What you'll learn in this article
- The end-to-end test I run every time, and why it costs a click on purpose
- What Tag Assistant proves, what it doesn't, and the gap where most setups fail
- How to test if google ads conversion tracking is working when the number in the column looks perfectly normal
- The silent breakages I keep finding months late, and the two checks that catch them early
Nobody asks how to test google ads conversion tracking when conversions are zero. Zero is loud. You notice zero on a Monday morning and someone fixes it by lunch. The accounts that hurt me are the ones where the column was full the whole time and every number in it was wrong.
I inherited an account once where the thank-you page tag had been duplicated in a theme update. Conversions doubled. Nobody complained, because the CPA halved and everyone assumed the new creative was working. Smart Bidding believed it too, and spent six weeks chasing a phantom. The tracking never showed an error status. It was firing beautifully. It was just firing twice.
I'm assuming here that the setup already exists the mechanics of building it live in my pillar on how Google Ads conversion tracking works. This piece is about the part that comes after: proving it.
So the question I actually care about isn't whether the tag exists. It's whether a real click, from a real ad, produces exactly one conversion, attributed to the right action, with the right value. That's a different test, and it's the one almost nobody runs.
How to test Google Ads conversion tracking: click your own ad and pay for it
Every shortcut around this test is a shortcut around the thing that breaks. Google's own troubleshooting guidance says it plainly: a test conversion performed in Tag Manager preview mode, or by navigating straight to your conversion page without an ad click, may never show up in the account at all Google Ads Help on verifying your tag. That isn't a footnote. It's the entire reason people think their tracking is fine when it isn't.
The mechanism explains why. The click writes an identifier into the browser. The tag, later, reads it back. If you skip the click, the tag fires into a void with nothing to attribute to, and the account stays quiet. So the test has to start where a customer starts.
The sequence, in order
Search one of your live keywords on Google. Click your own ad and accept that you're paying for it that click is the price of a real answer, and it's cheaper than a quarter of bad data. Land on the site and check the URL still carries the gclid parameter. Give consent exactly as a normal visitor would. Then complete the conversion for real: submit the form, finish the checkout, whatever the action is. Not a dry run, the actual path.
The step everyone skips
Follow the gclid through every hop. Landing page to cart, cart to a payment processor, processor back to the thank-you page. That parameter dies somewhere in a startling number of accounts, and it almost always dies at a redirect somebody added for an unrelated reason. If it's present on the landing page and gone on the confirmation page, you've found your bug and you didn't need a single tool to find it.
Then wait, and count
Reporting isn't instant, and expecting it to be causes half the false alarms I see. Give it a few hours, then look at the conversion action, not the campaign total. You're checking three things: it recorded, it recorded once, and the value matches what the transaction was actually worth. One conversion, right action, right number. Anything else is a finding.
Tag Assistant: what it proves and where it stops
Tag Assistant is the right first tool and the wrong last one. Google surfaces it directly from the conversions table when an action reads "Unverified," "Tag inactive," or "Needs attention," and it will walk you through triggering the action and reporting back what fired. It's free, it's fast, and it doesn't cost a click. Use it.
What it tells you is narrow but valuable: the tag exists, it executes on the page you expected, and the request going out carries a conversion ID and label. That last part matters more than people realise I've seen tags copied between two accounts during a migration, firing perfectly, into someone else's conversion action. Green light. Wrong destination. Only checking the ID and label against the interface catches that.
Where the tool goes quiet
Here's the inference I'd draw from every audit I've done: Tag Assistant validates the tag, not the attribution. It cannot tell you the gclid survived the checkout, because you didn't come from an ad. It cannot tell you the tag fires twice under conditions you didn't reproduce. It cannot tell you the value passed is a string with a currency symbol in it instead of a number Google documents that exact failure, and it silently kills the conversion value while the conversion itself sails through.
None of that makes the tool useless. It means Tag Assistant answers half the question, and anyone learning how to test google ads conversion tracking properly needs to know which half.
And it cannot tell you about consent. If ad storage and ad user data stay denied for your real traffic, the tag fires and the conversion never attributes. Test it the way a visitor experiences it, banner and all, or you're testing a path nobody walks.
The layer underneath
When the tag looks right and conversions still don't arrive, the next place I go is the network request itself: Chrome DevTools, Network tab, find the googleadservices call, read the ID and the label off the URL. It's unglamorous and it's definitive. If your implementation lives in a container, the same discipline applies one level up the container publishes, the tag conditions change, and the reasoning in my breakdown of deploying conversion tags through Tag Manager is where I'd start reading before touching anything.
How to test if Google Ads conversion tracking is working when it looks fine
This is the part of how to test if google ads conversion tracking is working that no checklist covers, because there's no error state to look for. These are the ones that cost money, because nothing alerts on them. Every symptom below is something I've found late, in an account that had been running on it for months.
The CPA that improved for no reason
A conversion rate that steps up overnight, with no campaign change behind it, is a duplicate tag until proven otherwise. Real improvements arrive as a slope. Tracking bugs arrive as a step. Whenever I see the step, I go look at what the site deployed that week, and I'm usually right.
The gap that nobody reconciles
Google's number and the CRM's number will never match, and part of that is legitimate modeled conversions exist, attribution dates to the click rather than the sale, and time zones drift between systems. But "they never match" becomes cover for "they're wrong." I want the gap measured once and explained. An unexplained gap is a finding. An explained one is just reality.
Fractional and impossible counts
Decimals are usually data-driven attribution behaving normally. But a conversion count that exceeds your actual orders isn't attribution, it's double firing. Pull the raw number the business has and compare. This takes ten minutes and I have never regretted it.
The action that quietly went secondary
Tracking works, bidding ignores it. Someone reorganised goals, the action stopped being counted in the Conversions column, and the algorithm has been optimizing toward nothing you care about. Nothing is broken. Everything is wrong. If you're feeding Smart Bidding from these actions, this is the failure with the highest cost and the lowest visibility, because the setup looks configured and the data looks populated.
The seasonal ghost
Conversions drop, and the drop lands on a week where something plausible happened a holiday, a budget change, a competitor. The story fits, so nobody tests. Then you find the container was republished the day before. Test first, explain second. The explanation is always available and it's frequently wrong.
When I re-test, and how much it's worth
Knowing how to test google ads conversion tracking is half of it. Knowing when to do it again is the half that saves accounts. Once, at setup, isn't a policy. It's a formality. The tracking didn't break on the day you built it, it broke on the day someone else touched the site.
The triggers
Any site deployment, theme update, or plugin change. Any checkout or form change, however cosmetic. Any container publish. Any consent banner change, which is the one people never connect to conversions and which breaks them completely. Any new conversion action. And every time I take over an account from someone else, before I read a single performance report the full audit logic is in my write-up on what I check first in an inherited account, and tracking verification is step one for a reason.
The cheap alarm
An automated rule that emails me when conversions fall more than half against the prior week costs nothing and catches the loud failures within a day. It won't catch a double-firing tag, because that failure moves the number up. But the asymmetry is fine: automate what's detectable, test manually for what isn't.
What I actually write down
One line per conversion action: where it fires, which ID and label it carries, what a normal week looks like, and the date I last tested it end to end. That last field is the whole point. When someone asks whether the data can be trusted, "I tested it on the 3rd and it recorded once at the right value" is an answer. "It says Active" is not.
The honest summary: how to test if google ads conversion tracking is working comes down to one refusal refusing to accept a status badge as evidence. Click your own ad, follow the gclid, complete the real action, count what arrives. Everything else is checking that the machine is plugged in, which is worth doing and proves nothing about what it's measuring.