In a recent email to Google Ads administrators, Google announced an upcoming update to Demand Gen Lookalike segments. Google has begun phasing out traditional Lookalike audiences within Demand Gen campaigns, and from March 2026, Lookalike segments won’t act as hard targeting boundaries anymore. Advertisers who previously relied on manually built Lookalike lists are now being guided toward automated targeting powered by Google’s machine learning systems.
This means less manual control over who your ads are shown to, and more reliance on Google’s algorithms to do the heavy lifting and decide which users resemble your existing customers. Over the last few years, Google has steadily reduced granular targeting controls in favour of AI-led optimisation. Demand Gen, built to drive awareness and conversion across YouTube, Discover, and Gmail placements, is now firmly part of that automated future.
The Shift Toward AI-Driven Targeting
Until now, when you set up a Lookalike audience in Demand Gen, you’d choose from three reach levels based on how similar you wanted your audience to be to your seed list:
- Narrow targeted roughly 2.5% of users in your location – the closest match to your existing customers
- Balanced reached around 5% – a middle ground between similarity and scale
- Broad extended to 10% – wider reach with less strict matching.
Those same reach levels still exist, but they’re no longer constraints. Google’s treating them as signals now – hints about your preferences rather than rules it must follow. If you’ve been using narrow Lookalike targeting to keep your ads tightly focused on highly similar users, Google might now reach beyond that group if it thinks doing so will improve your results. The system will use your seed list and selected reach level to prioritise conversions, but it can expand beyond those boundaries if its AI predicts better performance or lower cost per action.
The upside? AI can identify patterns beyond what manual audience building can achieve. It looks at behavioural signals, engagement trends, and intent signals at scale – more variables than a typical Lookalike list ever could. The trade-off? Reduced transparency (over who is being targeted) and less control over how far your ads can spread beyond your core audience.
Efficiency vs. Control
For businesses running Demand Gen campaigns, this update may initially feel like losing a layer of control (requires giving up some of the targeting precision you might be used to). But it’s also an opportunity to rethink where performance actually comes from. Performance will increasingly depend on:
- Strong first-party data: clean customer lists and accurate conversion tracking
- Clear conversion signals: well-defined goals for the algorithm to optimise toward
- High-quality creative: messaging that resonates across broader audience pools
- Ongoing testing: measuring incrementality rather than relying on surface metrics.
Google’s official line is that strict similarity thresholds can limit reach to qualified users that AI-driven targeting could otherwise find. This update is designed to improve conversion performance and help campaigns hit their goals more effectively. For some brands, this might improve efficiency. For others, it may require closer monitoring and stronger data foundations. If your tracking, data, and creative foundations aren’t solid, automation will likely amplify the gaps.
Optimised Targeting
If you’re already using Optimised Targeting alongside Lookalike audiences (and many advertisers are), you’re now running two overlapping AI-driven expansion systems simultaneously. Google’s documentation acknowledges this, noting that “ad groups with both Lookalike-as-a-suggestion and Optimised Targeting enabled will find higher performance gains being captured by Lookalike-as-a-suggestion mode.” This layering represents a significant shift toward letting Google’s algorithms make more decisions about who sees your ads, similar to what we’ve seen with Performance Max campaigns.
Can you opt out? Yes, but it requires action. Google offers a form-based opt-out – you’ll need to submit a request if you want to maintain the old behaviour.
What Happens Now?
Rather than focusing on what’s been removed, it’s more productive to focus on what you can control. Well-structured accounts with:
- Proper conversion tracking
- Clear value-based optimisation
- Strong creative differentiation
- Defined audience inputs
are more likely to benefit from AI expansion.
Here’s what we’d recommend:
- Audit your conversion tracking and attribution
- Review your seed lists – your seed list becomes the primary lever for shaping who the AI prioritises, so make sure it’s high-quality and represents your best-converting users
- Review creative to ensure it speaks to intent, not just demographics
- Monitor performance data closely during this transition period.
The Last Word
If you’re running Demand Gen campaigns with Lookalike audiences and you’re not sure how this change affects your strategy, we’re here to help. Get in touch with the Herdl team to review your PPC setup.