Lookalike audiences are one of the most powerful targeting tools available to digital advertisers. Instead of manually defining who should see your ads based on age, interests, or demographics, you let the algorithm find new people who closely resemble your best existing customers. The result: you reach cold audiences who are far more likely to convert than a generic interest-based segment.
This guide explains exactly how lookalike audiences work, how to build them correctly on Meta and Google, what seed audiences produce the best results, and how to avoid the most common mistakes that burn ad spend.
What Are Lookalike Audiences?
A lookalike audience is a targeting segment created by an ad platform that finds new users who share characteristics with a source group you provide — your "seed" audience. The platform analyzes hundreds of signals from the seed group (purchase behavior, browsing patterns, app usage, content engagement, demographic patterns) and builds a statistical model of what those people look like. It then searches its broader user base to find people who match that model.
The key idea is that people who behave similarly to your existing buyers are more likely to become buyers themselves. Lookalike targeting systematically surfaces those people without requiring you to manually predict who they are.
Meta popularized the term "Lookalike Audience" and it remains the most well-known implementation, but Google, TikTok, Pinterest, Snapchat, and LinkedIn all offer equivalent products under different names (Similar Audiences, Audience Expansion, Predictive Audiences, etc.).
How Lookalike Audiences Work
The process follows three phases:
1. Seed audience ingestion. You upload or select a source: a customer list, a website custom audience (pixel-based), an app event audience, or an engagement audience. The platform processes the seed, identifies the people in it, and builds a behavioral and demographic profile.
2. Model training. The platform's ML model looks at hundreds of data points for seed members: what they buy, what content they engage with, what apps they use, when they're online, how they respond to ads, and many signals that aren't disclosed publicly. It finds the statistical pattern that distinguishes seed members from non-members.
3. Audience expansion. The model scores every eligible user in the platform's database and selects the top-scoring users who aren't already in your seed. Those users become your lookalike audience. You typically control audience size via a percentage slider (1%–10% on Meta) or a size selection, which determines how closely the lookalike matches the seed versus how broadly it expands.
The audience is refreshed periodically (Meta updates lookalikes every 3–7 days) as the platform's data changes and new users join or leave the pool.
Types of Lookalike Audiences
Not all lookalikes are equal — the type of seed audience you use determines how well the resulting lookalike performs.
Purchase-based lookalikes use a list of people who have already bought from you. This is the gold standard seed because it directly tells the algorithm what a converting customer looks like. The more purchase events in the seed (ideally 1,000–5,000), the stronger the model.
High-value customer lookalikes (Value-Based Lookalikes on Meta) go one step further: instead of treating all buyers equally, you weight seed members by their lifetime value or average order value. The algorithm then tries to find people who resemble your highest-spending customers, not just any buyer.
Website visitor lookalikes are built from pixel-based audiences — people who visited key pages (product pages, checkout, pricing) without necessarily buying. Useful when you don't yet have enough purchase data to build a robust buyer seed.
Engagement lookalikes use people who have engaged with your content: watched videos, interacted with your Facebook Page or Instagram profile, or opened your lead form. These tend to be weaker seeds than purchase data but can work for awareness campaigns.
Email list lookalikes upload your CRM contacts (hashed email addresses) as the seed. The platform matches them to profiles and builds the lookalike. Quality depends heavily on how clean and recent your email list is.
Lookalike Audiences on Meta (Facebook and Instagram)
Meta's lookalike audience tool is the most sophisticated and widely used. Here's how to use it effectively:
Creating a lookalike in Ads Manager: Go to Audiences → Create Audience → Lookalike Audience. Select your source, choose the country or region you want to target, and select the audience size (1%–10% of the population in that country). A 1% lookalike is the most similar to your seed; 10% is larger but less precise.
Seed size matters: Meta recommends seeds of 1,000–50,000 people. Below 100, the model doesn't have enough signal to be reliable. Above 50,000, you're often including too many people who aren't genuinely your best customers, which dilutes the model.
Value-Based Lookalikes: When uploading a customer list, include a "value" column representing LTV or purchase amount. Meta uses this to build a lookalike that prioritizes finding high-value buyers. This feature is available when using a customer list with value as the source and selecting "Create a value-based lookalike" during setup.
Stacking lookalikes: You can target multiple lookalike percentages in the same ad set (e.g., 1%–3%) to balance precision and scale. Alternatively, create separate ad sets for 1%, 2–4%, and 5–7% to test which range performs best for your product.
Advantage+ Lookalikes: Meta's Advantage+ campaign type now uses algorithmic audience selection by default and treats lookalikes as a signal rather than a hard constraint. For many advertisers, this Advantage+ approach outperforms manually-defined lookalike targeting because it gives the algorithm more flexibility to find converters.
Lookalike Audiences on Google Ads
Google offers lookalike-style targeting under the name Similar Segments (previously Similar Audiences). These are auto-generated by Google based on your existing remarketing lists and Customer Match lists.
How Similar Segments work: Google analyzes the search behavior, YouTube activity, and browsing patterns of users in your remarketing lists and finds new users with similar patterns. Similar Segments are automatically created when your remarketing list meets minimum size requirements (typically 500+ users for Search, 1,000+ for Display).
Where to use them: Similar Segments work across Search campaigns (as observation or targeting), Display Network, YouTube, and Gmail campaigns. They're most effective for Display and YouTube, where broad reach matters more.
Optimized Targeting: Google's Optimized Targeting feature (available in Display and Video 360) functions similarly to lookalike targeting — it automatically expands your audience beyond your defined segments to find users likely to convert, using conversion signals from your account.
Performance Max: Google's Performance Max campaigns use audience signals (which can include your customer lists and website visitors) as seeds for their algorithm, effectively building lookalikes automatically across all Google inventory.
Building the Best Seed Audience
The seed audience is the single most important factor in lookalike performance. A poor seed produces a poor lookalike, regardless of budget.
Use purchasers, not visitors. If you have enough purchase data (aim for 500–1,000 minimum, 2,000+ ideal), use buyers as your seed. Visitors include window shoppers and people who bounced immediately — they produce noisier lookalikes.
Filter to your best customers. Don't upload all buyers — upload your repeat buyers, high-AOV buyers, or customers who haven't returned within 30 days. This focuses the model on the pattern of your most valuable customers rather than all customers.
Keep the list fresh. Upload customer data from the last 90–180 days, not all-time. Buying behavior patterns shift, and older data may reflect a customer type that's less relevant to your current business.
Match rate matters. When uploading email lists, include as many identifiers as possible: email, phone number, first name, last name, city, state, country, date of birth. More identifiers = higher match rate = larger, more accurate seed on the platform side.
Exclude existing customers from targeting. When you run ads to a lookalike, exclude your existing customers and warm audiences (website visitors, engaged followers). You don't need to pay to reach people who already know you with prospecting ads.
Lookalike Audiences vs. Interest Targeting
Many advertisers wonder whether to use lookalike audiences or interest-based targeting. The short answer: lookalikes usually win at conversion, but the comparison is nuanced.
Lookalike audiences are data-driven: the model is trained on actual buyers or engagers, so it captures subtle behavioral patterns that no interest category can replicate. They tend to produce higher ROAS and lower CPAs for mid-to-large advertisers with enough seed data.
Interest targeting is useful when you lack enough first-party data to build a meaningful seed. New businesses, new product categories, or new market expansions often need interest targeting as a starting point before they accumulate enough conversion data.
Combined approach: Some advertisers layer interests on top of lookalikes (AND logic) to narrow the audience. This can improve relevance but reduces scale significantly. Test both with and without the interest layer to see which produces better results for your specific offer.
In 2024–2025, Meta's Advantage+ and Google's broad-match/optimized targeting have made the distinction less important: modern campaign types use your conversion data to find lookalike-style audiences automatically, even if you don't explicitly create one.
Common Mistakes to Avoid
Using too small a seed. A seed of 50 people gives the algorithm almost nothing to work with. The resulting lookalike will be essentially random. Wait until you have at least 500 purchase events, or use a broader proxy (email subscribers, Add-to-Cart events) as an interim seed.
Targeting too broadly. A 10% lookalike in a large country is an enormous audience — often tens of millions of people. At this size, the lookalike characteristic is diluted to the point where it barely outperforms random targeting. Start at 1%–2% and expand only after you have strong performance data.
Not refreshing customer lists. Customer list-based lookalikes are only as current as your last upload. Set a recurring process (monthly at minimum) to re-export and re-upload your buyer list so the seed stays fresh.
Forgetting audience overlap. If you're running multiple lookalike audiences in separate ad sets (1%, 2–5%, etc.), they may heavily overlap, causing you to bid against yourself. Use the Audience Overlap tool in Meta to check, or consolidate overlapping segments into a single ad set.
Ignoring post-iOS 14 signal loss. Apple's App Tracking Transparency (ATT) framework reduced the quantity and accuracy of pixel-based event data for iOS users. This directly affects the quality of website-based lookalike seeds. Mitigate by using server-side events via the Meta Conversions API, which doesn't rely on browser-based pixel tracking and produces higher-quality purchase signals for seed building.
Shopify Audiences: Lookalike Targeting Powered by Shopify's Network
If you run your store on Shopify, there's a purpose-built lookalike product worth knowing about: Shopify Audiences. It takes a fundamentally different approach to audience building compared to standard platform lookalikes — and for eligible merchants, it often outperforms both Meta and Google's native tools.
What Shopify Audiences is. Shopify Audiences is a first-party audience product built into the Shopify platform. Instead of using only your store's own customer data as the seed, it leverages aggregated, opt-in purchase signals from across the entire Shopify merchant network — millions of verified buyer transactions. From this collective dataset, Shopify builds high-intent audience lists of users likely to purchase products in your category, then lets you export those lists directly to your ad accounts on Meta, Google, Pinterest, TikTok, Criteo, and Snapchat.
Why it outperforms standard lookalikes. A standard Meta lookalike is only as good as your seed — and for small-to-mid size stores, that seed is limited to your own customer base. Shopify Audiences enriches the seed with cross-merchant purchase data, which means even a store with 500 buyers benefits from a model trained on a vastly larger and higher-quality signal pool. Shopify reports that merchants using Shopify Audiences typically see 2–5x lower CPAs compared to standard interest or lookalike targeting.
Eligibility and setup. Shopify Audiences is available to merchants on the Shopify Advanced or Shopify Plus plan in the US and Canada (with other markets being added). To use it: go to your Shopify Admin → Marketing → Audiences, connect your ad accounts (Meta and/or Google), and generate an audience. Shopify creates the audience list and pushes it directly to your connected ad platform. You can then target it in a new prospecting campaign just as you would a standard custom or lookalike audience.
Shopify Audiences vs. Meta Lookalike: when to use each. Use Shopify Audiences as your primary cold prospecting tool if you're on Advanced or Plus — especially if your own customer base is smaller than 2,000 buyers. Use a Meta purchase-based lookalike as a complement or fallback if you're not eligible for Shopify Audiences, or if you want to test both in parallel. Many advertisers run both simultaneously in separate ad sets to compare CPAs, then reallocate budget toward the better-performing segment.
Automating Shopify lookalike campaigns with Adwisely. Adwisely integrates with Shopify to automate prospecting and lookalike campaigns — using your store's purchase data to continuously refresh and optimize audience targeting across Meta and Google, so you don't need to manually rebuild audiences as your customer base grows.
Frequently Asked Questions
How many people do I need in my seed audience?
Meta recommends 1,000–50,000 people for optimal results. You can technically create a lookalike from as few as 100 people, but quality degrades significantly below 500. For purchase-based seeds, aim for at least 500–1,000 buyers. If you don't have that yet, use a proxy like Add-to-Cart events or email subscribers as an interim seed while you accumulate more purchase data.
Does lookalike audience targeting work for small businesses?
Yes, but you need enough data first. If you're a new business with fewer than 100 purchases, start with interest targeting to build your initial customer base. Once you have 500+ buyers, switch to lookalike audiences as your primary prospecting method. Many small e-commerce businesses find that a purchase-based 1% lookalike dramatically outperforms interest targeting once they have sufficient seed data.
How often does Meta update lookalike audiences?
Meta automatically refreshes lookalike audiences every 3–7 days. However, if your seed is based on a customer list (uploaded CSV), the lookalike will only update when you re-upload a new list. Pixel-based and engagement-based seeds update automatically as new people qualify.
What's the difference between a 1% and 5% lookalike?
A 1% lookalike contains the users most similar to your seed — highest match scores — making it the smallest but most precise audience. A 5% lookalike is five times larger but includes users who are less closely matched to your seed. In practice, 1%–2% lookalikes tend to deliver the best conversion rates; 3%–10% are better suited for awareness campaigns or when you need more scale and are willing to accept a slightly higher CPA.
Can I use lookalike audiences for retargeting?
Technically, lookalike audiences are a prospecting tool — they target new people who don't already know you. Retargeting targets people who have already interacted with your brand. That said, you can use retargeting audiences (cart abandoners, product viewers) as the seed for a lookalike, which lets you prospect for new users who resemble your highest-intent visitors. This "retargeting lookalike" approach often outperforms a simple all-visitors lookalike because the seed is highly qualified.