Google's documentation confirms that lookalike audiences in Demand Gen campaigns now function as suggestions during a phased transition, which allows delivery outside the similarity thresholds advertisers select. Advertisers can apply to retain strict targeting.

The transition began in March 2026. This is not a new October policy, and the October reporting that surfaced it does not describe a change that happened this month.

If you have not checked your audience reporting since spring, the practical question is whether your campaigns have been delivering more broadly than you configured them to.

What the Transition Actually Does

Google's guidance describes lookalike segments shifting from a targeting constraint to an input signal. A campaign set to a narrow similarity threshold can serve impressions outside that threshold if Google's system identifies other users it expects to convert.

The reporting label changes accordingly. Audience segments operating as signals appear differently in your audience reporting than segments operating as strict targets.

Advertisers who need strict targeting preserved can apply for it. Google has not published approval criteria or typical response times in the reviewed documentation.

Why This Shows Up as an Acquisition Quality Question

Broader delivery usually produces more conversions at a lower reported cost per acquisition. Whether those conversions represent the customers you were trying to reach is a separate measurement.

A lookalike built from high-value repeat purchasers has a specific purpose. If delivery expands beyond that similarity threshold, the campaign may find cheaper conversions from a different kind of buyer.

That can be fine. A lower-value customer acquired cheaply still contributes. The problem is when reported performance improves while contribution per customer falls, and the account reads as healthier than it is.

Check cohort behavior rather than campaign-level cost per acquisition. Compare repeat purchase rate, average order value, and return rate for customers acquired before and after the transition reached your account.

What to Check in Your Account

Open audience reporting and look for the signal label on your lookalike segments. That tells you whether a given segment is operating as a constraint or an input.

Pull conversion data by month going back to February. If delivery expanded, the shift should appear as a change in volume and cost that does not match any budget or creative change you made.

Then segment by customer value rather than conversion count. The relevant comparison is what those customers did after the first order, not how many of them converted.

If the quality difference is material, applying for strict targeting is the available remedy. If it is not, the broader delivery may be doing useful work.

The Pattern This Fits

Platforms have been converting advertiser controls into inputs for several years. Amazon's Full-Funnel Campaigns consolidate budget allocation across formats with different purchase intent. Meta has reduced manual placement controls.

The common structure is that the platform's system optimizes against its own objective, and the advertiser's configured constraint becomes something the system considers rather than obeys.

That is not automatically worse. Automated systems frequently find conversions manual targeting misses. It does mean the advertiser's ability to verify what the system is doing depends on reporting the platform chooses to provide.

For Demand Gen specifically, the audience signal label is the available verification. Whether it is sufficient depends on how much your acquisition strategy relies on reaching a specific customer profile rather than maximizing conversion volume.