The Unspoken Truth: Why Importing GA4 Conversions to Google Ads Is a Data Minefield

In April 2026, Google quietly made GA4 the default conversion source for a lot of Google Ads accounts. No big announcement. A lot of advertisers woke up to conversion numbers that had shifted and did not know why.

I have audited Google Ads accounts for years, and the GA4-to-Google-Ads import is the single setup I find broken most often. Not "slightly off." Structurally broken, in a way that quietly poisons Smart Bidding.

This is not a "how to import GA4 conversions" post. The import is easy. This is a post about why the data you are importing is already wrong before it ever reaches Google Ads, and why feeding it into Smart Bidding makes the problem worse, not better.

Here is the lie buried in the official guidance: that importing GA4 conversions gives you a richer, more attribution-aware signal. It can. It also routinely sends a double-degraded number into Google's bidding brain.

The fix is not a checkbox in the conversions menu. It is architectural, and that is where DataCops comes in.

See the Google Conversion API layer, and for the underlying signal problem read why your Google Ads aren't converting.

Quick stuff people keep asking

Should I import GA4 conversions to Google Ads or use native tags? For most accounts, the native Google Ads tag, ideally server-side, is the more reliable bidding signal. GA4 import is fine for cross-channel reporting. The mistake is using a reporting tool as your bidding source.

Why are my GA4 conversions different from Google Ads conversions? Different attribution models, different conversion windows, different counting rules, and GA4 applies its own consent and modeling layer. They will never match exactly. If they match perfectly, something is misconfigured.

What causes duplicate conversions in Google Ads? Running the native Google Ads tag and a GA4 imported conversion for the same action at the same time. Both fire, both count, your numbers inflate. Pick one source per conversion action.

How does the April 2026 GA4 update affect conversion tracking? Google shifted many accounts to GA4 as the default conversion source. If you did not audit your setup after that, you may be bidding on GA4 imported data without having chosen to.

What happens when GA4 data-driven attribution falls back to last-click? Data-driven attribution needs roughly 400 conversions in 30 days per conversion action to run. Below that, GA4 falls back to last-click. Your attribution model silently changes, and so does which clicks get credit, without any warning in the UI.

How do I fix inflated conversion numbers in Google Ads? Find duplicate conversion actions, confirm one source per action, check that you are not double-counting native plus imported. Then ask the harder question: is the remaining number itself trustworthy.

Is it better to use GA4 or Google Ads native conversion tracking? For bidding, native, server-side. For reporting and cross-channel context, GA4.

They serve different jobs. Trouble starts when you let GA4's reporting number drive bids.

How do I audit my Google Ads conversion tracking setup? List every conversion action, its source, its attribution model, and its 30-day volume. Flag anything below 400 conversions, anything with two sources, and anything where the source is GA4 but you never decided that.

The minefield is a stacked signal-degradation problem

The reason this topic deserves a real article is that the GA4 import does not have one problem. It has a stack of them, and they compound.

Layer one. Before GA4 records anything, consent mode and ad blockers have already eaten a slice of events.

On a typical site, 20 to 40% of conversion events never make it into GA4 cleanly. Some get modeled back in by Google's estimation, some just vanish.

Layer two. What does land in GA4 includes bot traffic.

Of the events reaching a typical analytics endpoint, 24 to 31% are non-human. GA4's bot filtering catches the obvious known crawlers and misses the rest, especially the AI agents that have exploded across the web.

Layer three. GA4 then applies an attribution model.

If a conversion action sits under that 400-conversions-in-30-days threshold, data-driven attribution quietly falls back to last-click. So the credit assignment changes based on volume, invisibly.

Layer four, the expensive one. You import that number into Google Ads and point Smart Bidding at it. Now Google's bidding algorithm is learning from a signal that is missing 20 to 40% of real conversions, padded with bot events, and attributed by a model that may have silently switched on you.

Smart Bidding does exactly what it is told. It optimizes hard toward the picture it is given.

Feed it conversions inflated by bots, and it learns the patterns of bot traffic look like success. It bids up to find more of it.

“

Garbage in, and the algorithm does not just store the garbage, it goes hunting for more.

Here is a concrete picture of how bad the bot half gets. A signup product ran a honeypot, a hidden registration path no real person would ever reach.

It collected 3,000 signups. 77% were fraudulent. 650 of those accounts came from a single device fingerprint. One machine wearing 650 faces.

If that kind of traffic flows through your analytics into your conversion feed, Smart Bidding treats one bot farm as 650 wins and spends to clone it.

That is the minefield. Not duplicate conversions, that is the beginner trap. The real damage is a confidently wrong number teaching Google's algorithm to chase the wrong traffic.

What a clean conversion signal actually requires

Fixing duplicates is hygiene. It does not touch the deeper problem. A genuinely trustworthy conversion signal needs three things, and a reporting-tool import gives you none of them.

It needs first-party collection. Events captured from your own infrastructure, on your own subdomain, instead of relying on a client-side tag that browsers and blockers keep breaking. This recovers the real conversions GA4 was losing.

It needs bot filtering before the signal is sent. Non-human events identified and stripped at ingestion, against IP reputation, device fingerprint, and behavior, so the bot share never enters the feed Google bids on.

It needs two separated data tiers. Anonymous, aggregate analytics that flow unconditionally because anonymous measurement is always legal.

And identifiable conversion data, the stuff Google uses to match and optimize, governed by consent. Separated at the source, not blended and untangled later.

This is the architecture DataCops is built for. First-party collection on your own subdomain, bot filtering at ingestion against a 361.8 billion-plus IP database, and clean Conversions API delivery into Google Ads, Meta, TikTok, and LinkedIn. You stop importing a reporting estimate and start sending Google a filtered, first-party signal.

The honest limitation: DataCops is a newer brand than GA4 itself, and SOC 2 Type II is in progress. If your procurement requires that certification right now, factor that in. The trade is a far cleaner bidding input.

Decision guide

You run Smart Bidding and import GA4 conversions as your source. This is the highest-risk setup. Move bidding onto a server-side native signal and keep GA4 for reporting.

Your conversion action gets under 400 conversions in 30 days. Assume data-driven attribution has fallen back to last-click. Bid and read results with that in mind.

Your numbers jumped or dropped around April 2026. Audit immediately. Google likely switched your default conversion source and you are bidding on a source you did not pick.

You see duplicate conversions. Quick fix first: one source per conversion action. Then go deeper on whether the remaining number is bot-clean.

You run paid in the EU. Make sure anonymous analytics and identifiable conversion data are split at the source, so the legal anonymous tier keeps flowing while consent governs the rest.

You cannot tell whether your conversion data is bot-contaminated. That uncertainty is your answer. You cannot optimize a signal you cannot trust. Get filtering in before ingestion.

You are not bidding on conversions, you are bidding on a story about conversions

Here is the mistake almost everyone makes. They treat the conversion number in Google Ads as a fact.

It is not a fact. It is the end of a long chain: consent filtering, ad-blocker loss, bot inflation, an attribution model that may have silently switched, then an import.

Every link bends the number.

Smart Bidding does not know any of that. It treats the story as gospel and spends your budget to produce more of whatever the story rewards. If the story is half-fiction, your bidding is optimizing the fiction.

Importing GA4 conversions is not the sin. Importing them blind, without knowing what got lost, what got faked, and which attribution model was actually running, that is the minefield.

So go look. Pull every conversion action, its source, its attribution model, its 30-day volume.

Which ones are below 400? Which have two sources?

And the real question: of the conversions you are bidding on right now, how many do you actually know are human?