Key Insight
Performance Max and AI Max campaigns hide most of their search term data. Google shows themed clusters, not individual queries. This makes traditional auditing impossible. But there are ways to audit these campaigns using the signals you do have, and catch wasted spend that Google does not show you directly.
The Visibility Problem: What PMax and AI Max Show vs What They Hide
In standard Search campaigns, you can see every query that triggered your ad, how much it cost, and whether it converted. That transparency is the foundation of traditional PPC auditing.
Performance Max and AI Max do not provide this level of detail.
What you can see: Search Terms Insights (themed query clusters, not individual queries), asset performance labels (low, good, best), audience segment insights, placement reports (with limited detail), geographic performance, device splits, and time-of-day patterns.
What you cannot see: Individual search queries that triggered your ads, exact spend per query, which asset combination was shown for which query, channel-level budget allocation in real time, and why the algorithm chose a particular placement or audience.
This opacity is by design. Google argues that AI needs freedom to optimise without advertisers micromanaging every decision. But it also means you cannot audit PMax and AI Max the same way you audit Search campaigns. You need a different framework.
The 8-Point PMax and AI Max Audit Framework
You cannot audit what you cannot see. But you can audit the inputs you control and the outputs Google reports. Here are the eight checkpoints that cover both.
- Conversion action quality. Are you tracking the right conversions? Are there duplicates, spam, or micro-conversions inflating your numbers?
- Asset group performance. Which asset groups are driving results? Are any underperforming consistently?
- Audience signal effectiveness. Are your audience signals attracting buyers or just traffic?
- Placement reports for spam. Are your ads appearing on legitimate sites or junk placements?
- Geographic performance. Is spend concentrated in your target areas or leaking to irrelevant locations?
- Device splits. Is one device type consuming disproportionate budget with poor results?
- Time-of-day patterns. Are conversions concentrated in specific hours? Is budget being spent overnight on low-quality traffic?
- Product-level ROAS (for Shopping). If you run PMax for e-commerce, which products are driving return and which are burning budget?
Let us break down the most critical of these checkpoints.
Conversion Tracking Is the Foundation
If your conversion tracking is wrong, everything PMax and AI Max do is wrong. These campaigns optimise towards whatever you define as a conversion. Feed them bad data, and they optimise towards bad outcomes.
The conversion tracking audit should be your first step. Check for these common issues:
- Duplicate conversions. The same form submission counted twice because both Google Ads and Google Analytics fire separate conversion tags. This doubles your reported conversion volume and halves your reported CPA, making campaigns look twice as effective as they are
- Spam submissions. Bot traffic submitting forms with fake data. PMax and AI Max count these as real conversions and optimise to find more traffic that behaves the same way
- Internal traffic. Your own team clicking through ads and filling out test forms. If your IP filters are not set up, these count as conversions
- Micro-conversions counted as macro. Newsletter signups, PDF downloads, and page scroll events tracked as primary conversions alongside lead form submissions. Smart Bidding treats them all equally, so it will happily drive cheap newsletter signups instead of expensive qualified leads
For a complete audit of your tracking setup, our Google Ads audit includes a full conversion tracking review as the first checkpoint.
Asset Group Audit: What to Check and When to Replace
Each asset group in PMax should target a distinct product, service, or audience theme. If you have one catch-all asset group with generic creative, the algorithm lacks the signals it needs to match the right creative to the right audience.
Check asset performance ratings in the Google Ads interface. Each asset (headline, description, image, video) gets a label: Best, Good, Low, or Pending. The audit process:
- Remove "Low" assets that have had sufficient impressions (at least 1,000). If an asset consistently underperforms, it is dragging down the entire asset group
- Test replacements. Replace removed assets with new variations. Test different angles: benefit-focused vs feature-focused, specific vs general, urgency vs authority
- Maximise asset slots. Provide the maximum number of assets per type. More assets give the algorithm more combinations to test, which accelerates learning
- Align creative with intent. Each asset group should have creative that speaks to a specific audience need. Generic creative that could apply to any product or service gives the AI no differentiation signal
For strategic guidance on how creative drives performance in AI-powered campaigns, see our article on why creative is the new targeting.
Audience Signals Audit: Suggestions, Not Targeting
PMax audience signals are suggestions, not hard targeting constraints. Google uses them to inform the initial learning phase but will expand beyond your signals once the campaign has enough data.
This means your audience signals influence who sees your ads early on, but they do not limit reach long term. The audit question is: are your signals pointing the algorithm towards your actual buyers, or towards assumptions that do not match reality?
- Customer match lists. Upload your actual customer email list, not a list of newsletter subscribers or webinar attendees. The closer your customer match data is to people who actually paid, the better the signal
- Custom segments. Review the search terms and URLs you used to build custom segments. Are they still relevant? Do they reflect how your buyers actually research and purchase?
- In-market and affinity audiences. Check whether the audiences you selected match your CRM data. If your audience signal says "In-market for Business Software" but your actual buyers are small business owners researching accounting tools, the signal is too broad
Revisit audience signals quarterly, or whenever you notice a shift in lead quality from PMax campaigns.
The Negative Keyword Workaround
PMax now supports account-level negative keywords. AI Max uses campaign-level negatives. Both need regular updates because AI-driven campaigns continuously discover new queries, and some of those queries will be irrelevant.
Since you cannot see individual search terms in PMax, build your negative lists from these sources:
- Search Terms Insights themes. If you see a themed cluster about jobs, careers, or tutorials, add those category terms as negatives
- Standard Search campaign data. Your Search campaigns still show individual queries. Use them as a proxy for what PMax might also be matching
- Google Search Console. Check which queries are driving traffic to your landing pages. If you see irrelevant queries appearing, they may be coming from PMax
- CRM lead quality data. If leads from PMax are consistently low quality in specific categories, build negatives around those themes
For a comprehensive starting point, download our Performance Max management guide which includes negative keyword recommendations specific to PMax campaigns.
When to Restructure vs When to Pause
Not every underperforming PMax or AI Max campaign can be fixed. Sometimes the best move is to pause and start fresh. Here are the signals that indicate each path.
Restructure when:
- Performance was good but has declined over two to three weeks, suggesting a specific change caused the drop
- You can identify a clear issue (wrong conversion actions, poor asset groups, weak audience signals) that has a specific fix
- The campaign has historical conversion data the algorithm can learn from
Pause and rebuild when:
- CPA has risen consistently for four or more weeks with no clear cause and no improvement despite optimisation
- Impression volume is spiking on irrelevant placements (check the placement report for junk sites and mobile apps)
- Conversion quality has declined steadily, confirmed by CRM data showing lower lead scores or close rates
- The campaign was built with weak foundations (poor conversion tracking, generic creative, broad audience signals) that infected the algorithm's learning data
A fresh campaign with better inputs (clean conversion data, specific creative, precise audience signals) often outperforms trying to rehabilitate a campaign built on bad data. The algorithm carries its learning history, and if that history is built on poor signals, it is sometimes faster to start over.
If you are unsure whether your PMax or AI Max campaigns are salvageable, our PPC management team can run a full diagnostic and recommend restructure or rebuild based on your account data. For the broader context of how AI is reshaping campaign management, see the AI PPC Playbook 2026.