Create Facebook Lookalike and Engagement Audiences With AI

Retargeting reaches people who already know you. Lookalike audiences help you find new people who resemble them. You give Meta a starting group, such as your buyers, and Meta finds others with similar patterns. Engagement audiences are the other half of the story: people who interacted with your Facebook Page, Instagram profile or videos.

This guide covers both, using AdChat's create_lookalike_audience and create_engagement_audience tools.

Who this is for

Advertisers who have some existing customers, website visitors or social followers and want to reach more people like them.

What you need

• Write access turned on in your AdChat dashboard. Creating audiences is a change.
• A source audience for lookalikes: a custom audience in your ad account, such as a website audience (see retargeting website visitors) or a list you uploaded in Ads Manager.
• AdChat connected to your AI app or the built-in chat.

How lookalikes work

You pick three things:

1. Source: the custom audience Meta copies from. Meta needs at least 100 people from the target country in it.
2. Country: where Meta should look for similar people.
3. Size: a percentage of that country's population. 1% is the closest match and smallest group. Larger percentages reach more people but are less similar.

The better your source, the better the lookalike. Buyers usually make a stronger source than "anyone who visited the homepage".

Step 1: Pick your source

List my custom audiences with their type, approximate size and when they were created. Which would make the best source for a lookalike if my goal is more purchases?

What a good answer looks like: your audiences in a table and a recommendation with a reason. For example: "Purchasers 180d is the strongest source because it's built from actual buyers and has enough people."

Step 2: Create the lookalike (write access on)

Create a 1% lookalike audience in the United States from my "Purchasers 180d" audience. Call it "LAL 1% Purchasers US". Show me the details before you create it.

You can also create a wider band to test later:

Create a lookalike from "Purchasers 180d" in the US covering the 1% to 3% range, called "LAL 1-3% Purchasers US".

AdChat lists what it's about to create and asks first. Lookalikes can take a while to fill after creation, sometimes a few hours.

Step 3: Build engagement audiences

Engagement audiences are people who interacted with your content on Meta. They're handy if your website traffic is low but your social following is active. AdChat supports three sources:

• Facebook Page: people who engaged with your Page, visited it, interacted with posts or clicked a call-to-action button.
• Instagram business profile: people who interacted with your profile, visited it or sent you a message.
• Videos: people who watched any amount of a video, or 25%, 50%, 75% or 95% of it.

People can stay in an engagement audience for up to 365 days.

Find my Facebook Pages and Instagram account, then create an engagement audience called "IG Engagers 90d" of everyone who interacted with my Instagram business profile in the last 90 days.

For video viewers:

Create an engagement audience called "Video 50% Viewers 60d" of people who watched at least half of the video in my ad "Founder Story", over the last 60 days.

Step 4: Use the new audiences

Engagement audiences work for warm retargeting. They also make good lookalike sources once they're large enough.

Create a paused ad set in my "Prospecting" campaign targeting "LAL 1% Purchasers US", ages 21 and over, with a $20 daily budget. Exclude "Purchasers 180d" so we don't pay to reach existing customers. Show me the setup first.

New ad sets start paused. Check the setup, add an ad, then ask the AI to turn it on.

Step 5: Compare after a week or two

Compare my lookalike ad sets with my interest-based ad sets over the last 14 days: spend, results, cost per result and CTR. Which audience type is working better?

CTR (click-through rate) is the share of people who saw an ad and clicked it. Cost per result (often called CPA) is what you pay for each lead or sale. Judge by cost per result first.

What to do next

• Check the size of a new lookalike before launching: audience size estimation.
• Combine with interest targeting ideas from audience targeting.
• Give your best new audience more budget once it proves itself: budget optimization.

Common mistakes

• Using a weak source. A lookalike of random visitors finds random people. Use buyers or high-intent actions where you can.
• A source that's too small. Meta needs at least 100 people from the target country. More usually works better.
• Not excluding existing customers. Otherwise you pay to reach people who already bought.
• Expecting instant results. Lookalikes take time to fill, and new ad sets need time to leave Meta's learning phase, the bumpy period after launch.
• Testing many lookalike sizes at once with a small budget. Start with 1% and expand once it works.

FAQ

Can AdChat upload my customer email list to build a lookalike? No. Customer list uploads aren't supported. Upload the list in Ads Manager, then AdChat can use it as a source like any other custom audience.

Can I create lookalikes in several countries at once? Create one per country by asking for each. Keeping them separate also makes results easier to compare.

What's a good starting size? 1% is the usual starting point. It's the closest match to your source.

Can I delete old audiences? Yes, with write access on. Deleting can't be undone, so AdChat asks first.

Try it free for 7 days. Create your account and build your first lookalike today. Questions? Email [email protected].