Cleaning up false conversions in Google Ads: spotting double counts and ghost conversions

Smart Bidding is only as smart as the data you feed it. Google’s bid model sees no difference between a genuine request from a qualified prospect and a form that a bot spammed full in three seconds. Both come in as a conversion, both get the same weight, and both steer your bids. If a measurable share of your conversions is false, your entire campaign optimizes toward noise. You then pay more and more for worse and worse traffic, while your dashboard turns green.

This article is part of our overview of what SEA is and the acquisition layer of your growth engine. It is about data hygiene: how to spot and clean up false conversions in Google Ads, double counts, bot leads and autoclicks included, so your bid model learns on real pipeline and not on ghost conversions.

Why false conversions in Google Ads poison your bid model

Smart Bidding is a learning system. It gathers signals, predicts the likelihood that a click leads to a conversion, and adjusts your bid accordingly. That only works if your conversion data is clean. Feed the model polluted data and it draws the wrong conclusions.

Suppose a certain search term, region or audience attracts a lot of bot forms. To the bid model, that looks like a well-performing segment: high conversion volumes, low cost per conversion. So the model bids more there. You attract more of the same bad traffic, your conversion volume rises, and you think things are going well. In reality you have trained a bid model to shift budget toward noise.

The insidious part is that false conversions behave like growth. You do not see that the numbers are polluted, because they point the right way. Only when you look deeper, at the number of leads your sales team can actually follow up, does the floor drop out. That is why conversion hygiene is not a technical afterthought but the foundation of reliable bidding.

The three big sources of false conversions

False conversions almost always come from one of three angles. It pays to look at them separately, because each calls for a different fix.

Double counts from duplicate tags

The most common source is also the dullest: you count the same conversion twice. This happens as soon as the same conversion is sent along two paths. Think of a conversion tag hard-coded on the thank-you page that also fires via Google Tag Manager. Or a conversion that comes in both directly in Google Ads and via an imported Google Analytics goal.

A second classic is the thank-you page that people can reload or revisit. If your conversion tag fires on page load without protection, every refresh counts as a new conversion. The same goes for a thank-you page that ends up in navigation or in a bookmark.

Spotting it is simple once you know where to look: a conversion action with a counting setting on “every” instead of “one” per click, or a striking gap between the number of conversions in Google Ads and the number of real requests in your CRM. Always start your audit here, because double counts are the fastest to fix and deliver the biggest effect right away.

Bot leads and ghost forms

The second source is forms no human filled in. Bots crawl the web, find your form and submit it with random or fake data. Sometimes it is malicious, often it is just automated traffic that clicks everything it comes across.

You recognize these ghost conversions by patterns: fill times of a few seconds, impossible field combinations, email addresses that make no sense, or spikes that do not track with your traffic. A single bot can submit dozens of forms in a day, and your bid model swallows them all as real leads.

Autoclicks and invalid click traffic

The third source sits on the click side. Automated click traffic, click farms and certain kinds of bot activity generate clicks and sometimes even follow-up actions that count as a conversion. Google filters out part of the obviously invalid traffic itself, but not all of it. What slips through lands in your conversion data and steers your bids along with it.

Clean up at the source instead of correcting after the fact

The temptation is strong to write off polluted conversions afterward in a report. That fixes your reporting, but not your bid model. Smart Bidding learns on the raw data it receives, not on the cleaned-up version in your spreadsheet. So you have to filter at the source.

Start with your tag setup. Map which conversion action comes in via which path and remove every duplicate route. One source of truth per conversion. Set the counting setting to “one” per click for lead forms, because one visitor making one request is one lead, no matter how often they reload the thank-you page.

Then tackle the forms. An invisible validation field catches a large share of bot submissions before they even reach your conversion tag. A minimum fill time filters out forms submitted in a few seconds. And server-side validation ensures that only forms with valid, complete data count as a conversion.

Where possible, build in a meaningful delay or qualification step. Instead of counting every raw form submission as a conversion, you can count only once a lead passes a first qualification. That keeps the volume your bid model learns on close to what sales can actually use.

Finally, exclude known bot and data-center traffic at the campaign level and keep an eye on your invalid-clicks report. That does not catch everything, but it keeps the coarsest pollution out of your bid signal. This whole pattern of clean measurement starts at your setup; for that, also read how to measure conversions after going live and how consent mode v2 affects your measurability.

Measure what counts: from form submission to pipeline

The root cause of polluted bid steering is that you measure the wrong thing. Whoever optimizes on raw form submissions optimizes on a number that bots and double counts can easily inflate. Whoever optimizes on qualified pipeline optimizes on something noise cannot fake.

That is exactly the logic behind offline conversion import and lead-to-deal attribution. Instead of rewarding Google for every request, you send back which leads actually became qualified and which ones eventually turned into revenue. Your bid model then learns on the outcome that matters, and false conversions automatically weigh less because they never become a real deal. This is the core of paid that buys pipeline instead of clicks: the fast-acquisition layer of one orchestrated growth engine, with measurement that runs all the way into your CRM.

That requires a measurement chain that reaches further than your conversion tag. If you do not have one, you optimize by definition on a proxy that lets itself be polluted. We set up that chain as a google ads specialist so your bid model learns on real deals, not on the volume of forms. If you want to understand how these numbers connect, also read how to build a conversion funnel in Google Analytics.

Make conversion hygiene a fixed rhythm

Conversion hygiene is not a one-off cleanup. Tags shift with every website change, bots find new forms, and a new thank-you page can introduce a double count out of nowhere. So build in a fixed moment to check your conversion actions: does the volume match your CRM, are there strange spikes, are the counting settings still correct.

A bid model that learns on clean data buys pipeline. A bid model that learns on ghost conversions buys noise, and does so ever more efficiently. The difference is not in your budget but in the quality of the signal you send.

Want to know what share of your conversions is polluting your bid model and how to connect your measurement to real deals? Get in touch and we will look at your conversion setup together.