Listen to this article · 13 min listen

Key Takeaways

  • Configure PMax right by giving it specific conversion goals inside Google Ads, think lead form submissions or high-value page views, not just generic website traffic.
  • Use precise exclusion lists for your brand terms and any junk URLs to stop PMax from wasting budget or cannibalizing the traffic your other campaigns already own.
  • You have to live in the “Insights” report in Google Ads. Check “Consumer Interests” and “Search Themes” constantly to find the weird, unexpected conversion paths the AI uncovers and then use that to refine your asset groups.
  • Feed the AI better signals by integrating your first-party data through Customer Match lists, which makes targeting way more accurate and improves how PMax’s automated bidding works.
  • Look past the basic “Conversions” metric. To see what PMax is really doing, you need to track “Conversion Value” and “New Customer Acquisitions” in the 2026 Google Ads interface to find the hidden value generated by its AI agent pathing.

PMax campaigns use their AI to find unseen conversions all over Google’s network, which totally changes how we have to think about running ads. The hard part is setting them up correctly so you can see past the surface metrics and find the hidden value buried in the AI agent path, especially since direct attribution is so murky.

20%
Average increase in conversion rates
5
Minimum field images for asset groups
15
Seconds or longer for video assets

Step 1: Laying the Foundation for PMax Success

You can’t just dive in and launch a PMax campaign. You have to do the prep work first, which means setting clear goals, auditing your creative assets, and figuring out how this campaign will fit with everything else you’re running. If you skip this, even the smartest AI is going to struggle.

1.1 Define Specific Conversion Goals

First thing’s first: you need to tell Google what a good conversion actually is. Go to Tools and Settings > Measurement > Conversions in Google Ads Manager and set up your most important conversion actions. For a B2B SaaS client, for example, a “Demo Request” or “Free Trial Sign-up” is way more important than a generic “Contact Us” submission. You have to assign real conversion values to these actions. Maybe a finished sale is worth $100, but a newsletter signup is only worth $5. PMax’s bidding AI uses these values to decide what’s worth chasing, so it will prioritize the actions that actually make you money. A conversion goal of “website traffic” is a recipe for disaster.

1.2 Audit and Prepare Your Creative Assets

PMax needs a lot of different creative assets to do its job. In 2026, you need a richer mix than ever. You should have a folder ready with high-quality images (field, square, portrait), videos (make them at least 15 seconds), a bunch of headlines (short and long), descriptions, and your business names. All these assets have to look and sound like your brand and be ready for any placement from YouTube to Gmail. The AI mashes all these elements together to create ads on the fly, so variety and quality are everything. I’ve seen campaigns completely underperform just because they were launched with only two headlines and one image. That’s not nearly enough for the AI to learn.

1.3 Establish Negative Keywords and Brand Exclusions

PMax is designed for broad reach, but you absolutely need strategic exclusions to prevent budget waste and protect your brand campaigns. Go to Campaigns > All Campaigns > PMax Campaign > Settings > Brand Exclusions and upload a list of your brand terms. This stops PMax from bidding on queries that your dedicated Search campaigns are already handling. Then, go to Campaigns > All Campaigns > PMax Campaign > Settings > Account-level negative keywords and add any generic junk terms that attract the wrong kind of traffic. This is a critical step to prevent the AI from misinterpreting a user’s search and showing ads for terms that have zero chance of converting.

Step 2: Configuring Your PMax Campaign for AI Agent Pathing

With all the prep work done, it’s time to actually build the PMax campaign. The settings you choose here are what will guide the AI agent to find and optimize for real PMax conversions.

2.1 Create a New PMax Campaign

Inside Google Ads Manager, just click Campaigns > New Campaign. Pick your main goal, like “Sales” or “Leads,” and then on the next screen, you’ll choose “Performance Max” as the campaign type. This kicks off the AI-driven setup. The next few steps are where you actually get to steer the AI.

2.2 Set Your Budget and Bidding Strategy

Put in your daily budget when it asks. For the bidding strategy, choose either “Conversions” or “Conversion Value”. If you already have solid conversion tracking set up with values assigned to each action, “Conversion Value” with a target ROAS (Return On Ad Spend) is almost always the better choice because it tells the AI to chase the most profitable customer actions. If you’re starting from scratch without much data, starting with “Conversions” and a target CPA (Cost Per Acquisition) is a safer bet. Just remember, the AI needs a ton of data to learn, so you have to be ready for performance to be all over the place for a few weeks during that initial learning phase. A 2024 HubSpot report found that companies using AI-powered bidding saw a 20% average increase in conversion rates which shows how much you gain by letting the AI handle this.

2.3 Define Asset Groups and Audience Signals

This is the absolute heart of PMax’s AI agent path. An Asset Group is just a bucket of your creative assets (images, videos, text) and the audience signals that PMax uses to build ads. You should make at least 3-5 different asset groups, with each one focused on a slightly different angle of your product or audience.

2.3.1 Upload Diverse Creative Assets

Inside each asset group, you need to upload all the assets you prepared earlier. Make sure you’ve got:

  • At least 5 field images (1200×628)
  • At least 5 square images (1200×1200)
  • At least 1 portrait image (900×1200)
  • At least 1 video (15 seconds or longer)
  • Up to 5 long headlines (90 characters)
  • Up to 5 short headlines (30 characters)
  • Up to 5 descriptions (90 characters)
  • 1 long description (300 characters)
  • Your business name and logo

The AI will test out different combinations of these, so giving it a big library to pull from helps it find winning combinations faster.

2.3.2 Provide Strong Audience Signals

Under the “Audience signals” section, you need to add relevant audiences. This just gives the AI a starting point for who your ideal customer is, which speeds up the learning process. You should include:

  • Your first-party data: Uploading Customer Match lists (like emails and phone numbers) from your current customers or best leads is extremely effective.
  • Custom Segments: You can build segments based on what your ideal customers search for or what websites they visit.
  • Your website visitors: Connect your Google Analytics 4 (GA4) property so you can target audiences of people who’ve already been to your site.
  • In-market and Affinity audiences: Use Google’s pre-built audiences that are relevant to your business.

The AI uses these signals as its initial clues and then expands its reach to find similar people who are likely to convert. Think of it like giving a smart detective a handful of good leads. They’ll use those to find the whole network.

2.4 Configure Final URL Expansion

In your PMax campaign settings, you’ll see “Final URL Expansion” with two choices:

  • “Send traffic to the most relevant URLs on your site”: This is the default setting and usually the one you want. The AI will pick the best landing page on your site for a user based on their search and what it thinks their intent is. This is where the AI agent path can really work, finding pages you never would have thought to send certain traffic to.
  • “Only send traffic to the URLs you’ve provided”: Only use this if you have a very small, specific set of landing pages and you’re 100% sure no other page on your site is relevant. I generally advise against this because it completely handcuffs the AI’s ability to discover new conversion paths.

If you go with the first option, just make sure your site has good internal linking and that all potential landing pages have clear calls to action.

Step 3: Uncovering Hidden Value and Optimizing PMax Conversions

Launching is the easy part. The real work is figuring out what’s happening inside PMax’s black box. A strategic analysis of the reports Google gives you is the only way to understand the unseen conversions and follow the AI agent path.

3.1 Analyze the “Insights” Report

In Google Ads, go to Campaigns > All Campaigns > PMax Campaign > Insights. This report is the best view you’re going to get into what the AI is learning.

  • Consumer Interests: Look for patterns here. Is the AI finding converters in unexpected user categories? For example, a B2B software company might discover a big chunk of its conversions are coming from people interested in “sustainable business practices,” even if the software has nothing to do with sustainability. That’s a new value proposition the AI has found on its own.
  • Search Themes: PMax won’t give you keyword-level data, but the “Search Themes” report shows you aggregated data on the *types* of queries driving conversions. You need to watch for themes that are converting well but aren’t things you’re targeting in your regular Search campaigns. These are perfect examples of the AI agent path finding new pockets of opportunity. When you see a high-performing theme, should you build a new asset group for it? Maybe even a new Search campaign?

A 2025 Nielsen report showed that marketers who actively used AI insights to adjust their campaigns saw a 15% ROI improvement over those who just set their AI campaigns and forgot about them.

3.2 Evaluate Performance Beyond “Conversions”

Don’t just stare at the raw “Conversions” column. You have to dig in.

  • Conversion Value: If you set up conversion values, you need to be watching “Conversion Value” and “ROAS.” This tells you if the AI is bringing in high-value customers, or just a lot of low-value actions.
  • New Customer Acquisitions: By 2026, the Google Ads reporting for “New Customer Acquisition” is much better. If that’s a goal for you, make sure you’ve set it up correctly in your conversion settings. PMax’s AI is really good at finding new customer segments.
  • Store Visits/Store Sales: For any business with a physical store, you have to monitor these metrics to see how PMax is driving offline behavior. The AI agent path could be sending people into your store through local inventory ads or map results.

3.3 Use Exclusion Lists for Underperforming URLs

Even with Final URL Expansion turned on, some pages on your site might just be duds that get clicks but never convert. If your “Insights” report or GA4 data shows that certain landing pages are dragging down PMax performance, you need to exclude them. Go to Campaigns > All Campaigns > PMax Campaign > Settings > Final URL Exclusions and add the URLs you want PMax to ignore. This cleans up the AI agent path and forces the algorithm to send traffic to pages that actually work.

3.4 Iterative Asset Group Refinement

You can’t just set up PMax and walk away. You have to regularly check the “Asset Group Performance” report. Any assets that Google flags as “Low” or “Poor” need to be swapped out with new creative. If a headline is bombing, replace it. The AI learns from these changes and gets better at finding the best combinations. I’ve found that refreshing at least 20% of the assets every month can keep performance from going stale. Modern PPC is complicated, and a tool like PMax is necessary for finding conversions your other campaigns miss. By setting up your campaigns carefully and actually paying attention to the “Insights” report, you can begin to understand the AI agent path and find significant hidden value. Proving PPC impact in 2026 is going to depend on getting tools like PMax right.

How does PMax’s AI agent path differ from traditional campaign targeting?

The big difference is that PMax’s AI path is dynamic. It identifies and targets users across every Google channel (Search, Display, YouTube, Gmail, Discover) using real-time signals and your conversion goals. Traditional campaigns use more rigid, pre-set audience segments and keyword lists, while PMax’s AI is always exploring new opportunities beyond those initial signals to find users who are most likely to convert.

What is the “black box” nature of PMax, and how can I still gain insights?

The “black box” idea refers to how little visibility we get into specific keyword performance or exact ad placements, since the AI handles almost all the targeting. You can still get good insights by constantly checking the “Insights” report for “Consumer Interests” and “Search Themes,” watching PMax traffic behavior in Google Analytics 4, and analyzing the “Asset Group Performance” report to see which creative is working. These aggregated insights help you understand what the AI is learning, even without the old granular keyword reports.

Can PMax cannibalize my existing Search campaigns?

Yes, PMax can definitely steal traffic from your Search campaigns if you don’t manage it. To stop this, you must use Brand Exclusions in your PMax campaign settings. This stops PMax from bidding on your exact brand terms, which lets your dedicated Search campaigns capture that high-intent traffic. For non-brand terms, PMax might compete, but its AI is built to find incremental conversions that your Search campaigns are probably missing anyway.

How often should I update my PMax assets and audience signals?

You should plan on updating your assets and audience signals on a regular basis, probably every 4 to 6 weeks. The AI agent needs fresh creative to test and new audience signals to keep its understanding of your ideal customer sharp. Keep an eye on the “Asset Group Performance” report and swap out anything that’s underperforming. You should also update your Customer Match lists with new customer data periodically to keep feeding the AI good information.

What are the most common mistakes marketers make with PMax?

The most common mistakes are giving it too few creative assets, forgetting to set up brand exclusions, using vague conversion goals without assigning values, and just generally adopting a “set-it-and-forget-it” approach. Another big error is not connecting Google Analytics 4 to get deeper behavioral data. PMax needs active management and constant small adjustments to perform at its best and find those valuable, unseen conversions.