Before iOS 14.5, Facebook lookalike audiences were among the most powerful targeting tools available to digital advertisers. The system was straightforward: you uploaded a seed audience of your best customers or high-value converters, and Meta's algorithm analyzed hundreds of data points to find new users who shared similar characteristics, behaviors, and interests.
The pixel tracked everything. Every page view, every add-to-cart, every purchase was fed back into the system, creating rich behavioral profiles that powered precise audience modeling. Advertisers could confidently scale campaigns by creating 1%, 2%, or 5% lookalike audiences, knowing the algorithm had access to comprehensive user data across the web.
Similar audiences on Google Ads worked on comparable principles, leveraging browsing behavior, search history, and YouTube engagement to identify users who resembled your existing customers.
The rollout of iOS 14.5 in April 2021 fundamentally disrupted this model. Apple's App Tracking Transparency (ATT) framework required explicit user consent for cross-app tracking, and roughly 75-80% of iOS users opted out. This created massive signal loss for Meta and other platforms relying on pixel data.
Key impacts include:
Simultaneously, third-party cookie deprecation across browsers further eroded the data foundation that traditional lookalikes relied upon. Google officially sunsetted Similar Audiences in May 2023, acknowledging that privacy changes made this targeting approach unsustainable.
Lookalike audiences on Meta still exist, but their effectiveness has fundamentally changed. In 2026, pixel-based lookalikes built from website visitor behavior show significantly weaker performance compared to pre-iOS 14 benchmarks. Many advertisers report that traditional lookalikes now perform only marginally better than broad targeting.
Meta has responded by pushing advertisers toward Advantage+ audience targeting, which essentially allows the algorithm to expand beyond your selected audiences when it detects better opportunities. This represents a shift from advertiser-controlled targeting to machine-learning-driven optimization.
The platform now prioritizes first-party data inputs and encourages the use of Conversions API (CAPI) to supplement pixel data with server-side signals. Advertisers using Meta automation tools can better manage this complexity while maintaining consistent data flows.
Google Ads no longer offers Similar Audiences as a standalone targeting option. The platform has transitioned to audience expansion features within Performance Max and Demand Gen campaigns, where machine learning handles audience discovery automatically.
Customer Match remains the primary way to leverage your owned data on Google. You can upload hashed email lists, phone numbers, or mailing addresses, and Google will match these to signed-in users across Search, YouTube, Gmail, and Display. For best results with Performance Max campaigns, consider reviewing the PMax structure recommendations to understand how audience signals interact with asset groups.
Optimized targeting in Demand Gen campaigns serves a similar function to what lookalikes once provided, using your signals as a starting point while expanding to find likely converters.
Not all audience signals have degraded equally. In 2026, certain inputs consistently outperform traditional pixel-based approaches:
Tip: When building customer match lists, include multiple identifiers (email, phone, address) to maximize match rates. A single customer with three identifiers is more likely to be matched than three customers with one identifier each.
Several strategies have emerged to fill the gap left by degraded lookalike performance:
This allows Meta's algorithm to find converters without strict audience boundaries. When combined with strong creative and conversion optimization, it often outperforms legacy lookalike approaches.
Instead of targeting narrow audiences, many advertisers now use broad targeting while letting creative serve as the segmentation mechanism. Different ad variations naturally attract different user types.
Layering multiple relevant interests creates pseudo-lookalike audiences based on declared behaviors rather than modeled predictions.
Using engaged website visitors as a starting point, then allowing platform expansion features to find similar users within controlled parameters.
Leveraging feed management to serve personalized product ads to broad audiences, letting the catalog and algorithm work together to find relevant matches.
The advertisers seeing the best results in 2026 have fundamentally restructured their audience strategy around first-party data collection. This requires a mindset shift from relying on platform data to owning your customer intelligence.
Effective first-party data strategies include:
Tip: Create separate seed audiences for different customer cohorts—first-time buyers, repeat purchasers, and high-LTV customers will generate distinct lookalike models with different applications in your funnel.
