Performance Max now drives 45% of Google Ads conversions across all channels. The advertisers getting the best results aren’t the ones with the biggest budgets — they’re the ones who give Google’s algorithm the strongest starting signals.

Audience signals are the primary lever you have to influence who PMax shows your ads to. But they work differently from traditional audience targeting, and most advertisers either misunderstand them or configure them badly. Upload “all website visitors” as your signal and you’ve given Google almost no useful information. Upload a customer match list of your top 500 buyers by lifetime value, and you’ve given the algorithm a clear pattern to replicate.

This guide covers what audience signals actually do, which signal types produce the best results, how to structure asset groups around them, and the critical complementary features — search themes, brand exclusions, URL controls — that most guides ignore but that directly determine whether your PMax campaigns find new customers or just cannibalize your existing traffic.

What Audience Signals Actually Do

Audience signals are suggestions, not targeting constraints. When you add a customer list, a custom segment, or an in-market audience to an asset group, you’re telling Google: “People like this tend to convert for my business. Start here.”

Google’s algorithm takes that starting point and expands beyond it. If it finds users outside your specified signals who convert at a higher rate, it will shift spend toward those users. You cannot prevent this — PMax is designed to optimize for conversions, not to stay within your signal boundaries.

This distinction matters for how you evaluate performance. If your Audience Insights report shows that 40% of conversions come from audiences outside your signals, that’s the algorithm working as intended — not a malfunction.

The signals serve their most critical function during the learning phase (typically 2-4 weeks for initial optimization, 6-8 weeks for full performance). Without signals, Google starts from zero and tests broadly. With strong signals, it starts with a pattern of what your ideal customer looks like and reaches optimal performance faster.

When Audience Signals Matter Most

New campaigns with no conversion history. PMax needs conversion data to learn. Without it, audience signals are the primary guidance the algorithm has.

Accounts with fewer than 20 monthly conversions. If your account doesn’t generate enough conversion volume for the algorithm to learn independently, audience signals compensate by providing the patterns Google can’t yet derive from your data. If you’re below 20 conversions per month total, consider building conversion volume with Search campaigns first before launching PMax.

High-value, infrequent conversion events. Lead generation, B2B sales, or luxury purchases where individual conversions happen rarely — signals guide the algorithm toward the right user profile when it can’t learn from volume alone.

Signal Types Ranked by Impact

1. First-Party Data (Highest Value)

Your own customer data is the strongest signal type because it represents people who have actually purchased from or engaged with your business. Google uses these lists to find similar users.

Customer match lists are the most powerful signal. Upload your buyer email list, segmented by value tier. A list of customers who spent $200+ outperforms a generic “all customers” list because it gives Google a more specific pattern to replicate.

Segment by value, not just status. Instead of one “all buyers” list, create:

• Top 20% customers by revenue (your best signal for finding high-value lookalikes)
• Repeat purchasers (signals loyalty and product-market fit)
• Recent purchasers (last 90 days — signals current buying patterns)
• Cart abandoners (high-intent users who didn’t convert)
• Product page viewers (category-specific signals for category-specific asset groups)

2025 update: Google removed the previous 1,000 user minimum for Customer Match lists. You can now use smaller lists, making this signal viable even for B2B accounts with modest CRM databases. Update your lists regularly — stale data degrades signal quality over time.

2. Custom Segments (High Value)

Custom segments are the closest thing to keyword targeting available in PMax. You build audiences based on:

Search terms: Add terms your best customers search for. Create a segment with your top 25-50 converting search queries. This directs PMax toward users with demonstrated purchase intent.

URLs: Specify competitor websites or industry publications. Google finds users with browsing patterns similar to visitors of those sites. This isn’t targeting visitors of those specific URLs — it’s finding people with matching interests.

Apps: Specify apps your target audience uses. Google builds a profile of users who download similar apps.

Custom segments work exceptionally well for competitor targeting. Add your top 5-10 competitor brand URLs, and Google will find users who research those competitors — intercepting them during the consideration phase.

3. In-Market Audiences (Medium Value)

Google’s pre-built audiences of users actively researching or shopping in your product category. These are users showing current purchase intent through their search and browsing behavior.

In-market audiences work well as a supplement to first-party data but shouldn’t be your only signal type. They’re broad by nature — “In-Market for Running Shoes” includes everyone from casual joggers to ultramarathon runners.

Limit yourself to 5-10 highly relevant in-market categories. Adding too many dilutes the signal.

4. Affinity Audiences (Lower Value for Direct Response)

Affinity audiences represent long-term interests, not active purchase intent. Users in “Fitness Enthusiasts” may have been interested in fitness for years without currently shopping for fitness products.

These work better for brand awareness objectives. For direct response campaigns with conversion goals, prioritize first-party data, custom segments, and in-market audiences over affinity.

5. Demographics (Supplementary)

Age, gender, household income, parental status — useful as refinement layers but rarely strong enough as standalone signals. A jewelry brand targeting high-income households should layer income demographics on top of in-market and custom segment signals, not rely on demographics alone.

Structuring Asset Groups Around Signals

Asset groups are the only meaningful structural control you have inside PMax. How you organize them determines whether you can identify what’s working, test different approaches, and match creative to audience intent.

The Architecture That Works

Each asset group should align three elements: audience signal, creative assets, and landing page. All three should serve the same customer segment and tell the same story.

Wrong approach: One asset group with all your products, all your creative, and broad audience signals. This forces Google to serve the same creative to every user across every channel. You can’t tell what’s working because everything is blended.

Right approach: Separate asset groups by product category, customer segment, or purchase intent.

Ecommerce example:

• Asset Group 1: Running shoes — audience signal: custom segment (running search terms + competitor URLs) — creative: running-specific images and headlines — landing page: running shoe category page
• Asset Group 2: Casual shoes — audience signal: custom segment (casual/lifestyle search terms) — creative: lifestyle imagery — landing page: casual shoe category page

Lead gen example:

• Asset Group 1: Bottom-funnel — audience signal: customer match (past leads who converted) — creative: offer-focused headlines, testimonials — landing page: demo request page
• Asset Group 2: Mid-funnel — audience signal: custom segment (problem-awareness search terms) — creative: educational messaging — landing page: comparison or resource page

What Not to Do

Don’t duplicate asset groups with different signals but identical creative. This was a common tactic — creating three asset groups with the same images and headlines but different audience signals to test which signal performed best. In 2026, performance of these duplicates converges to the same average over time. The algorithm treats them as interchangeable.

Don’t use more than 7-10 asset groups per campaign. Beyond that, you split your conversion data too thin for the algorithm to learn effectively in any single group.

Search Themes: The Signal Most Advertisers Skip

Search themes were added to PMax in 2024 and allow you to specify up to 25 keyword themes per asset group. They guide PMax’s behavior on the Search and Shopping channels specifically — telling the algorithm which search queries are relevant to each asset group.

Most advertisers skip search themes or don’t know they exist. This is a mistake, especially for niche products where Google’s algorithm may not immediately understand your product category.

How to Use Search Themes Effectively

Add your top converting search queries from your Search campaigns as themes. If you’re running Search alongside PMax, export your best-performing keywords and use them as themes in the corresponding PMax asset group.

For new campaigns without Search history, use the terms your target audience would search. Focus on commercial and transactional intent terms — these direct PMax toward the Search inventory that drives conversions.

Search themes and audience signals work together: audience signals tell Google WHO to target, search themes tell Google WHAT queries are relevant. An asset group with both a strong customer match signal and targeted search themes gives the algorithm clear direction on both dimensions.

Protecting Your Brand Traffic

PMax will, by default, show ads on your brand search terms. This means your PMax campaign can claim credit for conversions that would have come through your organic listings or standard brand Search campaigns — inflating PMax’s reported performance while cannibalizing your other channels.

Brand Exclusions

Google added brand exclusion lists for PMax in 2023. Set them up:

1. Go to your PMax campaign settings
2. Navigate to Brand Exclusions (under Additional Settings)
3. Add your brand name and common misspellings
4. Optionally add any brand terms you don’t want PMax competing on

This forces PMax to focus on non-brand, incremental traffic — which is where its real value lies.

URL Expansion Controls

URL expansion is enabled by default and allows PMax to serve ads pointing to any page on your site — including career pages, blog posts, privacy policy, discontinued product pages, or any other URL that shouldn’t be an ad destination.

Add URL exclusions for pages that aren’t appropriate landing pages. Or disable URL expansion entirely if your campaign requires specific landing page control.

Monitoring and Optimizing

Channel Performance Report

This report shows how PMax distributes your budget across channels: Search, Shopping, Display, YouTube, Gmail, Maps, and Discover. Check it monthly.

If Display is consuming 40%+ of your budget with low ROAS, your audience signals may be too broad — PMax defaults to cheap Display impressions when it can’t find enough high-quality search traffic. Tighten your signals and add more specific search themes.

Audience Insights Report

Shows which audience segments drive conversions. Use it to:

• Identify over-converting segments you didn’t include in your original signals — add them explicitly
• Spot segments that receive spend but don’t convert — these indicate signal drift
• Discover unexpected audience patterns worth reinforcing or investigating

Search Terms Report

Limited in PMax compared to standard Search campaigns, but still valuable. Review monthly to:

• Identify high-converting query categories you should add as search themes
• Spot irrelevant queries that indicate signal or URL expansion issues
• Find patterns that inform your Search campaign keyword strategy

Creative Refresh Cadence

PMax creative fatigues. Competitors recommend replacing your 3-5 lowest-performing assets every 6-8 weeks. Don’t do a full creative overhaul — incremental refresh lets you identify which new assets improve performance.

Accounts with sparse asset groups (3-4 images, 2-3 headlines) see 25-40% lower performance than fully populated groups. Fill every slot: upload 10-15 images, 8-12 headlines, 4-5 descriptions, and 3-5 videos. If you don’t upload your own video, PMax will auto-generate one from your images — and the quality is usually poor. In 2026, Google enabled AI voice-over on PMax video assets by default, making custom video uploads even more important for brand control.

Common Mistakes

Using broad, generic signals. “All website visitors” or “newsletter subscribers” gives Google almost no useful pattern to work from. Use specific, high-value segments: “customers who spent $200+” or “users who viewed product category X.”

Starting PMax without enough conversion data. If your account generates fewer than 20 conversions per month, the algorithm doesn’t have enough signal to optimize effectively. Build conversion volume with Search campaigns first.

Skipping search themes. For niche products especially, search themes help the algorithm find relevant queries faster. Use all available slots.

Not setting brand exclusions. Without them, PMax claims brand traffic and reports inflated ROAS while adding no incremental value.

Making changes during the learning phase. PMax needs 2-4 weeks to process new signals. Significant changes during this period restart the learning phase. Make changes in batches and wait for the learning period to complete before evaluating results.

Ignoring URL expansion. If you see conversions from irrelevant landing pages in your reports, URL expansion is sending traffic to pages that shouldn’t be ad destinations. Set up exclusions or disable the feature.

Frequently Asked Questions

Are audience signals required in Performance Max?

No, they’re optional. But campaigns without audience signals take significantly longer to exit the learning phase and typically perform worse during the first 4-8 weeks. For new campaigns or accounts with limited conversion data, audience signals are effectively mandatory for reasonable performance.

How long does it take for audience signals to impact performance?

Google’s algorithm typically needs 2-4 weeks to process new audience signals and begin optimizing based on them. Full performance optimization usually takes 6-8 weeks. Don’t make significant changes during this learning period.

Should I use one audience signal per asset group or combine them?

Separate signals into different asset groups whenever possible. This lets you see which signal types drive the best results through the asset group report. Combining all signals into one group makes it impossible to attribute performance to specific signal types. Exception: layering demographics on top of interest or custom segments within the same group is fine — that’s refinement, not mixing.

What’s the difference between audience signals and search themes?

Audience signals tell Google WHO to target (user profiles, behaviors, demographics). Search themes tell Google WHAT search queries are relevant (keyword themes). They complement each other — use both in every asset group for the strongest algorithmic guidance.

How do I prevent PMax from cannibalizing my brand traffic?

Set up brand exclusions in your PMax campaign settings. Add your brand name and common misspellings. This forces PMax to focus on non-brand incremental traffic. Also review your Channel Performance report — if PMax is spending primarily on Search with brand-related queries, brand exclusions aren’t working properly or aren’t set up.

Has the Customer Match minimum list size changed?

Yes. Google removed the previous 1,000 user minimum for Customer Match in 2025. You can now use smaller lists, making this signal type viable for B2B accounts and businesses with smaller customer databases.