LinkedIn Lookalike Audiences: what happened to them, and what to use now

If you came here looking for LinkedIn's Lookalike Audiences, here's the part nobody updated the old blog posts to tell you: LinkedIn retired Lookalike Audiences in early 2024. The feature that let you upload a seed list and have LinkedIn find demographically similar professionals is gone. In its place LinkedIn pushed everyone toward Predictive Audiences — same underlying goal ("find more people like my best customers"), different, more AI-driven machinery.

So if your real question is "how do I reach more people like the customers I already have on LinkedIn?", this is the current answer — what changed, what to use instead, and the part that actually determines whether it works.

Why this matters (and why the old advice will burn you)

The idea behind lookalikes was always sound: you know a lot about your best customers, so use a platform's data to find more people who resemble them. Meta has done this since 2013. LinkedIn was characteristically late — it only rolled lookalikes out to all advertisers in 2019 — and then, just as people got comfortable, folded the feature into its newer Predictive Audiences and sunset the standalone version in early 2024.

The takeaway: any guide still walking you through the old "create a Lookalike Audience" flow in Campaign Manager is describing a button that no longer exists. The strategy survives; the feature name and the clicks changed.

What replaced it: Predictive Audiences

A Predictive Audience does what a lookalike did, but LinkedIn's AI does more of the work. You give it a data source to learn from — a matched audience (uploaded list or website retargeting), a Lead Gen Form audience, or a conversions-based audience — and LinkedIn models the professional traits of those people and assembles a fresh audience of similar members you can target.

The mental model is the same as a lookalike: a high-quality seed in, a larger similar audience out. What you've lost is some manual control; what you've gained is LinkedIn leaning on richer professional signals (company, role, seniority, skills, industry) than a consumer platform can.

How to build one today

The flow lives in LinkedIn Campaign Manager and shifts over time, so treat this as the shape rather than exact button labels:

1. Create the data source first — a matched audience from a customer/CRM list, a website-retargeting audience, a Lead Gen Form audience, or a conversions audience.
2. Make sure that source meets the minimum match threshold and has finished processing (uploaded lists can take up to ~48 hours to match against LinkedIn members).
3. Create a Predictive Audience and point it at that data source; LinkedIn builds the modeled audience.
4. Use it as your targeting in a campaign, and refine sparingly (see the warning below about over-layering).

The seed data is the whole game

This hasn't changed and never will: the quality of what you feed in dictates the quality of what comes out. LinkedIn needs a minimum of 300 matched members to build from a list, and because not every email matches a LinkedIn account, your raw list has to be comfortably larger than 300 (the share that matches is your match rate). Strong seeds, roughly best to worst:

• A list of your top revenue-generating customers (highest quality — but watch that it's big enough to match 300+).
• A handpicked list of target accounts from sales.
• Closed-won contacts exported from your CRM (e.g. Salesforce/HubSpot).
• People who hit a high-intent page (pricing, demo, login).
• Leads who converted on a real offer (an eBook download, a webinar signup).

Quality and size pull against each other: your most valuable-customer list is the best signal but often too small to clear the threshold. Blend for quality and volume.

Best practices that still hold

• Test more than one seed. A website-sourced audience and a list-sourced audience will behave differently — run both and let performance decide.
• Match the seed to the funnel stage. A broad, awareness-stage seed for top-of-funnel; a converters/customers seed for lower-funnel efficiency.
• Don't over-layer targeting. Every extra filter you stack on top of a predictive audience shrinks it and undercuts the modeling. Let the audience be the targeting.
• Read the demographics report. It tells you which companies, roles and industries are actually engaging, so you can refine the next seed.

Does it work for B2B?

This is where LinkedIn earns its (considerable) premium. Lookalike-style targeting has always been strong for B2C on Meta and weaker for B2B, because consumer platforms model lifestyle signals, not careers. LinkedIn models the professional context directly — title, seniority, company, industry, skills — which is exactly what B2B targeting needs. If your buyer is defined by their job, a well-seeded LinkedIn Predictive Audience is one of the better tools you have.

Running LinkedIn at any real scale rewards hands-on management — seeds, tests, demographics, exclusions. If you'd rather have seniors do that, see how we run paid social.