PMAX AND SEARCH: AVOIDING CANNIBALIZATION

PMax Search cannibalization is one of the most frequent, and most silent, problems in Google Ads accounts that use both campaign types simultaneously. Performance Max (PMax) is silent in how it cannibalizes because it produces no errors in the interface, generates no alerts, and total account conversions can appear stable or even growing, while beneath the surface PMax is quietly eroding the volume and quality of carefully built Search campaigns.

The problem stems from a priority asymmetry: when a PMax campaign and a Search campaign compete for the same query, Google assigns precedence to PMax in most cases, and this is where the tendency to cannibalize Search campaigns with covered keywords arises.

This article analyses Performance Max Search cannibalization operationally: starting from the official priority rules documented by Google, flagging where those rules have exceptions and grey areas that only emerge from direct use on real accounts, and providing the concrete configuration to prevent the problem.

How priority between PMax and Search works: the official version

According to the official Google documentation, the priority rules between Performance Max and standard Search campaigns are as follows:

• PMax takes priority over generic Search campaigns for the same queries, that is, when the same query could trigger both PMax and a Search campaign running broad match keywords or phrase match.
• Keywords in Search campaigns with exact match retain priority when the user's query is identical to the exact match keyword, but only for that exact query, not for semantic variants.
• Ad quality influences the outcome, if PMax's Ad Rank is higher than the exact match Search campaign's Ad Rank for the same exact query, PMax can still prevail.

These rules appear clear on paper. Operational experience reveals two grey areas that the documentation does not highlight:

First grey area, exact match variants: even with exact match keywords, Google automatically includes semantically equivalent variants (plurals, synonyms, word order). A Search campaign with [men's running shoes] in exact match does not necessarily cover every search query variant such as "running shoes men" or "men running shoes", and PMax can capture these variants. The coverage of exact match is less rigid than it appears.

Second grey area, Ad Rank as override: the documentation mentions that the final Ad Rank determines the auction outcome. In practice, PMax tends to have highly competitive Ad Ranks because it uses automatic bidding with full access to the history of the Google Ads accounts it operates on. A Search campaign with manual bidding or a conservative target CPA can have a lower Ad Rank than PMax even for exact match keywords, and lose the auction.

What experience reveals: the "protected exact match" rule works reliably for exact brand keywords (e.g. [company-name]) but is less reliable for non-brand exact match keywords, especially when the semantic variants of the query are numerous. The only way to guarantee that PMax does not interfere with a specific keyword is to explicitly exclude it from PMax, not to rely solely on exact match in the Search campaign.

Brand cannibalization: the most critical case

Brand protection PMax Search is the area where cannibalization has the most significant, and hardest to justify, consequences. Brand queries, the company name, main product names, the website, have characteristics that make them particularly vulnerable to PMax interference:

• Very high conversion rate: someone searching directly for a brand name almost always has a purchase or contact intention. These conversions would have occurred anyway, with or without ads. When PMax captures them, it claims credit for conversions it did not generate.
• Apparently low CPA that inflates PMax performance: the CPA on brand queries is structurally low because the conversion rate is high. PMax capturing brand queries shows very efficient CPA in its own report, but this is not a result of algorithm optimisation, it is an effect of natural brand traffic.
• Loss of control over the message: brand Search campaigns have titles, descriptions and sitelinks calibrated for the brand. PMax uses an automatic combination of assets that may not reflect brand communication consistently.

The pattern observed most frequently on accounts with PMax and brand Search campaigns running simultaneously without protections: in the first 4–6 weeks after PMax activation, the impression share of brand Search campaigns progressively drops, by as much as 20–40%, while PMax shows growing impressions on the same queries. PMax's CPA falls (because it captures high-converting brand traffic) and the brand Search campaign's CPA can rise (because it receives less volume). Reports show an apparently positive situation, but the reality is a redistribution of credit, not an improvement in performance.

Inference on brand cannibalization: on accounts where brand keywords have been added as exclusions in PMax, the total account CPA in most cases does not worsen, or improves slightly. Brand conversions continue to occur through the dedicated Search campaigns. PMax's CPA increases because it no longer captures easy brand traffic, but the brand Search campaign's CPA stabilises. The net effect on the account is neutral or positive, despite what PMax reports might suggest before the intervention.

Non-brand cannibalization: when it is genuinely a problem

PMax Search cannibalization on non-brand keywords is more nuanced than the brand case. It is not always a problem, in some scenarios it is simply PMax optimising budget allocation more efficiently than the Search campaign. Distinguishing the two cases is essential to avoid intervening where it is not needed.

When non-brand cannibalization is a real problem

• PMax uses non-brand keywords with irrelevant messaging: when PMax is triggered by non-brand queries for which the advertiser has built specific Search ads, and serves them with generic asset combinations, the click rate and traffic quality can be lower than with dedicated Search ads. The problem is not coverage, it is the quality of the ads being served.
• Search campaigns cover strategic segments with specific bids: if a Search campaign has higher bids for certain keywords because that segment is strategically prioritised (e.g. high-margin product keywords), PMax capturing those queries with its own automatic bidding does not respect that strategic priority. Budget can go towards less valuable conversions.
• Search keyword impression share drops without recovery in other channels: if Search campaigns lose impression share on specific keywords and PMax does not generate incremental conversions from channels other than Search, the cannibalization is not adding value, it is only redistributing credit.

When it is not a problem (but looks like one)

If PMax captures non-brand queries that Search campaigns were not covering (long-tail, semantic variants) and generates incremental conversions, it is not cannibalization, it is coverage expansion. This is exactly the desired behaviour when adding PMax to an account already structured with Search campaigns. The test to distinguish the two: check for search terms overlap, verify whether PMax conversions on those queries come from the same queries as Search campaigns or from different ones. The PMax search terms report (available since the 2024 update) is the primary tool for this diagnosis.

How to configure PMax Search keyword exclusions

According to the official Google Ads documentation, since the 2024 update it is possible to add negative keywords directly to Performance Max campaigns from the Google Ads interface, without going through support. PMax Search keyword exclusions are configured via campaign settings: Campaigns → selected PMax campaign → Keywords → Negative keywords.

Negative keywords in PMax apply to the search network and prevent PMax from being triggered for those queries. They do not apply to Display or YouTube placements, a keyword exclusion does not exclude visual channels. This is a point that the documentation specifies but that is often overlooked: excluding "brand-name" from PMax stops PMax from appearing on brand queries in Search, but does not prevent any display or YouTube ads with brand content from reaching those users.

Practical configuration for brand protection

The brand exclusion list to add to PMax must include:

• The brand name in its main variants, exact name, with and without spaces, with common spelling variants. Example: "brandname", "brand name", "www.brandname", "brandname.com".
• The names of main products or services if they are searchable as brand terms, e.g. the name of a proprietary software product, an exclusive product line. If a user searches for that specific name, they are searching for the brand.
• Variants with "reviews", "opinions", "prices", navigational queries indicating users already in an advanced stage of the funnel who will almost certainly convert through the brand Search campaign without needing PMax.

Configuration for strategic non-brand keywords

For non-brand keywords you want to keep under the exclusive control of the Search campaign, the practical rule is to add as negatives in PMax only those for which a dedicated Search ad with specific messages exists that you want to guarantee will be served. Note that search themes, the intent-based signals available in PMax, can partly overlap with non-brand keyword coverage, making selective exclusions even more important. Adding too many non-brand negative keywords in PMax reduces its ability to find incremental traffic, it is a selective measure, not a blanket one.

Inference on exclusion granularity: on accounts where precise brand exclusions (10–20 keywords) have been added without touching non-brand keywords, brand Search campaign performance stabilised within 2–3 weeks , with impression share recovering and CPA stable. On accounts where massive non-brand exclusions were added, PMax showed a significant reduction in total conversion volume without a corresponding improvement in Search campaigns. The principle is: exclude what is necessary, not the maximum possible.

How to diagnose cannibalization in the account

Monitoring PMax Search coexistence requires looking at specific reports in an integrated way, not the reports of individual campaigns in isolation. When analysing Search and PMax campaigns together, the concrete signals of active cannibalization can be identified with three checks:

1. Search campaign impression share over time

The most direct report: add the "Search impression share" column to Search campaigns and check the trend in the 30–60 days before and after PMax activation. A drop in brand Search impression share of 15% or more coinciding with PMax activation is a strong signal of cannibalization. For non-brand campaigns, a drop is expected, what matters is whether total account conversions have grown or have simply been redistributed.

2. PMax search terms report

Since the 2024 update, the PMax search terms report is accessible via Campaigns → PMax campaign → Search terms report. Filter for queries containing the brand name: if they appear, PMax is capturing brand traffic and exclusions need to be configured. Filter for any keyword in your search campaigns: if those same terms appear in PMax results significantly, consider adding selective exclusions.

3. Brand conversion attribution

In the conversions report segmented by campaign, check whether PMax shows conversions attributed to brand queries even before the user has interacted with non-brand ads. If the majority of PMax conversions are brand conversions , clicks on brand ads as the last touchpoint, it is a strong signal that PMax is capturing conversions that would have occurred anyway, not generating incremental demand.

Inference on diagnosis: in almost all accounts analysed with PMax and Search active simultaneously without brand exclusions, at least 25–40% of conversions attributed to PMax came from brand queries. This percentage drops below 5% after brand exclusions are added, and total account conversion volume remains substantially unchanged. This is the most direct confirmation that brand cannibalization redistributes credit, it does not generate added value.