For four years Performance Max answered every "where did my money go?" question with a shrug. In the Google Ads API, the field that should have told you the channel returned a single value: MIXED. That's over.
Google Ads now reports Performance Max results by channel — Search, Search Partners, YouTube, Display, Discover, Gmail, and Maps — at campaign, asset group, and asset level. It showed up in the interface as a beta in mid-2025, and the Google Ads API opened it up properly in v23 in January 2026. So you can finally see the split.
Here's the part nobody puts in the headline: the report is last-touch inside Google's own model. It tells you where a conversion got credited, not where it got created. We've now watched several accounts get worse after their team read this report for the first time and started cutting. This is how to read it without doing that.
What the channel report actually gives you
Three things, and they're not equally useful.
Spend and conversions by channel, at campaign level. This is the one in the UI. It answers the basic question — how much of your Performance Max budget went to YouTube versus Search versus Display. Most advertisers find the Search share smaller than they assumed and the Display share larger.
The same split at asset group and asset level, via the API only. This is the good one, and it's genuinely not in the interface. Campaign-level data tells you Display took 28% of the budget. Asset-group-level data tells you which asset group's imagery took it. Those are different problems with different fixes, and you can only see the second one through the API.
Companion segments for creative type. You can layer ad_using_video and ad_using_product_data on top of the channel split, which is how you find out that your one decent video asset is carrying YouTube entirely while three others do nothing.
One constraint that trips people up: channel data only exists from June 1, 2025 onward. Google didn't backfill it. Any year-over-year channel comparison you try to build before that date returns nothing, so run your first analysis on a rolling window instead of a YoY view and save yourself the confusion.
The trap: this report is last-touch, and it's grading its own homework
Every conversion in the channel report is attributed to whichever surface the user touched last before converting. That's it. There's no assist column.
Think about what that does to a normal customer journey. Someone sees a Discover placement, doesn't click. Sees a YouTube pre-roll two days later, doesn't click. Searches your brand name on day four and converts. The channel report hands 100% of that conversion to Search and zero to Discover and YouTube — and then shows you a Discover CPA that looks indefensible.
Cut Discover and Display on that basis, and the pattern is predictable: your Search CPA stays fine, your Search volume drops, and your total conversions fall. You didn't remove waste. You removed the top of your own funnel and kept the part that harvests it.
This is the same measurement problem we write about constantly in a different costume — it's why GA4 and Google Ads report different conversion numbers for the same campaign, and why we push clients toward an independent attribution layer instead of grading a channel with the platform that sells it. A platform's own last-touch report is evidence. It isn't a verdict.
To be fair to Google: they're not hiding this. The channel prioritization control they've been testing carries the same caveat in its own documentation. But the report doesn't warn you at the moment you're looking at a red CPA and reaching for the exclusion button.
The four-step pass we run before touching anything
We treat the channel report as the first input, not the decision.
1. Get the shares, ignore the CPA
Pull spend share and conversion share by channel over a clean 60- to 90-day window. Don't look at CPA yet — CPA is where the last-touch distortion lives, and looking at it early anchors you. You want the shape: which channels are taking budget, which are getting credit.
2. Write down the gaps as questions
A channel taking 30% of spend and returning 5% of credited conversions is a question. So is a channel taking 4% of spend and returning 20%. Both are worth explaining. Neither is an instruction.
3. Check it against something Google doesn't own
Pull the same window from your server-side tracking or attribution platform. The tell for an assisting channel is simple: its spend correlates with lift in direct and branded search, even though its own credited CPA looks terrible. If you see that correlation, the channel is doing upper-funnel work and the report is mislabeling it. If you don't — if the spend goes up and nothing else moves anywhere — then you've found real waste.
4. Change one thing, then wait
Adjust one lever. Let it run a full conversion lag window, which for most of our B2B and high-ticket clients is two to four weeks, not four days. Then read total conversions and blended CAC, not the channel's own CPA. Changing three things at once and reading it on Friday is how agencies convince themselves of things that aren't true.
What you can actually control in PMax right now
Visibility improved faster than control did. Here's the honest inventory as of this month.
Campaign-level negative keywords, up to 10,000. Real, useful, and mostly affects the Search side. This is the cleanest lever you have for cutting genuinely irrelevant queries.
Brand exclusions, native and shareable at account level. Better than negatives for the specific job of excluding a brand, because they match variants and common misspellings automatically instead of making you guess every typo. Use these for brand work, not negatives.
Placement exclusions. Available, and worth a monthly pass on the placement report — Display and YouTube inventory still surfaces the occasional placement you'd rather not be on.
Channel prioritization — in alpha, and not what people think it is. Google's been testing an adjustment control across Search, YouTube, Display, Discover, Gmail, and Maps. It is not a budget split and not a bid multiplier. A positive adjustment relaxes the CPA the system is willing to accept on that channel; a negative one tightens it. You're nudging the economics the algorithm optimizes against, not allocating dollars. Very few accounts have it yet, so don't build a strategy on it.
What you still can't do: hard-cap a channel's budget, or turn one off outright. If a channel is genuinely broken for you, your options are still the indirect ones — starve it through asset choices, exclude the placements, or move that objective into a dedicated campaign. Which is part of why the Display-to-Demand-Gen migration matters more than it looks: a standalone campaign is still the only way to get real control over a surface.
When the data does justify a change
Not everything in the channel report is a mirage. Three patterns we act on with reasonable confidence:
A single asset group absorbing a channel's budget with no conversion share anywhere. Pull the asset-group-level split from the API. If one group is eating Display spend and neither it nor the account shows movement, that's usually a weak image set the algorithm found cheap inventory for. Fix the creative or retire the group.
Search Partners running high spend with a flat total. Search Partners is the one channel where we're quickest to be skeptical, because there's less assist behavior to give it the benefit of the doubt.
Query-level irrelevance. If the Search share is bringing in queries that have nothing to do with your business, that's not attribution nuance, it's a targeting problem. Negatives and brand exclusions, and it's a legitimately fast win. The same discipline applies to the newer AI-driven Search formats we covered in our AI Max breakdown.
The through-line: act on absence of total effect, not on a bad channel-level CPA. One of those is a measurement of reality. The other is a measurement of Google's attribution model.
What to do this week
If you've never pulled this report, start here.
Segment your largest Performance Max campaign by channel over the last 60 days and just look at it — no changes. Then pull the same window from whatever independent tracking you have and put the two side by side. For most accounts, that single comparison is more useful than the next three optimizations you were planning, because it tells you how much of your reporting you should actually believe.
If your asset-group-level data isn't accessible because you're only using the interface, that's the gap worth closing first. Campaign-level channel data tells you that you have a problem. Asset-group data tells you where it lives.
And if the answer to "what does our independent attribution say?" is that you don't have any — that's the real finding, and it's a bigger lever than any channel adjustment in this post.
Get a free automation audit and we'll pull your Performance Max channel split against your actual tracking, tell you which channels are assisting versus wasting, and show you what we'd change first. If you'd rather see how we run paid media day to day, that's on our ads management page.