Should you exclude existing customers from PMax, and where should the lists live?

The decision is not whether Performance Max now has an exclusion toggle. It is whether that toggle does the job operators assign to it, and whether placement hygiene for PMax, Demand Gen and YouTube can live in one list without fragmenting the same signal three times.

Two changes in 2026 turned exclusions into an account design problem. In January, Google Ads began letting advertisers apply one placement exclusion list at account level across Performance Max, Demand Gen, YouTube and Display, instead of rebuilding the same blocks at campaign or ad group level. In March, first-party audience exclusions, documented as "Your Data Exclusions", arrived in PMax campaign settings. One control is account-scoped. The other is campaign-scoped. Treating them as the same lever is how lists proliferate, how existing customers still convert, and how Demand Gen keeps spending on inventory you thought you had already blocked.

Should existing customers leave Performance Max now the exclusion switch exists?

Most operators arriving at the new PMax setting are not looking for a feature tour. They already run a new customer acquisition goal, or they think they do, and they want a yes or no: should existing customers now be excluded from PMax because a first-party list can finally be attached?

No: not as the primary lever, and not as a substitute for the goal Google already documents for this job. Search Engine Land's March 2026 announcement framed first-party audience exclusions as a way to exclude customer lists and shift spend toward net-new acquisition rather than repeat conversions. That is the commercial point. Google's own campaign-settings documentation is narrower. Data exclusions only apply to Search, Display, Shopping, Video and Discover inventory within Performance Max campaigns. If the goal is specifically acquiring new customers, Google recommends the new customer acquisition goal instead, for additional reporting and value specification.

That gap, announcement language versus documented inventory scope, is the whole decision. A campaign checkbox that does not cover every PMax surface cannot be the account's definition of "existing customers off". The lifecycle goal can, because it is built for that objective and because it exists for Performance Max, Search, Shopping and Demand Gen rather than for one campaign type's settings panel.

What does a fragmented exclusion architecture actually cost?

When exclusions are rebuilt campaign by campaign, two costs show up immediately.

The first is media spent on people you already paid to acquire. If PMax is still converting existing customers while the rest of the account is trying to buy new ones, you are not looking at a creative problem. You are looking at a control that was applied in the wrong layer, or only on part of the inventory the campaign can use.

The second is operational duplication. Advertisers previously had to copy the same exclusion lists across multiple campaigns in an account. For Performance Max that duplication is especially expensive, because PMax does not allow manual placement selection, exclusion lists are the primary placement control. If Demand Gen and YouTube are governed by a different copy of the list, or by no list at all, you are not running one acquisition system. You are running three inventories with three opinions about what is allowed.

There is a quieter cost in measurement. A PMax campaign that still harvests returning customers will report a healthier ROAS than a campaign that has been pushed toward net-new acquisition, even when the second campaign is doing the job the business asked for. Practitioner coverage of net-new PMax work expects reported ROAS to fall while the new-versus-returning split aligns more closely with acquisition. That is an expectation, not a measured benchmark, but the directional warning is useful: if reported ROAS is the number you manage to, you will read a cleaner mix as a failure. How to read blended media efficiency without losing the plot is the right adjacent discipline, read platform efficiency beside contribution, break-even CPA and customer movement, not as a verdict on the exclusion itself.

Brand safety is the other bill. Account-level placement exclusions exist because once a site, app or YouTube placement is excluded at that layer, Google Ads prevents spend on it across eligible campaigns. Leave that list at campaign level, or leave Demand Gen outside it, and the commercial cost is not theoretical. It is impressions on inventory you have already decided you do not want, paid for by a campaign type that never inherited the block.

Why does duplicating lists across campaigns still fail?

The default response to both 2026 releases has been to flip every new switch. That fails in three predictable ways.

First, operators copy the same placement list into every campaign, then wonder why Demand Gen still serves inventory they thought they had blocked. Demand Gen's FAQ is explicit: topic and placement exclusions are not generally available at campaign or ad group level, and keyword exclusions sit at account level under Content Suitability, inside excluded content keywords. Core suitability settings such as inventory type, content types and labels are also not available at campaign level. A Display habit of parking blocks on the campaign will not transfer. Account level is not a convenience for Demand Gen. It is often the only layer the campaign type will honour for shared inventory rules.

Second, operators treat PMax data exclusions as a perimeter around existing customers. Google's campaign-settings documentation does not support that reading. "Your Data Exclusions" allows exclusion of advertiser-provided audience remarketing lists from PMax targeting, and only on the inventory listed above. It does not rewrite the campaign type, and Google points advertisers who want new customers toward the lifecycle goal rather than toward this checkbox.

Third, stacking brand exclusions, audience exclusions and Customer Match is useful and still imperfect. Those controls can be combined to push PMax toward net-new acquisition. They still do not match perfectly: audience exclusions miss people, brand exclusions miss variations, and Customer Match has gaps. An exclusion list is not a legal fence. Planning as if it were one is how operators over-promise a clean new-customer mix and then distrust the campaign when unmatched buyers still convert.

The failed pattern underneath all three is the same: feature availability is treated as a decision. Availability tells you a control exists. It does not tell you which layer should own the job, or whether the control even applies to the inventory you are trying to govern.

Are you solving a bidding problem or an inventory problem?

The useful diagnostic is not "do we have the new feature on?" It is "which layer owns this decision?"

Customer lifecycle, whether you want new buyers, returning buyers, or a weighted mix, is a bidding-goal problem. Customer lifecycle goals exist for Performance Max, Search, Shopping and Demand Gen. New Customer Value mode is what Google recommends for advertisers with purchase conversion goals: it prioritises new customers while maintaining engagement with potential returning customers, and it lets you specify additional value. New Customer Only mode bids exclusively for new customers, and Google advises a separate campaign to reach existing customers. Demand Gen documents the same fork in its own language: optimise towards new users either by targeting them exclusively or by bidding higher for them, and use audience exclusions from first-party data such as website visitors, app users and customer lists when the campaign must focus on new-customer acquisition only.

Audience data exclusions are a different class of tool. They are campaign-scoped and inventory-partial inside PMax. Hire them when a list must be kept off those documented PMax surfaces even if bidding would still find the people valuable. Do not hire them as a synonym for "new customers only". The public record in this set does not say whether data exclusions and the new customer acquisition goal can be combined without conflict. That is a genuine open question. Test it in a contained campaign rather than assuming the two controls stack cleanly.

Placement exclusions are an account-scoped override. Placement exclusions can be applied at account, campaign or ad group level, and account-level exclusions override campaign and ad group settings. Once excluded at account level, Google Ads prevents spend on those sites, apps or YouTube placements across eligible campaigns. Demand Gen cannot generally hold topic or placement exclusions at campaign level, so the account layer is the only consistent place a shared hygiene list can live across PMax, Demand Gen and YouTube.

If you diagnose a returning-customer problem as a placement problem, you will build enormous lists and still buy existing customers. If you diagnose a brand-safety problem as a PMax audience problem, Demand Gen will keep spending. Layer scope is the reframe; feature checklists are how the same list gets built three times.

This is also where judgement still sits with the operator rather than with the interface. How AI changes D2C research without replacing judgement makes the same point in a different desk: retrieval can surface the setting, but someone still has to own the question, the sources and the decision. Here the question is which commercial job the list is being hired to do.

How should one exclusion system be structured across PMax, Demand Gen and YouTube?

Start with existing customers, then place inventory hygiene, then respect the ceilings so the architecture does not collapse on upload.

Existing customers: goal first, data exclusion only for a hard block

Use New Customer Value mode when the business still wants some returning-customer engagement and you are trying to reweight the mix rather than seal it. Use New Customer Only mode when existing customers should not be bid for in that campaign at all, and stand up the separate existing-customer campaign Google advises. That separate campaign is not a reporting trick. It is how you keep lifecycle demand from leaking back into the prospecting vehicle. There is no published fencing pattern that should be treated as a rule. In practice, not as a lock, run the existing-customer work in a separate campaign on a separate budget and review the shared lists. Until that is verified in-account, assume leak is possible. The owned-side analogue is the same sequence: earn the relationship after acquisition before you ask retention mechanics to do prospecting's job, as Which ecommerce lifecycle flow should you fix first? argues for welcome and post-purchase.

Demand Gen can take a parallel audience exclusion: exclude users from first-party data so the campaign focuses on new-customer acquisition. Lookalike segments should sit in separate ad groups. That is campaign-level audience work, not a replacement for the shared placement list.

Reserve the PMax data exclusion for lists that must not see Search, Display, Shopping, Video or Discover inventory inside PMax for a reason bidding cannot express, a commercial or legal hard block, a remarketing list you would never want on those surfaces even at a lower bid. If you cannot write that reason down, you do not have a data-exclusion job. You have a bidding-goal job.

Expect the mix to look worse on ROAS if the exclusion and the goal do what you asked. Practitioner coverage expects reported PMax ROAS to fall as the new-versus-returning split improves. Whether paid media should optimise for ROAS or profit is the decision frame: ROAS can steer delivery inside the platform; profit decides whether the resulting mix is worth buying.

Placement lists: two tiers, not twenty campaign copies

Brand-safety and legal inventory blocks belong at manager-account level when they must travel with every client account. A manager account can have a maximum of three exclusion lists. Those lists can be applied across selected accounts from the Sub-account settings tab. Keep this tier small, slow-changing and boring: the domains, apps and YouTube channels that should never run, regardless of campaign type.

Performance pruning, placements that waste spend for this brand but would not be a policy incident, belongs at client-account level. Each client account can have up to 20 exclusion lists, and client-account exclusions can be applied across selected campaigns in the account. Use this tier for waste you are prepared to maintain for this advertiser.

Two tiers only hold if someone owns the maintenance. The list that survives is the one someone owns, usually the MCC brand-safety list. Client-account pruning dies unless it is on a weekly checklist.

Account-level exclusions override campaign and ad group settings. That is the point of putting hygiene there. Do not maintain a contradictory campaign-level list and hope it wins.

You cannot exclude the entirety of youtube.com as a placement, especially for PMax. Specific YouTube channels or videos can be added to the account-level list, and PMax will respect those account-level exclusions. URLs associated with a Google domain, including youtube.com and mail.google.com, cannot be added as placement exclusions. From March 2024, account- and manager-level site exclusions also apply to the search partner network across campaigns including Performance Max. Design the YouTube tier as channel and video hygiene, not as a fantasy of switching the network off.

Hard limits that shape the design

Plan the architecture against the ceilings, not after a bulk upload fails.

Google's account-limits documentation also states 128,000 total placement exclusions per ad group. At most 5,000 placements can be added at one time in the UI, and 20,000 at a time through the API. The account-level exclusions help page states that a maximum of 20,000 placement exclusions can be entered at a time, which does not sit cleanly with the 5,000 UI figure. Do not reconcile those pages in a spreadsheet. Check which limit the interface actually enforces before you paste a list.

1. Write down whether the job is a bidding mix, a hard audience block, or inventory hygiene
2. Set the new customer acquisition goal in Value or Only mode, and add the separate existing-customer campaign if you use Only
3. Apply PMax data exclusions only for lists that must not serve on documented PMax surfaces
4. Put never-run brand-safety blocks on the manager-account list and performance pruning on the client-account list
5. Keep Demand Gen content keyword exclusions at account level under Content Suitability
6. Confirm YouTube work is channel- and video-level, because youtube.com cannot be excluded as a domain

What does the public record actually support?

The January 2026 coverage is consistent on the headline. One exclusion list can be applied at account level and applies across Performance Max, Demand Gen, YouTube and Display. Previously, blocks were set at ad group or campaign level. Once excluded at account level, Google Ads prevents spend on those sites, apps or YouTube placements across eligible campaigns. That is the duplication problem the release was meant to solve, and it is why PMax needed it more than campaign types that still allow manual placement selection.

A practitioner write-up from April 2026 disagrees on scoping. It states that the MCC-list process applies to PMax campaigns, and that Display, Video or Demand Gen campaigns will continue using those placements unless a further child-account list is created. It also states that lists created at MCC level cannot be assigned to specific campaigns, whereas child-account lists can be for Display, Video or Demand Gen. That conflicts with the announcement coverage. Google's help pages in this set describe manager lists applied across selected accounts and client lists applied across selected campaigns. They do not settle a Demand Gen exception. The operating response is an in-account verification test, not picking a blog as the winner. The announcement says one list spans PMax, Demand Gen, YouTube and Display. Still check Demand Gen and YouTube residual spend before treating MCC coverage as done.

Several Google Ads Help pages in the public record carried no publication or last-updated date in retrieval. Date-check limits, inventory lists and YouTube wording before you treat them as current. Cite the version you actually built from.

No source in this record publishes a measured, independently verified performance impact of applying PMax data exclusions. Directional ROAS comments in practitioner coverage are expectations, not results. Do not borrow an uplift figure from a thread, and do not brief a client on a conversion mix you have not yet observed in the account.

What should you verify before you change a list?

Before you add a customer list to PMax data exclusions, write down which job you are hiring it for. If the job is "bid for new customers", turn on the lifecycle goal in the right mode and, if you use New Customer Only, stand up the separate existing-customer campaign. If the job is "this list must not appear on those PMax surfaces", apply the data exclusion and accept that it will not cover every PMax inventory.

Then prove the placement layer rather than trusting the setting. Apply the shared hygiene list at the account or manager level that matches the tier. Check account-level change history. New placement exclusions typically take effect within 12 hours. After that, use PMax placement reports segmented by network under "When and where ads showed". If Demand Gen or YouTube still spend on a blocked placement after that window, you have your answer on the unresolved MCC-scoping question, and you now know whether a child-account list is required for those campaign types.

Do not wait for a perfect match rate. Decide the layer, apply the list once, and verify it landed.