Google Analytics 4 (GA4) Audiences: Predictive Targeting for Ads (2026 Guide)

Old Google Analytics (UA) was about "What happened?" Google Analytics 4 (GA4) is about "What will happen?"

The killer feature of GA4 is Predictive Audiences. It analyzes user behavior (scroll depth, time on site, previous path) to identify users who are likely to purchase in the next 7 days—even if they haven't added anything to cart yet.

In this "Mega-Authority" guide, we cover:

1. The Tech: How GA4 prediction works.
2. The Lists: The 3 must-have predictive segments.
3. The Sync: Connecting GA4 to Google Ads.
4. The Strategy: Using them for PMax signals.

Part 1: Theory - Probabilistic Modeling

GA4 uses machine learning to score every user session. If User A reads your blog, visits pricing, reads "Terms of Service," and stays for 5 minutes, GA4 assigns a High Purchase Probability Score. Standard remarketing misses this user (they didn't trigger "Add to Cart"). Predictive Audiences catches them.

Part 2: The 3 Predictive Audiences

Find them in Admin → Audiences → New Audience → Predictive.

1. Likely to Purchase (7 Days)

• The "Hot" list.
• Action: Add to Demand Gen and YouTube with aggressive bidding.

2. Likely to Churn (7 Days)

• Active users likely to stop visiting.
• Action: Target with a "Win-Back" offer (e.g., "We miss you, here is 10% off").

3. Predicted Top Spenders (28 Days)

• Users likely to generate high revenue.
• Action: Create a "VIP" Asset Group in PMax.

Part 3: Execution - Creating Custom Segments

Beyond prediction, GA4 allows complex logic.

Recipe: The "Heavy Researcher"

• Condition: session_duration > 180 seconds.
• AND
• Condition: page_view contains "pricing".
• AND
• Exclude: purchase event.

This creates a list of people who are deeply interested but hesitating. Perfect for ads that address Objections (Trust, Warranty, Support).

Part 4: The Sync Mechanism

Lists in GA4 do nothing unless they are in Google Ads.

1. Admin → Product Links → Google Ads Links.
2. Link your account.
3. Important: Turn on "Enable Personalized Advertising".
4. Wait 24 hours. The audiences will appear in Google Ads Audience Manager automatically.

Part 5: Summary & Checklist

If you aren't using GA4 audiences, you are remarketing with a blunt instrument.

Your Action Plan:

1. Link GA4 to Google Ads today.
2. Create the "Likely to Purchase (7 Days)" predictive audience.
3. Apply it as a signal to your PMax campaign.
4. Create a "Heavy Researcher" custom segment.

Predict the future, don't just report on the past.

Membership Duration — The 7-Day vs 30-Day vs 540-Day Framework

Audience membership duration is not a default — it's a strategic decision.

• 7-Day Windows: Maximum urgency. These users are in active buying mode. If they don't convert within 7 days, the intent signal has expired. Accept higher CPAs here — you're targeting your hottest prospects.
• 30-Day Windows: Mid-funnel nurturing. Ideal for B2B prospects with longer evaluation cycles. Keeps you visible during comparison-shopping phases.
• 540-Day Windows: Brand ecosystem. Use for YouTube viewers and site visitors to maintain top-of-mind awareness. These campaigns run on minimal budgets — the goal is staying in the consideration set, not driving immediate conversions.

Rule: Create separate audiences for each duration tier. Do not lump 7-day and 540-day users together — their intent and required messaging are fundamentally different.

Custom Dimensions — Building Audiences Beyond Behavior

Predictive audiences are powerful but generic. Custom Dimension audiences let you target based on what users actually did inside your product.

SaaS Example:

• You fire an event called viewed_api_docs when users visit developer documentation
• Users who view API docs are Developers (influencers), not CEOs (buyers)
• Action: Create a GA4 audience: "Users who fired viewed_api_docs"
• Exclude this audience from "Request Demo" campaigns (stop paying for devs clicking demo ads)
• Include in "Start Free Trial" campaigns (developers want to self-serve)

Ecommerce Example:

• You track purchase_value as a custom parameter on every order
• Action: Create a "Whales" audience: "Users where lifetime purchase_value exceeds $500"
• Targeting: Apply as RLSA observation layer — bid aggressively when Whale-tier users search generic terms like "luxury shoes"

The combination of Predictive + Custom Dimension audiences is the most precise retargeting stack available outside a CDP.