How invalid clicks break Smart Bidding

Google pulls the invalid clicks it catches, and any conversions attached to them, from your campaign metrics before the billing cycle closes. Traffic past that filter is what Smart Bidding still learns from: junk leads Google cannot see after the click, invalid traffic found after the month's metrics are frozen, and remarketing lists that bots have joined.

ClickPatrol classifies each ad visit so Smart Bidding can learn from legitimate conversions only. If you run Target CPA, Target ROAS or one of the Maximize strategies in-house, this guide shows where bad data actually enters the account, how to check for it, and what to change.

Does Smart Bidding learn from invalid clicks?

Mostly not from the ones Google filters. Its automated filters catch most invalid traffic shortly after the interaction, Google's page on invalid traffic says. If the activity is identified before the billing cycle ends, it comes off your bill and out of your campaign reports. Impressions, clicks, interactions and conversions are included.

Those clicks were already taken out, so the Invalid clicks column is not the threat. What is left sits in three places Google's filter does not reach.

Junk conversions Google cannot see

Google's filters only run once an ad has served. Google says plainly that its systems cannot track or control what a visitor does on your site after the click, and that invalid traffic can still reach your website and lead forms.

A form fill from a bot or a paid clicker still lands as a conversion in your tag. Smart Bidding treats it like any other success. From there it starts looking for more visitors like that one. On a cost report this does not show up as waste. For that source, the CPA can even look better than average. That is exactly why the algorithm leans into it.

Invalid traffic found after the month closes

Some invalid traffic takes longer to detect. Google credits that activity on a later invoice. At the end of each month it freezes campaign metrics so they match the invoice. Credits for past invalid activity do not, for that reason, adjust the campaign data you see in Google Ads.

The history your campaigns report still holds those clicks. Google's Help pages do not say whether the bidding model corrects for them separately.

Remarketing lists that bots have joined

According to Google's Smart Bidding overview, remarketing list membership is one of the auction-time signals Smart Bidding uses. Automated visitors on those lists give the campaign a signal that says "this person has been here before" about traffic that will never buy. Keeping bots off your lists is the job of clean remarketing audiences.

Which Smart Bidding strategies are most exposed?

Four Smart Bidding strategies are listed on Google's current Help page. From June 2026 Google is also renaming two of them in the interface. "Maximize conversions with a Target CPA" becomes Target CPA, and "Maximize conversion value with a Target ROAS" becomes Target ROAS. The bidding behaviour stays the same.

The bigger risk usually sits with lead generation accounts. A purchase that needs a completed payment is much harder to fake than a contact form, so the bad data mostly comes in through enquiries, sign-ups and quote requests.

Why am I getting clicks but no conversions?

One possible cause is invalid traffic, and often it is not the first thing to check. Before you blame bots, rule out the boring explanations:

• Conversion tracking broke. A site release, a new form plugin or a changed thank-you page can stop the tag from firing. Run the tag through Tag Assistant.
• Something in the account changed. Google points to budget increases, new keywords, and audience or location edits as common causes of traffic shifts. Those edits show in your Change history.
• Demand moved. More clicks of a different quality can come in from seasonal swings, news events, or a competitor pausing their ads.
• The ad and the landing page have stopped matching. Slow load times, a changed offer or a broken form each lose people before they convert.

For Smart Bidding the reverse pattern is the more dangerous one. Conversions keep arriving while sales cannot reach those leads, and the algorithm is being trained on them.

How do you check whether invalid traffic is reaching your bidding data?

Work through these in order. Google's own reports cover the first three. The last three cover what Google cannot see.

1. Add the Invalid clicks column. Go to Campaigns, click the Columns icon, search for "Invalid clicks" and apply. These are clicks Google already filtered and did not charge you for. A higher number means Google caught more, not that you paid more. What those clicks usually are is set out on our page on invalid traffic.
2. In the Report Editor, open the Invalid Activity Credit Report. Adjusted clicks, cost and conversion rate appear after credits for later-detected activity. The gap between this report and your normal campaign view is activity your standard reports still count.
3. Check your invoice adjustments. In Billing, open Summary and look for "Invalid activity" under Adjustments in the Month card.
4. Match conversions to your CRM, per campaign. Count the leads that turned out fake, unreachable or duplicated. Google has no view of this. That count is what tells you what Smart Bidding is learning from.
5. Compare paid and organic behaviour in GA4. GA4 counts a session as engaged when it lasts longer than 10 seconds, includes a key event or has at least two page views. A campaign whose engaged-session rate sits far below organic deserves a closer look. The reports to use are in our guide to filtering bot traffic in GA4.
6. Ask Google to investigate what it missed. Requests for activity within the last 60 days go through the Click Quality Form. Include your customer ID, the dates, campaign names, IP addresses, GCLIDs and, for lead problems, a spreadsheet with your assessment of each lead.

Keep two limits in mind. Google will not confirm which individual clicks it judged valid or invalid. Doing so would help fraudsters avoid detection. Google notes that third-party tools may flag traffic it had already filtered and kept off your bill.

How do you stop junk conversions from training Smart Bidding?

Blocking clicks is the smaller part of the fix. Deciding which conversions the algorithm is allowed to see is the larger part. Three changes, roughly in order of effort.

Bid on outcomes that a bot cannot produce

Primary conversion actions are what Smart Bidding optimises towards. Secondary actions are for reporting only. They do not move bids. Make a raw form submission primary and the campaign chases form submissions, including the fake ones.

Google's own advice for lead accounts points the same way. Use the Leads objective, build goals around a qualified lead, converted lead, booked appointment or quote request, and set up enhanced conversions so actions are measured accurately. A primary conversion further down the funnel is harder to fake.

Protect the form itself

As website safeguards for lead forms, Google lists reCAPTCHA, server-side validation and double opt-in. Double opt-in means the lead has to click a link in a verification email before it counts. Each safeguard removes a share of automated submissions before they become conversions.

Count conversions on legitimate visits only

A click fraud tool does something a block list cannot: it decides which conversions Smart Bidding gets to learn from. Each ad visit is classified by ClickPatrol, and the result goes into a first-party cookie called cp_class. In Google Tag Manager you can add that value as a trigger condition, so a second conversion action fires only when the visit was classified as legitimate.

In the legitimate visit conversions guide, the setup runs in a set order:

1. Keep your existing conversion as it is and add the legitimate-only action next to it.
2. Set the new action to secondary for a week or two, then compare the two.
3. When you want a campaign to bid on legitimate leads only, make the new action primary for that campaign and set the original to secondary there.
4. Never make both primary in the same campaign, or the same lead counts twice.

The classification has limits. It is valid for 24 hours. It only covers ad traffic. It does not follow a visitor from mobile to desktop. If your consent banner holds the tag back, visitors who decline are never classified. Legitimate also does not mean qualified: if your sales team scores leads, that score is still the stronger signal. More on how this fits into Smart Bidding protection.

What should you expect after cleaning the conversion data?

At first, fewer conversions. A legitimate-only or qualified-lead action reports less than a raw form submission. A target CPA copied from the old action will be too strict and can push you out of auctions. Set the target from what the new action actually reports.

Wait before you judge it. Google recommends measuring Smart Bidding over a period with at least 30 conversions, such as a month or longer, and at least 50 conversions for Target ROAS. A bad week at the start tells you very little.

Change one thing at a time. After an exclusion goes live, ClickPatrol's audience guidance suggests watching cost per lead and conversion rate for two to four weeks before the next change. If the conversion action, the audiences and the target all switch in the same week, you cannot tell which one worked.

Do Google's invalid click credits repair the bidding data?

No. A credit is a billing adjustment on your invoice. It is not a correction to the campaign history. Campaign metrics do not adjust retroactively for past invalid activity, Google says, so the clicks behind a credit stay in your reports.

Credits are also slow and small. When ClickPatrol submits contested clicks to Google for manual review, the investigation typically takes about 4 to 8 weeks. On average roughly 5% of those clicks are eventually approved and credited. Keeping invalid traffic out of the conversion data matters more than the money that comes back, so the order of value runs the other way. For the full picture of blocking, reporting and audiences together, see our overview of click fraud protection.

Written & reviewed by

Abisola Tanzako

Content Manager, ClickPatrol (click fraud & invalid-traffic specialist)

Abisola covers bot traffic, ad fraud and PPC protection, drawing on ClickPatrol platform data from 1,793+ businesses.

Last updated: 19 October 2026

Related Articles

PopAds click fraud: risks for advertisers

2026 Click Fraud Statistics: Costs, trends, and prevention for marketers

Facebook click farm fraud: How to detect and stop click fraud in 2026