Campaign Cannibalization in Google Ads 2026: PMax, Search and Shopping Fighting for One Demand Pool

You launch Performance Max. A month later PMax reports 180 conversions and your Search campaigns report 240 — down from 400. Total: 420 versus 400, a 5% lift, on 60% more spend. On paper everything grew. In reality you just paid extra for the same sales. That is campaign cannibalization in Google Ads: two campaigns in one account chasing the same demand while you pay twice for one outcome.

This guide covers how Google decides which of your campaigns serves, the six signals that expose cannibalization in your reports, a 40-minute diagnostic, and the eight separations that actually work — plus four fixes that don’t.

Cannibalization vs. healthy overlap

Campaigns inside one account do not bid against each other. Google picks one before the auction. The damage is subtler: duplicate campaigns manufacture the appearance of growth, because reported conversions rise when you simply redistribute the same demand across more rows.

Healthy overlap looks like this: PMax finds people Search never reached — new queries, YouTube, Discover, Gmail — while Search captures explicit high-intent demand. Cannibalization is when PMax starts serving on your own brand terms and head commercial queries that Search was already winning.

There is exactly one test that matters: did incremental conversions change, not whether reported conversions went up. If they didn’t, you moved money from one pocket to another and lit the difference on fire.

How Google decides which campaign serves

Situation	What wins
Search campaign with an exact-match keyword identical to the query	Search — exact match takes priority regardless of PMax Ad Rank
Search campaign on phrase or broad match vs PMax	Higher Ad Rank wins, and that is frequently PMax thanks to its broader signal set
Standard Shopping vs PMax on the same products	PMax takes priority over standard product campaigns
Two Search campaigns with the same keyword	The higher Ad Rank campaign serves; the other simply doesn’t
Demand Gen vs PMax on shared placements	Normal auction dynamics apply, and conversion de-duplication blurs the reporting picture

The practical consequence: “exact match beats PMax” only protects queries you have literally built as exact match. Everything you catch on phrase or broad is open territory, and PMax will take it whenever its Ad Rank is higher. That is exactly why brand lists and brand exclusions moved from “nice to have” into basic account hygiene.

Six signals of campaign cannibalization in Google Ads reports

Symptom	Where to look
Search impressions and clicks drop in sync with the PMax launch while Search CPA rises	Day segment plus change history
Search impression share falls but lost IS (budget) and lost IS (rank) do not rise	Impression share columns
Your brand terms show up in the PMax search terms report	PMax search terms
Total conversions grew slower than total spend	Account-level period comparison
PMax CPA looks suspiciously low — brand-search low	Campaign CPA comparison
Auction Insights lists your own domain among overlapping advertisers	Auction Insights report

That last one is the most underused check. How to read the report properly — including whether you are losing position to budget, to rank, or to your own second campaign — is covered in our guide to impression share and Auction Insights.

The 40-minute diagnostic

1. Pin the launch date from change history. Every comparison anchors to that date, not to “last month.”
2. Compare equal 28-day windows before and after, so day-of-week seasonality doesn’t distort the read. Look at account level: spend, conversions, conversion value, CPA.
3. Pull the PMax search terms report. Sort by conversions and flag every term that duplicates your keywords or brand. If more than 20–30% of PMax conversions come from brand queries, cannibalization is confirmed.
4. Check Search impression share trend. A drop in impression share with flat lost-IS-to-budget and lost-IS-to-rank means the queries stopped reaching Search at all.
5. Split brand from non-brand. Build the comparison with a brand-term filter applied separately. Cannibalization almost always starts at brand, because that is where CPA is lowest and conversion rate highest — so automation goes there first.
6. Check product campaign overlap. Running standard Shopping alongside PMax on the same SKUs usually means Shopping is starved. Segment the catalog with custom labels in Merchant Center.
7. Reconcile with sales. Open the CRM for the same windows. Reported conversions up 20% and closed deals flat settles the argument.

Ad account reports show redistribution, not lift. The only way to size cannibalization honestly is to switch the suspect campaign off in a set of regions and compare against a control group.

Measuring the real damage

Reports give you a hypothesis. Only an experiment settles it, and you have two levels of rigor:

• Fast: a campaign experiment. Fine for comparing two Search campaigns against each other — mechanics in our walkthrough of Google Ads experiments.
• Rigorous: a geo holdout. Turn PMax off in one region group, keep it live in the control group, and compare total sales — not account-reported conversions — over three to four weeks. Methodology in our guide to incrementality and geo experiments.

A simple estimator: incremental value = (test-group sales − expected sales from control) / campaign spend. If incremental value hovers near zero while spend is significant, that campaign is claiming credit rather than creating demand.

Eight separations that work

1. Brand exclusions on PMax. The single highest-impact fix: apply a brand list covering your brand and its misspellings to PMax so brand demand routes to a dedicated Search campaign with predictable economics.
2. Account-level negative keyword lists. Broader reach than campaign-level lists and ideal for systematically cutting off irrelevant themes. Building the lists is covered in our piece on negative keywords and search terms.
3. Exact match on your core. Anything you want guaranteed to Search must exist as an exact-match keyword — that is the only priority rule that beats PMax by design.
4. Feed split via custom labels. In e-commerce, hand some product groups to PMax and others to standard Shopping so there is no overlap to fight over.
5. Customer-type goals. Use lifecycle goals and customer-type labels — new versus returning — so PMax optimizes for acquisition while retargeting lives separately.
6. Geo and language separation. If two campaigns target identical regions “just in case,” removing the overlap is pure savings.
7. Time separation. Sometimes staggering ad schedules is enough, particularly when the contested demand is a narrow slice.
8. Structural review. More near-identical campaigns means more risk, and fragmentation also slows bid-strategy learning — see our breakdown of account structure at scale.

Putting a number on it

“The campaigns overlap” is a hard sell to a finance director. Here is the same situation priced out.

Metric	Before PMax	After
Search spend	$40,000	$26,000
Search conversions	400	240
PMax spend	—	$38,000
PMax conversions	—	180
Total spend	$40,000	$64,000
Total conversions	400	420
Account CPA	$100	$152

In the campaign report PMax looks like a win: 180 conversions at $211. At account level the story inverts — 5% more conversions for 60% more spend. Price the increment: ($64,000 − $40,000) / (420 − 400) = $1,200 per additional conversion against a $100 baseline CPA. Even if some PMax conversions are genuinely new, the marginal cost is twelve times your baseline — and that single calculation is the one worth putting in front of stakeholders.

The inverse case exists too: spend up 60%, conversions up 55%. Cannibalization is present but acceptable — the channel adds demand at roughly the same price, and the decision moves to unit economics and your maximum affordable CPA.

The overlooked case: two Search campaigns with the same keywords

Search-on-Search cannibalization is more common than PMax-on-Search and gets noticed far less, because no new campaign type appears to blame. Usual origins:

• Campaigns split by region, but targeting was never restricted, so both run nationwide.
• A keyword set duplicated into a “test” campaign with a different bid strategy — and both left enabled.
• Broad match in one campaign intercepting exact keywords from another.
• Dynamic campaigns picking up queries already covered by standard ad groups.

Diagnosis is easier than with PMax: build a search terms report segmented by campaign and find terms serving in two campaigns at once. Then either add cross negatives or merge the campaigns. Merging often wins: one campaign with enough data learns better than two starved ones.

Cannibalization in retargeting and Demand Gen

The third layer usually goes unnoticed because it lives outside Search:

• Retargeting vs PMax. PMax already serves ads to your past visitors. Run a separate remarketing campaign alongside it and you pay twice to bring back the same user, with credit going to whoever touched them last. Separate the audiences with list exclusions.
• Demand Gen vs PMax. Both run on YouTube and Discover. This is exactly what campaign-type attribution — added in 2026 — is meant to untangle, by isolating Demand Gen’s contribution. Format mechanics are covered in our Demand Gen guide.
• Brand campaign vs dynamic formats. Dynamic formats pull in brand queries automatically, starving the brand campaign while you pay for the same traffic somewhere else in the account.

Four fixes that don’t work

• “I’ll lower PMax bids and Search will recover.” PMax has no manual bids. Loosening target ROAS changes aggressiveness but not query overlap.
• “I’ll add negative keywords to PMax.” Campaign-level negatives in PMax are limited in scope and do not replace brand exclusions. Verify with the search terms report, not with expectations.
• “I’ll split them into separate accounts.” That makes it worse: campaigns in different accounts owned by the same advertiser genuinely compete in the auction, and you start bidding your own CPC up.
• “PMax is cheaper, so I’ll pause Search.” It is cheaper precisely because it absorbed your brand queries. Pause Search and you lose control of your most profitable traffic.

A scaling sequence that avoids the trap

1. Before launching anything new, write down which demand it captures that existing campaigns cannot.
2. Apply brand exclusions and account negative lists on day one — not after the first invoice.
3. Launch in a subset of regions first; that hands you a control group for free.
4. After three to four weeks, compare sales between test and control regions instead of reading account reports.
5. Keep the campaign only if total results grew faster than total spend.
6. Re-audit the PMax search terms report quarterly — boundaries drift on their own, especially as reach expands. See our notes on optimized targeting and audience signals.

Separating each campaign type’s true contribution from de-duplicated reporting is a measurement problem in its own right. In 2026 Google added campaign-type attribution in-account, and at the strategic level the question belongs to media mix modeling — covered in our guide to MMM and Meridian for paid media measurement.

The hidden cost: cannibalization breaks bid strategy learning

Beyond the direct overspend there is a second effect that shows up later. Automated bidding learns from the data inside a single campaign. When the same demand is split across two campaigns, each gets half the signal — and both learn worse.

• Longer learning periods. A campaign with 30 conversions a month stabilizes far more slowly than one with 60. Split the flow in half and you double time-to-stability for both.
• Higher volatility. At low volume the strategy overreacts to random spikes, CPA swings, and the buyer starts “fixing” what is simply noise.
• Weaker optimization. With thin data the bidding model cannot separate expensive segments from cheap ones fast enough to act on the difference.
• A cascade of manual edits. Every intervention restarts learning, and fragmented campaigns invite more interventions — a closed loop.

The practical rule: if two merge candidates together produce fewer than 50–60 conversions a month, merging almost always beats separating, even if the structure looks less tidy on a slide. Treat those numbers as guidelines — the longer your conversion cycle, the more volume you need.

A five-point monthly check

1. Share of brand queries in PMax conversions — is it creeping up month over month?
2. Impression share for the brand campaign and the main Search campaign.
3. Auction Insights, checking for your own domain among overlapping advertisers.
4. Spend growth versus conversion growth at account level, not campaign level.
5. The list of enabled campaigns — any forgotten test duplicates still running?

FAQ: campaign cannibalization

Do campaigns in one account bid against each other?

No. Google selects one campaign before the auction using priority rules and Ad Rank, so you are not inflating your own CPC. Budget still gets redistributed between them, which is what creates the illusion of growth.

Does exact match really always beat PMax?

Yes, but only when the query matches your exact-match keyword identically. Phrase and broad match provide no such protection.

How do I quickly tell if PMax is eating brand traffic?

Open the PMax search terms report and calculate the share of conversions coming from brand queries. Roughly 20–30% is a reasonable alarm threshold, though the exact line depends on how strong your brand is.

Should I still run a brand campaign if PMax covers it?

Usually yes. A dedicated brand campaign gives you control over ad copy, cost and competitor defense on your own name. Handing brand to automation means losing control of your cheapest traffic.

What do I do with standard Shopping after launching PMax?

Either split the catalog by custom labels or run a single campaign type per SKU. Running both on the same products almost always means Shopping is idle.

Is cannibalization only a PMax vs Search problem?

No. Classic cases include two Search campaigns with overlapping keywords, retargeting versus PMax, Demand Gen versus PMax on YouTube and Discover, and brand campaigns versus dynamic campaigns.

How do I measure impact when conversions get de-duplicated?

Reports become nearly useless there — read account-level totals and CRM revenue instead. For a rigorous number you need a geo holdout.

Does capping the greedy campaign’s budget help?

As a stopgap, yes — some volume returns to the other campaigns. But it treats the symptom: the query overlap remains and everything reverts once the cap comes off.

How long should I wait before judging a change?

Two weeks minimum, three to four after changing goals or brand exclusions, because bid strategies re-enter a learning period. Judging week one is meaningless.

Can cannibalization ever be acceptable?

Yes, when total incremental results grow. PMax sometimes genuinely brings new customers while incidentally absorbing brand traffic — in that case the total is what matters, not clean separation.

Conversions dropped after restructuring — what do I check first?

Change history, bid strategy learning periods, and whether ads stopped serving for technical reasons. The diagnostic order is in our guide to ads that aren’t showing. Separately, check whether ads were disapproved under destination requirements — that looks like demand shifting between campaigns when the cause is purely technical.

If your constraint is not structure but the account’s ability to spend at all, look at our Google Ads agency accounts and Google Ads playbook — stable delivery and clean structure solve different halves of the same problem.