When Are Bots Filling Out Forms? (More Often Than You’d Think)

Back in 2013, the Legend himself, Dr. Augustine Fou, described working with university marketers who were buying leads from online forms. The forms came back complete: a high schooler’s name, address, phone number, even their school mascot. When the university called to follow up, the “student” overwhelmingly claimed that they had never heard of the university and never filled out a form. That happened often enough that these marketers had to routinely trash 50% to 70% of the leads they’d already paid for.

That was 13 years ago, and fraud has only gotten more sophisticated. Fou, an independent ad fraud researcher who’s spent more than a decade auditing digital campaigns, has said in interviews that roughly half of all digital ad traffic is bot-generated. His campaign audits go even further; one case study built on his FouAnalytics platform found that nearly 40% of the ads in one advertiser’s programmatic display campaigns landed on sites and apps flagged as suspicious or outright fraudulent. Nearly half!!

In one campaign audit he broke down, 74% of the ads that won a bid in a low-cost-per-impression campaign never got served to a real page, meaning three out of every four dollars were thrown out the window. Another campaign he reviewed in the same breakdown showed a 33% delta between bids won and ads served, a third of the entire spend gone before an ad was ever even loaded. He’s also run experiments to test whether tight targeting fixes the problem: with a single approved domain and only one ad exchange active, still only 73% of the ads reached the site he’d specified.

Screening For Bots Versus Confirming Humans

Fou’s core argument, made repeatedly on his FouAnalytics Substack, is that most fraud detection tools hand advertisers a single invalid traffic percentage with no detail on how it was measured or which sites caused it. A vendor reporting a clean 3% invalid traffic rate is only telling you what it managed to catch. In a companion piece, he’s made the case that fraud runs way beyond simple bot traffic into stacked ads, ad slot refreshing, pixel stuffing, and forced navigation, categories that a basic bot filter never even looks for. A vendor that can prove that every lead came from a real person offers a MUCH stronger guarantee than one that’s simply never been caught running fake traffic yet.

This is precisely what a verified HQL from LeadSpot is really paying for. A $65 to $95 lead that’s been through custom qualifying questions and a human verification step comes with proof that a real person confirmed there’s an actual buyer behind every lead before it’s delivered.

Agentic AI Scales Whatever Data It’s Given, Good Or Bad

Fou raised a newer concern on a podcast earlier this year: agentic AI is getting handed more autonomy over marketing decisions while the data it’s acting on still carries the same fraud problems that have existed for years. Automation speeds up whatever’s already in the system, fraud included. Applied to lead generation, it’s a strong argument for keeping a human review step in place as more of the qualification process gets automated.

Brands relying on automated lead scoring alone are trusting a system designed to flag the most obvious fakes. Confirming a human filled out the form is a much tougher job, one that still needs a person to do it correctly. A 2025 MarTech article on B2B lead gen fraud names FouAnalytics, the platform Fou created, as one of the tools marketers now install specifically to catch fake form fills, alongside qualifying questions and CAPTCHAs. As more vendors lean on AI to speed up qualification, the distance between “scored as qualified” and “verified by a person” keeps getting wider.

What This Means For How You Budget Lead Gen

If roughly half of ad traffic is bots, if nearly 40% of programmatic placements can land on fraudulent sites, and if even a single-domain campaign only reached its target 73% of the time, a cheap CPL can hide that fraud very nicely. Our research on how AI is changing B2B demand gen shows buyers arrive more skeptical and better informed than ever… before they talk to a rep, which makes a fake lead even more expensive since your team spends time chasing them.

Fou’s work makes a case marketers don’t hear often enough: a cheap lead and an expensive lead can look identical on a spreadsheet… an actual person checking every gated-content submission is what tells them apart.