Do Facebook lookalike audiences still work in 2026? Yes — but not the way they did five years ago. Lookalike audiences are still available in Meta Ads Manager, they still build a cold audience from a seed of your best customers, and in our experience they still outperform pure interest targeting for most small e-commerce and lead-gen accounts. What has changed is that Meta’s Advantage+ audience now uses your seed lists as a suggestion rather than a hard boundary, and for many advertisers that broader approach wins on cost per result.
The short answer for a small or mid-sized business in the US or Europe: keep building high-quality seed audiences, test lookalikes against Advantage+ audience in a controlled way, and let the data decide. Below we cover how lookalikes work today, which seeds produce good ones, how they interact with Advantage+ features, and a testing plan you can run in two weeks.
What a Facebook lookalike audience is (and what it isn’t)
A lookalike audience is a cold audience Meta builds by finding people whose behavior and characteristics resemble a source audience you provide — your customer list, website purchasers, people who engaged with your Instagram, or app users. You choose a country or region and a size from 1% to 10% of that population; 1% is the closest match, 10% the broadest. Meta’s official description is in Meta Business Help: About lookalike audiences.
What it isn’t: a list of your customers. Lookalikes exclude the source audience by default, so they’re a prospecting tool, not retargeting. If you want to reach people who already visited your site, that’s a custom audience, covered in our Facebook retargeting ads guide.
Why lookalikes got weaker, and why they still matter
Three things reduced the sharpness of lookalikes after 2021: Apple’s App Tracking Transparency cut the amount of off-platform behavior Meta observes, EU privacy rules (GDPR, the ePrivacy consent requirements) reduced how much website data reaches Meta without consent, and Meta shifted its own modeling toward broad, signal-driven delivery. A 1% lookalike built from a thin, consent-limited seed is simply less distinct than it used to be.
At the same time, the thing that made lookalikes valuable — a seed that describes your actual buyers — hasn’t lost value. It just gets used differently now. With Advantage+ audience, Meta treats your custom audiences and lookalikes as inputs to its model and will go outside them when it predicts better results. In practice, this means the quality of your seed still affects performance, even when the lookalike “boundary” is no longer enforced.
Which seed audiences produce the best lookalikes
Not all seeds are equal. The ranking below reflects what typically works in accounts we manage, from strongest to weakest:
| Seed source | Quality | Notes |
|---|---|---|
| Purchasers, last 180 days (via Pixel + Conversions API) | Highest | Needs at least a few hundred events; value-based version uses order value |
| Customer list upload (emails/phones, repeat buyers) | High | Match rates of 40–70% are typical; refresh monthly |
| Qualified leads (CRM-synced via CAPI or lead ads) | High for lead gen | Use closed/qualified leads, not every form submit |
| Add to cart / initiate checkout | Medium | Larger seed, lower intent; useful when purchases are scarce |
| Instagram/Facebook engagers, 90 days | Medium-low | Easy to build, but engagement ≠ purchase intent |
| All website visitors | Low | Too broad; mostly duplicates what broad targeting already does |
| Video viewers (3s) | Low | Fine for awareness, weak for conversions |
Two practical rules. First, Meta recommends a source of at least 100 people in a single country, but seeds under about 1,000 tend to produce noisy lookalikes; aim for more. Second, the seed is only as good as your tracking. If the Pixel is losing a third of your purchase events, your purchaser seed is biased toward whoever happens to accept cookies. The fix is server-side tracking — see Meta Conversions API: why your Pixel alone isn’t enough.
Lookalikes vs. Advantage+ audience vs. broad
You have three realistic cold-targeting options today:
- Classic lookalike (Advantage+ audience off): delivery stays inside the lookalike. Most control, smallest reach, often higher CPM.
- Advantage+ audience with your lookalike as a suggestion: Meta prioritizes your audience but expands beyond it. In our experience this is the best default for most accounts spending $1,000–10,000/month.
- Fully broad (no suggestions): age, country, and nothing else. Works well for mass-market products with strong creative and plenty of conversion volume; weaker for niche B2B or high-ticket services.
In Advantage+ sales campaigns (the evolution of Advantage+ shopping campaigns we covered in our Advantage+ shopping campaigns guide), targeting is largely automated and your lookalikes mainly influence delivery through the audience-suggestion and existing-customer settings. That is fine — it just means your energy belongs in creative and in seed quality rather than in audience segmentation.
How to set up a lookalike audience in 2026
- In Ads Manager, go to Audiences → Create audience → Custom audience. Build the seed (e.g., Website → Purchase, 180 days; or Customer list upload).
- Wait until the audience shows a size estimate (customer lists need processing time).
- Create audience → Lookalike audience. Select the seed, the country or countries, and a size. For most accounts, start with 1–3%; for small countries (Netherlands, Belgium, Switzerland) consider 3–5% to get enough reach.
- If you have purchase values, create a value-based lookalike from the purchase event so Meta weighs higher-value customers more heavily.
- In the ad set, add the lookalike. Leave Advantage+ audience on for the first test; create a duplicate with it off only if you want a strict comparison.
- Exclude recent purchasers (30–60 days) and active retargeting audiences so prospecting stays clean.
- Give each ad set a budget sufficient for at least 50 conversions per week if you can; if not, consolidate into fewer ad sets.
A two-week lookalike test plan
Use this structure if you’re not sure whether lookalikes still help your account:
- Ad set A: Advantage+ audience on, 1–3% purchaser lookalike as suggestion.
- Ad set B: Advantage+ audience on, no suggestions (broad).
- Ad set C (optional): Advantage+ audience off, same lookalike, strict.
- Same creative, same budget, same optimization event (Purchase or Lead), same dates. Run 14 days or until each ad set has 30+ conversions.
- Compare cost per result and downstream quality (average order value, lead qualification rate). Lookalikes often win on quality even when broad wins on cost per lead.
Use Meta’s A/B test feature rather than manual duplicates where possible; it prevents audience overlap from skewing the result.
EU specifics: consent, GDPR, and smaller seeds
If you advertise in Germany, France, the Netherlands, or elsewhere in the EU, your seeds are built only from users who consented to tracking. Consent rates vary by market and cookie banner design, but losing 30–50% of events is common. Practical adjustments: use longer lookback windows (180 days instead of 30) to grow the seed, upload CRM customer lists (first-party data you legitimately hold) as a complement, and prefer 3–5% sizes in smaller countries. Also make sure the customer list upload itself has a lawful basis under GDPR and that your privacy policy covers it.
Costs differ too: CPMs in the US are typically higher than in Southern Europe, so a lookalike that looks “expensive” in dollars may be perfectly efficient in euros. Judge by cost per result in local currency against your margin, not by CPM alone.
Common lookalike mistakes
- Building a lookalike from a seed of 150 people and expecting precision.
- Stacking five lookalikes of different seeds in one ad set with Advantage+ off — just use broad at that point.
- Never refreshing an uploaded customer list; six-month-old lists drift.
- Judging after three days. Learning phase plus attribution lag means a week is the minimum.
- Using a lookalike of engagers for a conversion campaign and concluding “lookalikes don’t work.”
FAQ
Do Facebook lookalike audiences still work in 2026?
Yes, especially when built from purchasers or qualified leads and used alongside Advantage+ audience. They are less precise than in 2019–2020 but still a useful signal for Meta’s delivery system.
What lookalike percentage is best?
Start with 1–3% in large markets like the US, UK, or Germany, and 3–5% in smaller countries. Test larger sizes once the smaller one is saturated or CPMs climb.
How many people do I need in a source audience?
Meta’s minimum is 100 in a single country, but in practice 1,000+ gives much more stable results.
Should I turn off Advantage+ audience to use a lookalike?
Usually not. Keeping it on with your lookalike as a suggestion typically delivers a lower cost per result. Turn it off only for strict testing or when compliance requires narrow targeting.
Lookalike vs. broad targeting — which is cheaper?
Broad often has lower cost per result for mass-market products; lookalikes often deliver higher-quality customers for niche or high-ticket offers. Test both with the same creative.
Can I build lookalikes for Instagram ads only?
Yes. Lookalikes are built at the audience level and work across Facebook, Instagram, Reels, and Audience Network; you control placement in the ad set.
Need help with your campaigns?
We build Meta campaigns with clean tracking, high-quality seed audiences, and a creative-testing structure so you’re not guessing which audiences work. Start with Meta Ads Setup (Facebook + Instagram) (€490), or hand over the whole account with Meta Ads Management. Not sure where to start? Book a free 30-minute call.