What Performance Max actually automates
Performance Max (PMax) is a goal-based campaign type that gives Google's AI access to all of its ad inventory from a single campaign. Per Google's own support documentation, PMax is designed to complement keyword-based Search campaigns, not replace them. It is meant to find converting customers across YouTube, Display, Search, Discover, Gmail, and Maps simultaneously, using one shared budget and one conversion objective. Google originally announced the global rollout of Performance Max campaigns in 2021 via the Google Ads blog, and the product has expanded significantly since then with the addition of asset performance reporting, search theme signals, and AI-generated video and text customisation.
The automation covers four domains inside that structure:
- Smart Bidding. PMax uses real-time auction signals, historical conversion data, and your target cost per action (CPA) or return on ad spend (ROAS) to set bids at each auction. No manual bid adjustment is needed or effective once a target is set.
- Placement decisions. The AI decides where your ads appear across Google's inventory. You cannot specify individual placements within a PMax campaign. You can set account-level brand safety exclusions.
- Asset testing and combination. PMax tests combinations of the text, image, and video assets you provide, serving whichever combinations drive the most conversions. Asset performance reports indicate which combinations are "best", "good", or "low" after sufficient data accumulates.
- Audience expansion. Starting from the audience signals you provide (customer lists, remarketing audiences, interest categories), PMax expands to lookalike and cold audiences it predicts will convert. You cannot restrict this expansion without reducing reach significantly.
How PMax and Search coexist
When a user query exactly matches an active exact-match keyword in your Search campaign, the Search campaign is prioritised. Search themes in PMax have the same prioritisation as phrase and broad match. PMax fills the gaps your Search campaign does not cover, which is why the two run best in parallel rather than as a direct switch.
The implication is direct: PMax is not a campaign type where a partner earns their keep by adjusting bids or selecting placements. Those decisions belong to Google. The work that remains is everything the AI needs from you as inputs, and everything the AI's outputs require a human to interpret.
What the AI cannot do: the four hard gaps
The automation ceiling in PMax is not where most buyers think it is. These are the four areas where the algorithm is structurally limited and where human input determines whether the campaign performs or stalls.
1. Conversion architecture
PMax optimises toward whatever conversion event you define. If that event is a form submission, the AI will drive form submissions. Whether those submissions represent genuine buyer intent, your CRM never sees, or your sales team cannot qualify, is invisible to the system. The AI works with what you measure, not with what you want. Building a conversion taxonomy that maps platform events to real business outcomes, then connecting CRM data back to Google via offline conversion imports or enhanced conversions, is the most consequential work in any PMax engagement. It is also the work that cannot be automated.
2. First-party signal quality
Google's documentation is explicit that audience signals, customer lists, and data feeds "significantly improve" campaign performance. The AI cannot build these inputs. A customer list from a CRM with 200 records, stale emails, and no segmentation is a weak signal. A refreshed list of recent high-value customers, segmented by product line and lifecycle stage, is a strong one. The difference in campaign behaviour is material. A partner who does not audit your audience signal quality in week one is not doing the job.
3. Creative quality and asset group structure
Asset testing only works when the assets being tested are genuinely differentiated. Ten headlines that say the same thing in different words are not ten signals; they are one signal with noise. A PMax partner's creative role is to bring enough variation in message angle, audience framing, and format that the AI has something meaningful to learn from. Asset group structure also matters: grouping assets by product category, audience type, or funnel stage gives the AI cleaner signals than a single flat asset group covering everything. This is editorial and structural work, not automation.
4. Incrementality and attribution judgment
PMax uses Google's data-driven attribution model, which credits touchpoints across Google's own inventory. It does not tell you whether those conversions would have happened without the campaign. This is an inherent limitation of any single-platform attribution model, not a deficiency unique to PMax. A capable partner designs incrementality tests, compares PMax conversion reports to CRM-confirmed outcomes, and makes informed recommendations about whether the campaign is driving net-new business or claiming attribution on traffic that was already converting. This judgment call is irreducibly human.
The AI Max shift: what changed in 2025 and 2026
If you have been researching PMax management, you may have come across references to "AI Max for Search campaigns." This is worth clarifying because the name creates genuine buyer confusion about whether it is the same as Performance Max or a replacement for it.
AI Max for Search campaigns is a separate campaign type. Google's AI Max for Search documentation describes it as a distinct offering in the campaign type menu. AI Max brings intelligent targeting expansion and text customisation to standard keyword-based Search campaigns, matching ad copy to landing page content and expanding reach beyond the literal keywords in your list. Performance Max, by contrast, eliminates the keyword structure entirely and spans all of Google's channels from a single objective-based campaign. The PMax overview in Google's own help centre frames this directly: PMax is designed to complement Search, not replace it.
The two products now represent a genuine choice point for Search-heavy advertisers:
| Dimension | AI Max for Search | Performance Max |
|---|---|---|
| Campaign structure | Keyword-based Search, with AI expansion | No keywords; goal-based across all channels |
| Inventory | Search (Google and Search Network) | Search, Display, YouTube, Discover, Gmail, Maps |
| Creative format | Responsive Search Ads (text); dynamic headline/description from landing page | Text, image, video; full asset groups |
| Right for | Advertisers who want Search intent coverage with wider reach, retaining keyword control | Advertisers whose primary goal is conversion volume across all inventory with minimal structural management |
| Human input required | Keyword strategy, negative match discipline, landing page quality | Conversion architecture, asset quality, audience signals, incrementality testing |
Neither is automatically superior. The right choice depends on your conversion volume, your creative production capacity, and whether intent-signal (keyword) control matters to your brand or vertical. For regulated verticals including financial services and insurance operating across Singapore, Australia, the US, and Canada, keyword control often matters more than Google's default guidance suggests, because policy restrictions and brand safety requirements are harder to enforce at a campaign level than at a keyword level.
See also: Do you still need a Google Ads agency in 2026? A buyer's guide for the broader question of how AI is reshaping the agency model across all Google campaign types, not just PMax.
Agency, consultancy, or in-house: the decision flowchart
The model that is right for you depends on three variables: your monthly conversion volume, whether your bottleneck is execution capacity or strategic clarity, and the internal competence of your team to interpret Google Ads diagnostics. The flowchart below walks the decision for a typical mid-market advertiser in Singapore, Australia, the US, or Canada.
A few notes on using this correctly. "Internal analyst able to run PMax" means someone who understands conversion taxonomies, can read Google Ads asset performance reports and search term themes, and has enough time to act on what they find monthly. It does not mean someone who can log into the platform. The in-house path only works if creative production keeps asset groups refreshed, because stale assets degrade performance without generating an obvious error state.
For further context on this decision across the full Google Ads product suite, not just PMax, see our post on Google Ads management and Do you still need a Google Ads agency in 2026?
What to demand from any PMax partner
The practical risk when hiring a PMax partner is not that they will do the work badly. It is that the work they do is the wrong work. If a partner spends their first month restructuring ad groups that PMax does not use, writing search queries you cannot target directly, or reporting on impression share metrics that are irrelevant to a goal-based campaign, you are paying for agency theatre while the actual performance variables go untouched.
Five questions to put to any PMax candidate before signing:
- How will you measure whether this campaign is driving net-new conversions? An answer that does not include incrementality testing, holdout groups, or geo lift is incomplete. PMax's own conversion reports are attribution reports, not incrementality reports.
- How will you connect our CRM data to the campaign? If the answer is "we'll use the conversion tag on the thank-you page," the signal quality will be limited to online actions only. Offline conversion imports or Enhanced Conversions for leads are the standard for anything with a sales cycle.
- How many asset groups do you plan to create, and why? The correct answer depends on your product range and audience diversity. A flat single-asset-group structure for a business with multiple product lines or markets is a signal that the partner is operating at template depth, not account depth.
- How will you handle brand keyword cannibalism? Brand exclusion lists at the campaign level are the standard fix. A partner who says "PMax doesn't really affect branded Search" has not run PMax through a budget-constrained period on a multi-campaign account.
- What will week eight look like? The learning phase for PMax typically runs four to six weeks. A good partner should be able to describe what diagnostic signals they will read at week eight and what decisions those signals will drive. If the week-eight plan is "review performance and optimise," that is not a plan.
Regulated markets note
If you operate in financial services, insurance, or fintech in Singapore, Australia, the US, or Canada, add a sixth question: how will you ensure PMax's placement and audience expansion settings comply with our sector's advertising policies? Account-level policy restrictions carry through to PMax, but placement-level brand safety requires explicit setup.
Retainer, project, or performance: which engagement model fits?
The engagement structure matters as much as who you hire, and the right structure for PMax management is different from the right structure for traditional search campaign management because the ongoing work is different in character.
| Model | What it covers | Right when | Watch for |
|---|---|---|---|
| Retainer | Monthly asset refreshes, signal maintenance, reporting cadence, ad hoc testing | You have continuous conversion volume and want consistent senior oversight month over month | Partners who use retainer hours on tasks the algorithm already handles, report on vanity metrics rather than conversion quality signals |
| Project | A defined scope: account restructure, asset group architecture, conversion tracking rebuild, PMax launch with internal handover | You have internal capacity to manage after setup, but need expert hands for the architecture phase | Handover documentation quality; an internal team inheriting a black-box setup cannot optimise what they cannot diagnose |
| Performance | Partner fee tied to conversion outcomes (cost per acquisition, revenue, or pipeline) | You have high conversion volume, clean attribution, and a conversion event that maps directly to revenue | Conversion event definition looseness: a performance-paid partner has an incentive to define the conversion event broadly. Insist on qualified pipeline or closed revenue, not raw form fills |
One practical note on retainers specifically: the monthly management work in PMax is lighter than it was in manual CPC Search management, because you are not setting bids, writing new ad variations weekly, or managing keyword lists at scale. A retainer priced as if it were a high-touch manual Search engagement is overpriced for what PMax actually requires. The monthly work is diagnostics, signal maintenance, creative refreshes, and test design. Scope it accordingly.
If you are evaluating a Google Ads partner more broadly, the considerations for Search campaigns, Shopping, and Demand Gen run alongside PMax in most accounts. Our Google Ads management page covers the full picture, and if you are thinking through the broader question of AI versus traditional agency models in 2026, read Do you still need a Google Ads agency in 2026?
For clients in the travel and insurance sector operating across Singapore and APAC, see also how PMax intersects with compliance requirements at our insurance industry page.
