. Advantage+: which is better?
Neither is universally better. They solve different problems. Lookalikes give you control over the signal (you define who your best customers are). Advantage+ gives Meta control over the delivery (the AI decides who to reach). The best-performing setup in 2026 combines both: use a lookalike as the audience suggestion inside an Advantage+ campaign. This gives the algorithm a strong starting point while letting it explore beyond your seed when it finds better segments.
Looking Ahead
Meta’s targeting tools will keep consolidating around AI-driven delivery. Lookalike audiences are not going away, but the advertisers who treat them as one input in a broader system (not the entire strategy) will get the best results. Build high-quality source audiences from real conversion data, start narrow, test methodically, and let the algorithm do what it is actually good at.
Need help with your Meta audience strategy?
If your Meta campaigns need a strategic reset, or you are not sure whether your current audience structure is leaving money on the table, our paid social team can help.
