Short answer: Google is changing how Lookalike segments work in Demand Gen campaigns. Instead of a hard boundary, a Lookalike will become a suggestion by default, so Google can show your ads to people outside that audience when it predicts they will convert. If you want the old behaviour, you can opt out and keep the Lookalike as a strict targeting constraint. The audience sizes stay the same: 2.5%, 5% and up to 10% of users in your chosen location.
This sounds like a small setting, but it changes who sees your ads and how you should read your results. Below I explain what is changing, who should opt out, and how to set up a clean test before you decide.
What exactly is changing
Today, when you add a Lookalike segment to a Demand Gen ad group, Google only shows ads to people inside that modelled audience. After the change, the Lookalike becomes a starting signal. Google’s system uses it to find likely converters first, then expands beyond it when it thinks performance will be better.
| Before | After (default) | |
|---|---|---|
| How the Lookalike is used | Strict targeting, ads stay inside the audience | Suggestion, Google can go beyond it |
| Who controls reach | You | Mostly Google’s AI |
| Can you keep the old way? | Yes | Yes, by opting out to a targeting constraint |
| Audience sizes | Narrow 2.5%, Balanced 5%, Broad up to 10% | Same |
Google has not published a single switch date for every account, so check your Demand Gen ad groups now rather than waiting for a notice.
How Lookalike segments work (quick refresher)
- Seed lists: Lookalikes are built from your first party data, such as customer lists, website visitors, app users and YouTube audiences.
- Minimum size: the seed needs more than 100 active matched users across your lists.
- Refresh: Google updates the modelled audience every one to two days.
- Sizes: Narrow reaches the most similar 2.5% of users in the location, Balanced 5%, and Broad up to 10%.
Because the seed is your own data, the quality of your customer list matters more than ever. A list of real buyers will produce a better Lookalike than a list of every email you have ever collected. My guide on Data Strength in Google Ads explains how to improve that data.
Who should keep strict targeting
Opt out and keep the Lookalike as a constraint if any of these apply:
- Your offer is narrow. B2B services, high ticket products or local services where the wrong audience wastes budget fast.
- You have compliance rules. Some industries need tight control over who sees an ad.
- You run a planned audience test. If you are comparing a Lookalike against another audience, expansion makes the test meaningless.
- Your conversion tracking is weak. Expansion relies on conversion signals. If tracking is broken, Google will optimise towards the wrong people. Fix tracking first with my conversion tracking guide for GA4 and GTM.
Who should try suggestion mode
Suggestion mode can work well when:
- You have at least 30 to 50 conversions a month in the campaign, so the algorithm has enough signal.
- You sell to a broad consumer market, such as eCommerce, where good customers are not limited to one modelled audience.
- Your goal is scale, and you already judge results on cost per sale or return on ad spend, not on clicks.
This matches what we have seen on Meta, where broad targeting with strong signals often beats tight audiences. I covered that shift in Meta Ads for eCommerce in 2026.
How to test it properly
Do not just accept the new default and hope. Run a simple test:
- Duplicate one Demand Gen ad group, or use a campaign experiment if available in your account.
- Keep the same creative, budget and bidding in both versions.
- Set one version to strict targeting and leave the other on suggestion mode.
- Run it for at least three to four weeks, or until each side has a meaningful number of conversions.
- Judge on cost per acquisition and conversion value, and check the quality of leads or orders in your CRM or store, not only in Google Ads.
A note on Demand Gen attribution
Google is also pushing new attribution options for Demand Gen. At Google Marketing Live 2026 it announced Campaign Type Attribution, which aims to give Demand Gen credit for the full value of conversions it drives, including later sales, and to let you compare it with paid social platforms. Reports from October say it is in beta and tied to data driven attribution.
That is useful, but be careful. More attributed conversions do not prove more sales. When a campaign gets more credit in its own report, it can look better without adding any new customers. Before you scale Demand Gen, compare total sales or leads across all channels, look at new versus returning customers, and if your budget allows, run a geo or conversion lift test. My article on ROAS vs MER vs CAC explains how to look at the whole picture.
Quick checklist for this week
- List every Demand Gen ad group that uses a Lookalike segment.
- Decide for each one: strict constraint or suggestion mode.
- Opt out where you need control, before the default changes.
- Check that your seed lists are fresh and contain real customers.
- Confirm conversion tracking and enhanced conversions are working.
- Set a review date three to four weeks out.
Frequently asked questions
What are Lookalike segments in Google Ads Demand Gen?
What does suggestion mode mean for Lookalikes?
Can I keep strict Lookalike targeting?
How big does my seed list need to be?
Does Campaign Type Attribution mean Demand Gen is working better?
Who can help me set up Demand Gen campaigns?
Need help with Demand Gen?
I can review your Demand Gen audiences, tracking and attribution and tell you which ad groups should stay strict. See my Google Ads management service, or request a free ads audit.