Half of Those Leads Are Bots, and Smart Bidding Is Buying More of Them

Most tracking problems are static. Something is misconfigured, it reports the wrong number, and it keeps reporting the wrong number until you fix it. Spam leads are different, and worse, because they create a loop. A bot fills in your form, your tag records a conversion because a form was genuinely submitted, the bid strategy concludes that whatever produced that conversion is working, and it goes and buys more of the same. The problem compounds while you watch your cost per lead improve.

Nothing is broken here in the technical sense, which is why this is almost never caught by a tracking audit. The tag fires correctly. The conversion action is healthy. The data is flowing exactly as designed. The input is garbage, and the measurement layer has no concept of lead quality, so it cannot tell you.

Why the platform counts it

A conversion tag answers one question: did the thing happen? A bot reaching your thank you page did the thing. There is no second question about whether the submitter was a person, whether the phone number was real, or whether anybody ever replied. Google’s invalid traffic filtering removes clicks it identifies as illegitimate, which is a different mechanism operating on a different signal, and it is not a lead quality filter.

So the spam flows into the same column as your real leads, and from there into the bid strategy. Google is explicit that Smart Bidding uses conversion data to optimise bids in each auction, which is the entire value proposition and also the mechanism by which this gets expensive. The algorithm is doing its job perfectly on a corrupted objective.

What the loop actually costs

It is worth being precise about the damage, because “some spam leads came in” sounds like a nuisance rather than a budget problem.

• Your reported cost per lead falls. This is the cruellest part. The number everybody is judged on improves, so nobody investigates.
• Budget shifts toward the worst placements. Bot traffic is not evenly distributed. It concentrates in particular placements, geographies, device types and times of day, and those segments start to look like your winners. With a maximise conversions or target CPA strategy, spend follows.
• Your real campaigns get starved. The segments producing genuine enquiries have a worse apparent cost per conversion than the ones producing bots, so they lose budget on the comparison.
• Sales stops trusting marketing. Once a sales team has worked a few weeks of junk, they discount the whole list, and genuine leads sit uncalled. This damage outlasts the fix.

Telling spam from a quiet month

Before you change anything, confirm what you are looking at. Lead volume rising while quality falls has several possible causes, and spam is only one.

1. Look at submission timing. Real enquiries follow business rhythms: weekday daytime in your target market, a dip at weekends. Spam does not care. A flat distribution across all twenty-four hours, or bursts of submissions seconds apart, is the clearest signal available.
2. Read the free text fields. Nothing beats actually looking. Gibberish, mismatched names and email addresses, a message field containing a URL, and the same body text under different names are all obvious once you read fifty submissions in a row.
3. Check the geography against your targeting. If you advertise in three cities and the leads are from everywhere, the leads are not from your ads.
4. Compare the form log to the conversion count. Your form backend, your CRM and the ad platform should roughly agree. Where they disagree tells you whether you have a spam problem or a counting problem, and those have completely different fixes.
5. Segment by campaign and placement.Spam is rarely uniform. If one campaign’s leads are 80 percent junk and another’s are clean, that is diagnostic in itself, and it usually points at display, broad targeting, or a particular network.

One warning: do not assume a sudden lead spike is spam just because the quality feels off. A broken form validation rule, a newly indexed landing page, or a conversion tag that started firing on the wrong trigger can all produce the same shape. Check the submissions themselves before you blame bots.

Stopping it at the form, which is the only real fix

Everything downstream is cleanup. The actual solution is that the submission never becomes a conversion in the first place.

• Validate on the server, not in the browser. Client-side validation is a courtesy to humans. Bots post directly to your endpoint and never execute it.
• Add a honeypot field. A field hidden from users that bots fill in anyway. Cheap, invisible, and it catches a surprising share of automated submissions with zero friction for real people.
• Use a scoring challenge rather than a puzzle. Modern invisible challenges score the interaction without asking the user to do anything, so you are not trading conversion rate for cleanliness. If you must use a visible challenge, measure the conversion rate impact before and after, because it is real.
• Fire the conversion after validation, not on page load. This is the tracking-side fix that most accounts are missing. If the conversion fires when the thank you page loads, anything that reaches that URL counts. If it fires from a server-confirmed, validation-passed submission event, your rejected submissions never enter the data.
• Require a field bots get wrong. A properly formatted phone number for the region you serve, or a dropdown that must be chosen, filters a lot of low-effort automation.

That fourth point is the one worth arguing for with a developer, because it fixes this problem and several others at the same time. A conversion keyed to a validated submission cannot be triggered by a page refresh, cannot fire on a failed submission, and cannot be reached by a crawler.

Cleaning up what already landed

Google Ads lets you adjust conversions after the fact, including retracting one entirely. Retracting removes the conversion completely from your reports, and for website conversions it cannot be restored afterwards, so this is a one-way door.

There is a prerequisite that catches almost everybody, and it is the reason this section is usually academic for lead gen accounts: you need an identifier for the conversion you want to retract. The documented route is the order ID that matches the original conversion, with the click identifier plus conversion time as an alternative in some cases. The conversion action name has to match exactly, including spelling and capitalisation.

Which means: if you are not already passing a unique identifier with every lead, you cannot retract anything. That is an argument for passing a lead ID with your conversions today, before you need it, because the ability to clean up spam retroactively is only available to accounts that planned for it.

The timing rules matter too. Google advises uploading adjustments at least 24 hours after the original conversion, notes that adjustments attempted within one day of the click may not be recorded at all, and says retractions and restatements count toward autobidding for up to 7 days after the conversion was first recorded. Adjustments can be made within 54 days. Practically, a weekly retraction run clears spam inside the window where it still influences bidding; a quarterly cleanup only tidies the report.

The better long-term shape: upload qualified leads

The strongest version of this is to stop treating the form fill as the conversion. Record the submission as a soft signal, and send the platform a conversion only once a human has qualified the lead. The bid strategy then optimises toward leads your client would actually accept, and spam never enters the loop because it never gets qualified.

This needs a CRM you can get data out of and a client willing to keep stages up to date, which is the real constraint. It is covered in more depth in our guide to offline conversion tracking, and it is the single highest-leverage change available to most lead gen accounts.

What to tell the client

Lead with the arithmetic rather than the diagnosis. “Of the 180 leads we reported last month, 74 had unreachable phone numbers, which means we have been optimising toward the wrong half” is a sentence that gets agreement. “We have a spam problem” sounds like an excuse for bad numbers.

Then be straight about what happens next: reported conversions will fall, cost per lead will rise, and both of those are the numbers becoming true rather than getting worse. Pre-announcing that is the difference between a credibility win and a difficult call.

If you want to check the mechanical side of a site first, our free tracking scan reads the public source and the published container in about thirty seconds and shows what fires on what, which is where a conversion wired to a thank you page load rather than a validated submission becomes visible. It cannot read your submissions, so judging lead quality is still the manual review above.

And a form that starts accepting junk usually does so after a change nobody announced. Taglert rechecks every connected account every thirty minutes and emails you when conversion tracking changes, so a conversion action that suddenly triples has somebody looking at it the same day.

Our conversion tracking monitoring page covers how those checks run across every client account at once.