Let me paint a picture you might relate to.
You’re running Performance Max. Or Smart Bidding on Search. Maybe even the newer AI Max on Search campaigns.
The setup looks clean. The campaigns are active. You know your campaigns are learning.
So you wait. Two weeks pass. A month passes. The CPA creeps up. The ROAS doesn’t move. You start questioning whether AI campaigns actually work, or whether you’re just not using them right.
Here’s the uncomfortable truth most advertisers don’t want to hear:
The algorithm isn’t broken. Your data is.
Google’s AI, or any platform’s per se, can only be as smart as what you feed it with. And in most accounts I audit, what’s being fed to the algorithm is incomplete, inconsistent, or sometimes completely wrong.
I said this before in my AI-Ready Search Campaigns guide — garbage in, garbage out. But I didn’t go deep enough on the “garbage” part. That’s what this post is about.
We’re going to tear apart the most common conversion tracking mistakes that silently destroy AI-powered campaigns’ performance. And then, I’ll walk you through exactly how to audit and fix them.
No fluff. Just the stuff that actually matters.
Why This Matters More Than It Ever Has
Google Ads has fundamentally changed.
Three years ago, you could get away with imperfect tracking. You had manual levers. You could adjust bids yourself, pause keywords, and course-correct based on your own judgment.
That’s not the game anymore.
Google’s AI now updates its systems several times per week, and the pace of change has accelerated many times compared to 2024. Smart Bidding, Performance Max, AI Max, every major campaign type is now driven by machine learning. And machine learning has one non-negotiable requirement:
It needs accurate conversion data to function.
This isn’t a suggestion. It’s the foundation everything else is built on.
In my experience and also validated by industry peers, for AI Max campaigns to work well and not be erratic, a minimum of 30 to 50 conversions in the past 30 days before launch is ideal. If you have 100 or more it delivers meaningfully better results. But here’s the thing — having enough conversions isn’t the problem for most advertisers.
The problem is that the conversions being tracked are wrong.
And wrong conversion data doesn’t just fail to help the AI. It actively teaches it the wrong lessons.
The Privacy Wall Making This Even Harder
Before we get into the mistakes, you need to understand what’s happening on the tracking side of things right now.
Research by Cookie Script shows that only about 31% of users globally accept tracking cookies. That number varies by country, some regions sit as low as 20%.
What does that mean in practice?
It means that in many markets, two-thirds of your traffic is invisible to standard cookie-based tracking.
Google rolled out Consent Mode V2 in late 2023 to deal with this, and enforcement started in early 2024. The idea is sound. If you implement it correctly, specifically Advanced Consent Mode, Google can use conversion modelling to estimate conversions from users who denied cookies. Advertisers typically see a 15–25% uplift in reported conversions from this modelling alone.
But for the modelling to activate and function, you need at least 700 ad clicks over 7 days per country and domain, plus a consent rate of roughly 20% or higher.
Most advertisers either haven’t set this up, or have set it up incorrectly, which means they’re flying blind on a majority of their traffic, while still spending real money.
Okay, that was the the backdrop. Now let’s talk about what’s actually going wrong inside your account.
The 5 Tracking Mistakes That Are Silently Killing Your AI Campaigns
I’ve audited hundreds of accounts over the years. These are the issues that show up again and again, and each one has a direct, measurable impact on how well your AI campaigns perform.
1. Duplicate Conversion Tracking
This is the most common mistake I find in accounts. And it’s surprisingly easy to miss.
It happens when the same conversion event fires more than once. A thank-you page that reloads. A Google Tag firing manually on the page and through GTM at the same time. A form submission that triggers both a GA4 event and a direct Google Ads conversion tag. The scenario can be different.
The result? Google sees 2 or 3 conversions where there’s actually 1.
Your campaign looks like it’s performing three times better than it actually is. The AI learns from this inflated signal. The worst part, when you eventually fix it, performance appears to drop, but in reality, you’re just seeing the truth for the first time.
A recent GTM update in April 2025 made this even trickier, because Google Tag now auto-loads in certain configurations. If you haven’t audited your setup since then, there’s a real chance you’re double-firing without knowing it.
How to check: In Google Ads, go to Goals → Conversions. Look at the “Repeat rate” column. If your repeat rate is significantly above 1.0, you’re likely tracking duplicates.
2. Tracking the Wrong Event as a Conversion
This one is even more dangerous, because it doesn’t just give the AI bad numbers, it teaches the AI to optimise for the wrong thing entirely.
I’ve seen accounts where “Add to Cart” is set as a primary conversion. Others where a button click on a contact page counts as a lead, regardless of whether anyone actually fills in the form. See, there can be specific scenarios where you might need those actions to optimize your campaigns for. But still, they do not need to be your primary goals. You can use them as secondary and them create a Custom Conversion to use them with your specific top of the funnel campaigns.
I recall an agency’s account which I was consulting for. It was a SaaS company that tracked eBook download button clicks as conversions. The click did not open the eBook, it simply opened a popup with a form to fill in. The link was delivered via emails on submission. First, a click on the button wasn’t the goal being completed. Second, people needed to fill in the form with valid email IDs to get the link, which about 36% filled with wrong/dummy emails. Only about 38% of total clicks translated into actual eBooks being delivered.
Even worse, all campaigns were using account default conversion goals for optimization and this conversion too was set as primary. The AI saw volume and optimised for it. The numbers shot up. Actual qualified leads went down. The budget kept flowing, and nobody noticed until the pipeline dried up.
The problem is, your advertising system doesn’t know the difference between a valuable conversion and a meaningless click. It only knows what you tell it to optimise for.
How to check: Open each active conversion action in your account. Ask yourself one question: Does this event represent real business value? If the answer is “sometimes” or “not really,” it should either be a micro-conversion (secondary, not primary) or removed entirely.
3. Broken or Delayed Conversion Signals
Timing matters more than most people realise.
It’s important to note that Smart Bidding makes bid decisions in real time, at the moment of the auction. If your conversion signal arrives hours or days later, the algorithm can’t connect it to the right auction. And, the learning loop breaks.
GA4-imported conversions typically have a 12–24 hour processing delay before they appear in Google Ads. For high-frequency campaigns, that delay can significantly weaken the bidding signal.
On top of that, iOS and Safari have been aggressively blocking gclid parameters, the Google’s way of linking a click to a conversion. If your URL structure or redirect chain strips the gclid before the conversion fires, that conversion becomes invisible to the algorithm.
How to check: Use Google Tag Assistant to verify your conversion tags fire immediately on the right page, with the gclid intact. Also check: are you importing conversions from GA4, or tracking them directly in Google Ads? Direct tracking is faster and more reliable for bidding purposes. I personally prefer it this way, almost always.
4. No Enhanced Conversions or Server-Side Tracking
This ties back to the privacy wall I mentioned earlier.
Standard cookie-based tracking is losing ground fast. Browser restrictions, ad blockers, and user consent choices are all eating into your conversion data. If you’re relying only on client-side tags, you’re already missing a significant chunk of your actual conversions.
Enhanced Conversions helps bridge this gap. It uses hashed first-party data, like email addresses, to match conversions back to Google accounts, even when cookies are blocked. The data is encrypted using SHA-256 before it ever leaves your site. So yes, it stays privacy-compliant.
Server-side tracking takes this a step further. Instead of relying on the user’s browser to send conversion signals, your server sends them directly to Google. No cookie blockers and Safari ITP issues. No browser restrictions. It makes those third-party cookies, the first-party. Just clean, reliable data.
For advertisers spending $10,000 or more per month on ads, the improvement in data quality from server-side tracking typically pays for itself within the first month.
How to check: Go to Tools → Conversions in your Google Ads account. Check whether Enhanced Conversions is enabled for your web conversions. If it’s not, that’s your first priority. If you’re in markets with strict privacy regulations, server-side tracking should be next on the list.
5. Conflicting Attribution Models Across Campaigns
This is the sneaky one. It doesn’t look broken on the surface.
You might have one campaign using Data-Driven Attribution. Another importing from GA4 with a different attribution window. A third still defaulting to Last Click somewhere in the setup.
Each model tells a different story about which touchpoint deserves credit for a conversion. When the AI sees contradictory signals across campaigns, it gets confused. It can’t build a consistent picture of what’s actually working.
How to check: Go through each conversion action and note the attribution model being used. If they don’t match, or if you’re importing from GA4 while also tracking directly in Ads, you have a conflict. Data-Driven Attribution is Google’s recommended default for 2026. Standardise on it unless you have a very specific reason not to.
The Conversion Tracking Audit: Do It Now
Alright. You’ve identified the problems. Now let’s fix them.
This is a step-by-step audit you can run on your own account today. No paid tools required beyond what Google already gives you.
Step 1: Inventory Every Conversion Action
Go to Google Ads → Goals → Conversions.
List every active conversion action. For each one, note:
- What event does it track? (Purchase, lead, call, button click?)
- Is it set as a Primary or Secondary conversion?
- When did it last fire?
- What’s the repeat rate?
If something hasn’t fired in 30+ days, it’s either broken or irrelevant. Either way, it needs attention.
Step 2: Verify Tags Are Firing Correctly
Download and install Google Tag Assistant.
Visit the pages where your conversions should fire, thank you pages, confirmation pages, download pages. Trigger the conversion action manually (fill in a test form, for example).
Check two things:
- Does the conversion tag fire at all?
- Does it fire only once?
If it fires multiple times, or doesn’t fire at all, you’ve found your problem.
Step 3: Cross-Reference with Your Backend
This is the step most people skip. Don’t skip it.
Pull up your actual business data for the last 30 days. How many real purchases, real leads, or real sign-ups did you get?
Now compare that number to what Google Ads is reporting.
If Google shows significantly more conversions than your backend, you’re over-counting. If it shows significantly fewer, you’re losing data somewhere in the tracking chain.
The gap is your problem. And it’s usually bigger than you think. Do not delay on this, take correct measure right away.
Step 4: Check for Duplicate Tags
Look at your website source code or GTM setup. Are you running both a manual Google Tag and a GTM-based tag? Are you importing the same conversion from GA4 and tracking it directly?
Any duplication here is inflating your numbers and misleading Google Ads system.
Step 5: Confirm Enhanced Conversions Are Active
Go to Goals → Conversions in Google Ads. Open a web conversion action. Check the “Enhanced conversions” section.
If it’s not set up, enable it. Make sure you’re collecting and sending at least one piece of first-party customer data, email or phone number, at the point of conversion.
Step 6: Audit Your Consent Mode Setup (If Applicable)
If you run ads in markets with privacy regulations, EU, EEA, UK, or anywhere else with cookie consent requirements, this step is non-negotiable.
Go to Tools → Data Manager → Consent Settings in Google Ads.
Verify the options you have selected.
Make sure you’re using Advanced Consent Mode, not Basic. Basic blocks all Google tags when consent is denied. Advanced allows cookieless pings, anonymised signals Google can use for conversion modelling.
Without this, you’re losing data from the majority of your consented-region traffic.
Step 7: Standardise Your Attribution Model
Open every conversion action. Check the attribution model.
Pick one. Data-Driven Attribution is the right choice for the vast majority of accounts in 2026. Apply it consistently across all conversions.
If you’re importing from GA4, be aware of the processing delay. For campaigns that rely on real-time bidding signals, direct Google Ads tracking will always be faster and more reliable. And, it’s okay to have both the setup in place. But make sure to keep Google Ads one as the Primary.
After You Fix It
Here’s where people make another mistake.
They fix the tracking. Great. But then they expect performance to immediately improve, or worse, they make other major campaign changes at the same time. Stop doing that, it’s muddying the water.
Once your tracking is clean, let the AI re-learn on its own.
Smart Bidding needs time to build a new model with the corrected data. For accounts generating 50 or more conversions per week, this typically takes 7–14 days. For lower-volume accounts, it could take longer.
During this period, resist the urge to change budgets, targets, or campaign structure. Every modification resets the learning clock. Let the clean data do the work.
If you’re running Performance Max alongside this, your ROAS structure becomes even more important once the data is accurate. I covered how to segment your campaigns by ROAS tier in the ROAS Segments Framework post, it’s worth reading with fresh eyes once your tracking is solid. The framework only works as intended when the conversion data feeding it is trustworthy.
The Bigger Picture: This Is Only Going to Get More Important
I want to leave you with some context on where things are heading. The tracking infrastructure you build today will determine how competitive you are over the next 12–24 months.
Privacy regulations aren’t going away. They’re expanding. More countries, more enforcement, more cookie consent requirements. The advertisers who have already built server-side tracking and Enhanced Conversions into their stack will have a compounding advantage. The client-side tracking will continue to erode.
AI Max is already here. As Google’s AI campaigns become more autonomous, the quality of your data infrastructure becomes the single biggest differentiator between advertisers who scale and those who spin their wheels.
The performance gap is already measurable. Accounts with clean, consistent conversion data are seeing CPA reductions, while those with inconsistent tracking are watching their CPAs climb.
The difference is not small. It’s the difference between a profitable campaign and a bleeding one.
The Bottom Line
The algorithm isn’t the problem.
The data you’re feeding it is.
Most advertisers set up conversion tracking once when they launch their first campaign and never revisit it. But the landscape has shifted dramatically since then. Privacy changes, platform updates, and the rise of AI-driven campaigns have made accurate tracking the single most critical factor in Google Ads performance.
Run the audit. Fix what’s broken. Then let the AI do what it’s actually built to do.
If you’d rather not do this yourself, or you want a second pair of eyes on your setup, I offer a Website Tracking Audit which you can order or discuss with me about.
Or start with the Free Google Ads Audit Template if you want to tackle it on your own.
