A lookalike audience is a copy machine. You hand Meta a source list, and it goes and finds more people who resemble that list. Which means the whole thing rises or falls on two levers, and most advertisers get both wrong.
Lever one: the seed
A lookalike inherits the quality of the list you built it from, exactly. Feed it 5,000 random newsletter signups and Meta faithfully finds you millions more people like your random newsletter signups. Feed it a few hundred clean, verified buyers, your actual best customers, and it goes hunting for more of them. Small and precise beats big and messy every time. The seed is the recipe: garbage in, garbage out, at scale. Building those converter lists is covered in the retargeting audience architecture guide.
This is why the source matters more than the size. The best seeds are tight customer lists of high-value buyers, or high-intent events captured cleanly by your pixel and Conversions API. If the signal feeding the seed is weak, everything downstream is weak too.
Lever two: the source size
When you build a lookalike, Meta lets you choose how wide to go, from 1% to 20% of the population in the locations your ad set targets (Meta no longer asks for a location when you create the lookalike). A 1% lookalike is the slice of that population most similar to your seed: the smallest reach, the closest match. As you climb to 5%, to 10%, up to 20%, the audience gets bigger and the resemblance gets weaker. Near the top of that range you are barely targeting your customer profile at all. You are just buying reach.
So people chase volume, set it to 10%, and then wonder why the leads feel nothing like their real buyers. They diluted the one thing a lookalike was supposed to give them: similarity.
How to actually use it
- Build the seed first. A clean, high-intent source list beats a large one. Quality of the seed sets the ceiling for everything.
- Start at 1 to 2 percent. That is the closest match to the people who actually pay you. Under Advantage+ audience, now the default, a lookalike is treated as a suggestion and delivery can go beyond it, so use original audience options when you need a 1% lookalike to stay tight.
- Widen only when you have exhausted it, and watch lead quality as you go, because every point you add trades similarity for size.
- Judge it on quality, not volume. A smaller audience of the right people outperforms a big audience of nearly-anyone, the same way a strong cost per lead means nothing if the leads never qualify.
Lookalikes are a demand-capture and scaling tool, not a fix for a weak source. Pair a tight lookalike with retargeting for a full-funnel setup, and remember that Meta is stronger at this kind of signal-based matching than at the self-declared targeting we cover in Meta's B2B targeting is built on self-declarations. LinkedIn works differently: it retired lookalikes in 2024, and predictive audiences replaced them.
The takeaways
- A lookalike copies its seed exactly, so a small, clean source list beats a large, noisy one.
- Source size runs 1% to 20%; 1% is the closest match and smallest reach, higher percentages trade similarity for reach.
- Chasing volume at 5 to 10 percent dilutes the resemblance that makes a lookalike useful.
- Start at 1 to 2 percent off a high-quality seed, widen only when exhausted, and judge on lead quality.
Common questions
What size should a Meta lookalike audience be?
For quality-led lead generation, start at 1 to 2 percent, which is the closest match to your source list. Larger sizes, up to 20 percent of the population in your ad set's locations, reach more people but resemble your source less, so they suit scaling for volume rather than precision.
Why is my lookalike audience low quality?
Almost always the seed. A lookalike copies the list you built it from, so a noisy or low-intent source produces a low-intent audience at scale. Rebuild it from a clean list of your real, high-value customers.
Does a bigger lookalike percentage find better customers?
No. A bigger percentage finds more people, but less similar ones. As you move from 1% toward 20%, resemblance to your source weakens, so you are buying reach, not precision. It never finds better customers than your seed, only more of whatever the seed already is.