Lookalike audience
Give an ad platform a list of your best customers and it finds strangers who behave like them.
See it
What it is
You hand an ad platform a seed audience (a customer email list, purchasers from the pixel, high-value trial users), and it builds a new audience of strangers whose behavior patterns resemble that seed. Meta popularized the name and lets you pick a size from 1% to 10% of a country's users: 1% is tightest and most similar, 10% is broad and closer to plain interest targeting.
It is the standard way to scale cold prospecting once retargeting has eaten its small warm pool. Seed quality beats seed size: 500 actual repeat buyers usually outperform 50,000 newsletter subscribers, because the model copies whatever signal you feed it.
Gotcha: seeding on 'all website visitors' produces a lookalike of people who browse, not people who pay. Seed on the outcome you want more of, refresh the list every quarter as your customer base drifts, and exclude the seed itself from delivery so you are not paying to reach existing customers. Broad automated targeting (Advantage+, Performance Max) now often matches or beats hand-built lookalikes, so test one against the other before assuming.
Ask AI for it
Design a lookalike audience strategy for [platform] in [country], using eligible seed counts [counts]. Pick three seed sources ranked by signal strength (repeat purchasers, single purchasers, high-intent page viewers) and give a refresh cadence for each. Recommend only the audience tiers and seed minimums that platform actually supports in that market today, say explicitly when a seed is too small to use, and never quote a threshold you cannot point to in current platform documentation. Run each tier as its own ad set so results stay readable. Add exclusion rules so existing customers and current retargeting audiences never see these ads, and state the metric that decides a winner (cost per purchase, not CTR).