Performance Max in 2026: What Changed and How to Adapt Your Campaigns

Performance Max in 2026 is a different product from the black box that launched in 2021. Google spent the intervening years answering the format's loudest criticism — lack of visibility and control — and the accumulated changes are substantial: channel-level reporting, search term insight with real negative keyword support, asset-group-level data, brand exclusions, and meaningful campaign-level levers. If your PMax setup was designed around the limitations of the early format, it's due for a structural review.

What Actually Changed

Transparency arrived first. You can now see how spend and conversions split across Search, Shopping, YouTube, Display, Discover, Gmail and Maps — ending the era of guessing whether PMax was quietly buying cheap Display conversions or doing real Shopping work. Search term visibility followed, showing the queries PMax matched against, and campaign-level negative keywords graduated from a support-ticket workaround to a standard feature available in every account.

Control deepened alongside. Brand exclusions let you keep PMax off your own brand queries so it can't take credit for demand you already own. Asset group reporting shows conversion data per group, making structure testable. Device and demographic signals sharpened, and customer value modes let acquisition-focused retailers bid differently for new versus returning buyers. None of these change what PMax fundamentally is — an automation wrapper over Google's full inventory — but together they change how you should run it.

Structure: Fewer Campaigns, Deliberate Asset Groups

The early-era pattern of splitting PMax into many small campaigns to force budget allocation mostly fights the machine. In 2026 the recommended shape for e-commerce is: consolidate by economic logic, segment by asset group. One PMax campaign per meaningful margin or strategic tier — not per product category — with asset groups inside it organized around genuinely different customer intents or product narratives.

Margin-based segmentation remains the highest-value structure decision. Group products by contribution margin using custom labels, set ROAS targets per campaign that reflect actual profitability rather than a blanket account target, and let high-margin lines run at aggressive targets while low-margin lines earn stricter efficiency. Blended targets subsidize your worst products with your best ones — the classic silent killer of Shopping-heavy accounts.

Use brand exclusions in every retail account. Measure PMax on incremental, non-brand performance. If removing brand queries makes the campaign look dramatically worse, you've learned what its reported numbers were hiding — and you now have a real baseline to optimize against.

Assets in the AI-Creative Era

Asset quality now visibly affects delivery. Feed every group a full complement: the maximum text variants, genuinely different images (not crops of one photo), and video — because if you don't supply video, Google auto-generates it, and auto-generated product slideshows represent your brand exactly as well as you'd expect. Short vertical product videos, even simply produced, outperform the templates and unlock YouTube inventory on your terms.

Google's AI asset generation matured to the point of being useful for volume — backgrounds, resizes, text variants — but concept and brand voice still have to come from you. The practical workflow: human-defined concepts and hooks, AI-assisted execution and variation, then let asset reporting kill the losers. Review asset performance monthly and replace "Low" performers rather than letting them rot in the group.

Audience signals remain signals, not targeting — but they matter at launch. Seed new asset groups with your customer lists and high-intent site audiences to shorten the learning phase. The algorithm outgrows the seed within weeks; the seed determines how expensive those weeks are.

Search Terms and Negatives: Use the Visibility

With search term insight and negative keywords available, PMax accounts should adopt the same query hygiene as Search campaigns. Review matched terms bi-weekly for the first months, then monthly at steady state. You're hunting three categories: irrelevant matches burning budget, near-miss queries that deserve dedicated Search coverage, and competitor or informational terms whose value depends on your strategy and margins.

The interaction with standard Search campaigns is now manageable instead of mysterious. Keep exact-match Search campaigns for your proven money keywords — Search takes priority when the query is an exact match to an eligible keyword — and let PMax mop up the long tail and cross-channel demand around them. Check the channel split monthly: if PMax leans heavily on Search-channel conversions for queries your Search campaigns already cover, tighten with negatives; if it's finding genuinely new query territory, leave it room.

Migration Checklist for Older Setups

If your PMax structure predates these controls, run through this list. Add brand exclusions and re-baseline performance on non-brand numbers. Consolidate fragmented campaigns into margin-tiered ones with asset groups carrying the segmentation. Replace auto-generated and low-rated assets; add real video to every group. Install a recurring search-term review with campaign-level negatives. Turn on new-customer value mode if acquisition is your bottleneck. Verify channel-level reporting matches your assumptions about where the money goes — and if Display or auto-video placements are consuming budget without conversion support, cut them via asset and exclusion controls.

The through-line of every 2026 change: Google handed back enough visibility that PMax performance is now an operator skill, not a coin flip. The format still rewards clean conversion data, honest margin math and good creative above all clever configuration — but for the first time since launch, you can see enough to manage it like a real campaign. Most accounts we audit haven't caught up to that. The ones that have are quietly compounding the advantage.