In the evolving landscape of digital advertising, the line between "targeted advertising" and "algorithmic discovery" is blurring. Recent scrutiny regarding Google’s Demand Gen campaigns has brought a critical update to the forefront: lookalike audiences are no longer rigid targeting constraints. Instead, they have been transitioned into "input signals," a change that allows Google’s AI to deliver ads well outside the similarity thresholds once set by advertisers.
For performance marketers, this shift represents a fundamental change in how budget is allocated and how audiences are defined. While many media outlets recently highlighted this as a new development surfacing in October, the reality is more nuanced: this transition began as a phased rollout in March 2026. For advertisers who haven’t audited their audience reporting since the spring, the critical question is no longer "who are we targeting?" but rather "who is Google showing our ads to?"
Understanding the Architectural Shift: From Constraint to Signal
To understand the current state of Google Demand Gen, one must first understand the distinction between a "constraint" and an "input signal."
Historically, when an advertiser uploaded a customer list to create a lookalike audience, they were effectively drawing a fence around a specific demographic or behavioral cluster. The platform was instructed to serve impressions exclusively to users who shared high-degree similarities with the source list.
Under the new paradigm, that fence has been removed. Google’s documentation now clarifies that lookalike segments serve as an "input signal." This means that while the system considers your lookalike parameters to identify high-potential candidates, it is explicitly empowered to deliver impressions to users outside of those similarity thresholds if the machine learning algorithm predicts a high probability of conversion.
The Mechanism of Delivery
When a campaign is set to a "narrow" similarity threshold, the system is no longer prohibited from reaching beyond that scope. If Google’s AI identifies a user who does not fit the lookalike profile but exhibits behaviors—such as search history, site engagement, or purchase intent signals—that suggest they are likely to convert, the ad will serve.
This change is not just theoretical; it is reflected in the reporting dashboard. Audience segments operating as signals appear differently in Google Ads reporting than those operating as strict targets. Advertisers must now pay close attention to these labels, as they indicate the degree of autonomy the platform is exercising over their targeting logic.
Chronology of the Transition
Contrary to recent industry buzz, this is not a sudden, mid-autumn policy change. The transition has been a deliberate, long-term migration of Google’s advertising infrastructure.
- Early 2026: Google begins the phased rollout of audience segments shifting toward an AI-driven "signal" model.
- March 2026: The official start of the transition period, marking the move away from rigid lookalike constraints.
- Spring 2026: Advertisers begin to notice shifts in delivery, though many attribute it to natural fluctuations in market demand or creative fatigue.
- October 2026: A surge in industry reporting brings the change into the spotlight, leading to a broader realization among the PPC community that their historical targeting settings may no longer be functioning as expected.
This timeline is crucial for account audits. If an advertiser compares their campaign performance against Q1 2026 data, they are essentially comparing two different operating models—one where the advertiser held the reins, and one where the algorithm dictates the scope of the audience.
The Quality Paradox: Acquisition Costs vs. Long-Term Value
The most significant implication of this transition is the potential for a "false positive" in campaign performance. When a system is allowed to broaden its delivery, it frequently identifies "low-hanging fruit"—users who are cheaper to convert but may not align with the brand’s ideal customer profile.
The "Cheap Conversion" Trap
Broader delivery often results in a decrease in the reported Cost Per Acquisition (CPA). On a surface-level dashboard, this looks like a win: the campaign is more efficient, and the system is finding more customers. However, the quality of these customers is the true test.
If a lookalike audience was built from high-value repeat purchasers, the goal was to acquire more of those high-value individuals. If the algorithm expands delivery to find cheaper conversions among a different, lower-value cohort, the CPA will drop, but the contribution margin per customer may plummet. This creates a scenario where the account appears healthier than it actually is, masking a decline in long-term customer lifetime value (LTV).
Assessing True Performance
To reconcile these discrepancies, advertisers must shift their focus away from campaign-level CPA. Instead, the focus should move toward cohort analysis:
- Repeat Purchase Rate: Are the customers acquired in the last six months coming back at the same rate as those acquired before the transition?
- Average Order Value (AOV): Has the basket size changed significantly?
- Return Rate: Are the "broader" audiences resulting in higher product return rates, suggesting a mismatch between the product and the customer?
Official Stance and Remedies
Google has acknowledged the shift and, for those who find the algorithmic expansion detrimental to their specific goals, has provided a potential remedy: the ability to apply for "strict targeting."
However, the process is not automated or transparent. Advertisers must apply to have strict targeting preserved, and Google has not provided clear documentation regarding the criteria for approval or the typical response times for these requests. This leaves many marketers in a state of limbo, forced to choose between the efficiency of AI-driven optimization and the control of manual targeting.
For brands with highly specific acquisition strategies—such as luxury retailers or niche B2B service providers—the "broadening" of audiences can be disastrous if not monitored. The lack of clarity around the approval process for strict targeting suggests that Google is firmly committed to a future where automation is the default.
Broader Industry Context: The End of Manual Control
The shift in Google Demand Gen is not an isolated incident; it is part of a systemic trend across the entire digital advertising ecosystem.
- Amazon: Through their "Full-Funnel" campaigns, Amazon has consolidated budget allocation across various formats, effectively reducing the advertiser’s ability to micromanage spend based on intent.
- Meta: The platform has systematically removed manual placement controls and audience targeting granularities, pushing advertisers toward Advantage+ campaigns that rely entirely on the platform’s ability to "find the right person."
The industry pattern is clear: platforms are converting advertiser controls into mere inputs. The system optimizes against its own objective—usually total conversion volume or ROAS—and the advertiser’s configuration becomes a suggestion rather than a mandate.
Strategies for Advertisers: What to Do Now
For those managing Demand Gen campaigns, immediate action is required to ensure that algorithmic expansion is not negatively impacting the bottom line.
1. Perform an Audience Audit
Log into your Google Ads account and review your audience reporting. Look for the "signal" label on your lookalike segments. This will immediately clarify which segments are currently operating as constraints and which have been converted into signals.
2. Historical Data Comparison
Pull conversion data by month, beginning in February 2026. Look for anomalies in volume and cost that do not correlate with budget or creative changes. If you see a sudden spike in volume accompanied by a lower CPA, investigate the quality of those cohorts.
3. Shift to Value-Based Measurement
Do not optimize for conversion count alone. Segment your data by customer value. If your business relies on repeat purchases, look at the post-purchase behavior of customers acquired post-March 2026. If the quality difference is material, you have a strong case for applying for strict targeting.
4. Evaluate the Utility of Broad Delivery
Not all broadening is bad. In many cases, automated systems can identify high-intent users that human intuition would miss. If the broader delivery is producing profitable, high-quality customers, it may be worth embracing the change. The goal is not to fight automation, but to verify that it is serving your business objectives.
Conclusion
The transition of lookalike audiences from constraints to inputs is a milestone in the "black-boxing" of digital advertising. While the change promises greater efficiency and broader reach, it demands a higher level of analytical rigor from advertisers.
In the era of AI-driven marketing, the advertiser’s role has fundamentally changed. We are no longer the architects of the audience; we are the auditors of the algorithm. By focusing on cohort behavior, challenging the "low CPA" narrative, and actively monitoring signal labels, marketers can ensure that their campaigns remain a tool for growth rather than a casualty of platform-wide automation.
