On 21 September 2026 the Google Ads API changes what it returns for the Network field on Performance Max campaigns. Instead of a single "Cross-network" value, the same campaign now reports the specific channel a row ran on — Search, Search Partners, YouTube, Display, Gmail, Discover, or Maps. Supermetrics, documenting the change for the tools it feeds, describes it plainly: Performance Max campaigns used to return a single "Cross-network" value for Network, but now report the specific channel. One PMax row can become as many as seven.
For anyone auditing where Performance Max actually sends the money, this is good news buried inside a landmine. The good news is real per-channel visibility, exposed at last through the API. The landmine is that any report, connector, or script that filters or segments PMax on Network = "Cross-network" will stop returning PMax data the moment the change lands — with no error, just an empty result. This post covers what changes, why the silent break is more dangerous than an outright failure, and a concrete checklist to run before the deadline so your spend audit does not quietly zero out.
What changes on 21 September
Three fields change, but only one matters for search-term hygiene. The Network field on Performance Max stops collapsing everything into "Cross-network" and starts naming the channel each row ran on. The other two changes are peripheral to this work: asset performance labels (the Good/Best ratings) are removed for Search and Display campaigns, and the click type segment on asset reports changes. Collective Measures, covering the same update, notes the underlying API mechanic — in v23 the ad_network_type field returns real channel values rather than a mixed placeholder — which is why every downstream tool inherits the new shape whether or not you asked for it.
The practical effect is a row-count explosion. A campaign that returned one Cross-network row per day now returns up to seven, one per channel that served an impression. That changes totals, data-transfer volumes, and any calculation built on the assumption of one row per campaign per day. Advertisers on the Google Ads API forum have been asking how to read Performance Max metrics split by network as the behaviour rolled out, which is the tell that this is not a documentation footnote but a shape change teams are hitting in production. If your reporting sums or counts rows to derive a campaign metric, that arithmetic now spans seven rows where it used to span one.
The silent break: why an empty result is worse than an error
The dangerous part is not the new values — it is the old one disappearing. After 21 September, for recent PMax data, "Cross-network" is no longer a value the Network field ever returns. A filter that says Network equals Cross-network therefore matches nothing, and the API returns an empty result set rather than raising an error. Nothing crashes. No alert fires. The report renders cleanly and shows zero Performance Max spend, and whoever reads it concludes the campaign stopped spending — when in reality the money is flowing exactly as before under seven new labels the filter was never written to accept.
A hard failure would be safer, because a hard failure gets noticed and fixed. A silent zero gets trusted. This is the same failure mode that makes a shrinking search terms report so corrosive: the number looks authoritative precisely when it has quietly stopped describing reality. If you run a wasted-spend scorecard that pulls PMax cost through a Cross-network filter, that scorecard reads zero waste on the channel where you most need eyes, and the campaign sails past every threshold because the data feeding the threshold is empty. The break is invisible until someone cross-checks against billing, and by then a reporting cycle or two has passed on bad numbers.
The 1 June 2025 boundary and the API-only caveat
Two constraints shape how the new data behaves. First, channel-level Performance Max data exists only for dates from 1 June 2025 forward. A query that spans that boundary returns "Cross-network" for the earlier rows and specific channel names for the later ones, so a single table can carry both schemes at once. Year-over-year and long-window comparisons are the obvious casualties: a chart that groups by Network will show a fat Cross-network bucket for the old period sitting next to seven thin channel buckets for the new one, and any naive sum double-counts or mislabels unless you normalise the two eras deliberately.
Second, this breakdown is exposed only through the Google Ads API — not in the standard Google Ads interface. If your audit lives entirely in the UI you will not see the channel split at all, and you will keep reading the campaign as a single opaque block the way Performance Max always presented itself. The visibility arrives only when you pull through the API or a tool built on it, which is exactly where the broken filters live. So the change is simultaneously invisible where you work by hand and disruptive where your automation works for you — the worst distribution for catching it early. Treat the API as the only source of truth for channel-level spend, and treat the UI as unchanged.
The pre-deadline checklist: find every Cross-network dependency
The work before 21 September is an inventory: locate every place that reads the PMax Network field and remove any dependency on the literal string "Cross-network". Start with the connectors and dashboards, because they fail most silently. Audit each Looker Studio data source and its filters, every Supermetrics or equivalent connector query, and any custom Google Ads API script or scheduled export. For each, find where Network is filtered, segmented, or grouped, and rewrite it to accept the seven channel values rather than matching one obsolete string. Where you need a campaign-level total, change the logic to sum across channels instead of selecting a single Cross-network row.
Then extend the sweep to the softer dependencies: saved reports in the interface that segment by network, automated rules that key off it, and any spreadsheet a human maintains by pasting an export into a tab with a Cross-network filter downstream. The failure is the same everywhere — a filter that once matched now matches nothing. When the inventory is done, validate rather than assume: run a test pull covering a date range after 1 June 2025 and confirm the seven channel values flow through your pipeline correctly, so you are checking against live data before the deadline rather than debugging a zeroed-out report the following Monday. This is the same discipline as a standing visibility-ratio audit: verify that the report still describes reality before you trust a number it produces.
Reading the new per-channel rows for spend hygiene
Once the plumbing survives the change, the payoff is a question you could never answer cleanly before: how much of a Performance Max campaign's budget lands on inventory your negative keywords can actually touch. Performance Max negative keywords apply only to Search and Shopping inventory — they do nothing to YouTube, Display, Discover, Gmail, or Maps placements. When everything reported as a single Cross-network figure, that scope limit was invisible; you added negatives and hoped. With per-channel rows you can measure it: sum the cost on Search and Search Partners against the cost on the display and video surfaces, and you know immediately what fraction of the campaign your query hygiene can move.
That number reframes the whole exercise. If a campaign runs most of its spend on Search and Shopping, mining its search terms report and adding channel-level negatives is high-leverage work. If most of the cost sits on Display or YouTube, no amount of negative-keyword cleaning will move it — that is a steering problem for asset groups, audience signals, and placement controls, and pretending negatives will fix it just burns hours. The channel split lets you triage: spend your search-term hygiene where the inventory is query-driven, and stop expecting PMax negatives to police surfaces they were never built to reach.
What this does not change
This is a reporting-shape change, not a change to how Performance Max spends or how negatives behave. The search terms report itself is unaffected; the queries you can see, and the ones Google redacts, are governed by separate rules. Your negative keyword lists keep working exactly as before, with the same Search-and-Shopping scope limit they always had. Bidding, asset groups, and audience signals are untouched. What changes is purely the label the API hangs on each row and, consequently, whether your tooling can still find the data it was built to read. The campaign is the same; your view of it is what needs updating.
So keep the response proportionate. You do not need to restructure campaigns or rewrite your negative strategy for this — you need to fix filters and validate pulls before a dated deadline, then take the upside of clearer channel visibility once the pipeline is safe. If you already run a disciplined search hygiene audit, fold the Cross-network sweep into it as a one-time task with a hard due date, and let the new per-channel data make the next audit sharper than the last. The break is avoidable with an afternoon of inventory; the visibility is worth keeping.
Related reading
For mining the report the channel data sits alongside, see the Performance Max search terms report and channel-level negatives. For the scope and limits of PMax negatives themselves, see Performance Max negative keywords, and for keeping the whole account's reporting honest, see the visibility-ratio audit and the wasted-spend metrics scorecard.