Google Ads Performance Max in 2026: What Works, What Doesn't, and How to Win

Performance Max became the dominant Google Ads campaign type in 2024, and in 2026 it accounts for approximately 60% of total Google Ads spend globally. The premise is compelling: one campaign type that accesses all Google inventory — Search, Shopping, Display, YouTube, Discover, Gmail, Maps — and lets Google's machine learning allocate budget to the best-performing channels in real time. The reality is more complicated. Performance Max is extraordinarily powerful when structured correctly and extraordinarily wasteful when structured incorrectly. The gap between a well-structured and a poorly-structured Performance Max campaign, running the same budget against the same products, routinely exceeds 3x in ROAS in our client audits.

Asset group architecture: the structure that determines performance

Asset groups are the primary organisational unit within a Performance Max campaign. Each asset group contains a set of creative assets (headlines, descriptions, images, videos, logos) and a set of audience signals. The optimal asset group architecture segments by product or service category first, then by audience intent second. Do not create one asset group for your entire product catalogue — create one per major product line or service. This allows asset performance to be evaluated meaningfully (you can see which headlines work for Product A vs. Product B), audience signals to be relevant to the specific products being advertised, and budget to be allocated at the product level rather than at the campaign level.

Asset quality: the lever most advertisers underestimate

Performance Max uses your asset quality scores to determine which combinations of assets to serve to which audiences. Low-quality assets — generic stock photography, short generic headlines, descriptions that repeat the headline verbatim — limit Google's ability to optimise because all asset combinations perform similarly (poorly). High-quality assets — original photography showing the product in use, specific benefit-driven headlines that differ meaningfully from each other, video assets that hook in the first 5 seconds — give Google the creative diversity it needs to find winning combinations for different audience segments. The minimum creative brief for a well-funded Performance Max campaign: 15 unique headlines with distinct angles (not 15 variations of the same message), at least 5 original images per asset group (no stock photography), and one 15-second video per asset group.

Audience signals: what they are and what they are not

Brand exclusions: the protection layer no one mentions

Performance Max, without brand exclusions, will allocate significant budget to branded search queries — people searching your brand name. This is budget that would have converted organically at zero cost if the Performance Max campaign had not captured it. The effect is a ROAS inflation: branded search conversions attribute to Performance Max, making the campaign's ROAS appear strong, while in reality the campaign is cannibalising free organic traffic and not generating incremental conversions. Add your brand name, all its misspellings, and all product names as negative keyword exclusions in Performance Max from day one. Run a separate branded search campaign at a much lower CPC cap to protect branded intent. This structural separation is the most common structural fix we make in Performance Max audits, and it typically reduces spend by 15 to 25% while maintaining total conversions.

• Create one asset group per major product category — never one asset group for your entire catalogue.
• Provide your CRM customer list and your converters from the last 90 days as audience signals.
• Add brand exclusions on day one — your brand name, all misspellings, and all product names.
• Do not evaluate performance or make budget changes in the first 4 weeks — the learning phase requires stability.

The optimisation cadence: when and what to change

Performance Max has a learning phase of 6 to 8 weeks after any significant change. Making frequent changes during this phase resets the learning and extends the period of suboptimal delivery. The optimisation cadence we recommend: weeks 1 to 4, observe only — review asset quality scores, check for brand query cannibalisation, verify conversion tracking. Weeks 5 to 8, make one structural change if clearly needed — add or remove an asset group, adjust budget allocation. Month 3 onwards, monthly review — refresh underperforming assets (any asset rated Poor by Google), review audience signal performance, and compare Search Terms Insight data to identify new negative keyword candidates. Never make more than one significant change per optimisation cycle.