October 7, 2026 - FouAnalytics and fraud0 announced a partnership to bring automated Google Ads refunds to large advertisers. The headline sounds like a new AI tool. The plumbing is something older and more specific: a one-third commission on money Google's own filters missed, collected by a Munich startup that got Google-certified to do it.
The fee structure first
fraud0's pricing page lists subscription tiers starting at 40 euros per month on an annual plan - or 50 euros per month billed monthly. That's the access fee. The revenue engine is the success fee: fraud0 takes one-third of every dollar recovered. The advertiser keeps the rest.
The partnership announcement cites one case study. Pharmaceutical company Uriach recovered $28,000 in Google Ads spend over three months. That means fraud0's cut on that account was approximately $9,300 - from a company that presumably pays a pittance to Google by enterprise standards. The scale of the economics only starts to matter when the machine runs against larger budgets.
What the machine actually does
Here's the gap it sits inside. Google spends enormous resources filtering invalid traffic - bots, click farms, automated abuse. But Google's own internal research and third-party audits suggest the filters don't catch everything. One industry analysis puts the miss rate at around 7.5% of invalid clicks. When Google's filters miss something, the advertiser can file a manual refund request.
The manual process is structurally hostile to advertisers. You need specific technical evidence - fraudulent IP addresses, server logs, Google Click Identifiers. The claims are limited to the last 60 days. Investigations take five to ten business days, and manual requests have an approximate 10% approval rate. Of the claims that get approved, Google typically grants only 30% to 50% of the requested amount. And refunds come as account credits, not cash - money you can only use for more Google Ads.
fraud0 automates this bottleneck. As a Google-certified Click Tracker, fraud0 works with Google's Click Quality team to investigate suspicious activity. The company's JavaScript tag instruments the advertiser's website, detects invalid traffic, builds the evidence package, and submits the claim. One fraud0 customer reported receiving 30x more refunds after activating the tool. Another said refunds went up 19x. The product claims to recover up to 10% of a client's total ad spend.
The classification question
The official description is "AI for ad refunds." In practice, this is closer to a claims-processing service: a way to mine a refund channel that Google built into its own billing system but made so painful to use that most advertisers never bother.
The Google certification is the key classification boundary. Without it, fraud0 is just another analytics dashboard telling advertisers their traffic looks suspicious - a useful observation but not a refund mechanism. With it, fraud0 works with Google's Click Quality team to investigate and submit evidence, while Google retains control over refund decisions. The certification is what separates a detection tool from a recovery service.
And here's the structural irony: fraud0 is certified by the same company it's auditing for billing errors. The relationship is symbiotic. Google gets a partner that helps clean up the traffic it didn't catch - reducing advertiser churn and complaint volume. fraud0 gets a working relationship with the refund process. The advertiser gets money back that would have taken weeks and probably never materialized.
Nobody is being defrauded. But the incentive alignment is unusual for a vendor that charges its clients for finding errors in their bill from a platform the vendor is certified by.
FouAnalytics' role
FouAnalytics, created by Dr. Augustine Fou, is the credibility and distribution layer. Fou has been publicly skeptical of Google, legacy verification vendors, and the broader programmatic advertising ecosystem for years - a 25-year track of calling out click fraud, inflated CTRs, and attribution manipulation. FouAnalytics itself is an enterprise SaaS platform that charges a flat $2 million annual subscription for its Unlimited tier, targeting the world's largest advertisers and agencies. Its client list includes Microsoft, Beiersdorf, and Georgia Pacific, plus over 10,000 smaller businesses.
The partnership is a logical fit. FouAnalytics has the enterprise relationships; fraud0 has the refund automation. FouAnalytics sells governance for digital ad spending - seeing exactly where your impressions go. fraud0 adds the recovery piece: when the governance finds something wrong, the AI agent actually files the claim.
The Google problem
This is where the plumbing matters for anyone who holds Alphabet stock.
Google's advertising arm generated roughly $295 billion in revenue in fiscal year 2025, with search advertising accounting for the bulk of it. In the second quarter of 2026 alone, search and other ads brought in $63.3 billion. The advertising ecosystem these companies orbit is enormous.
The structural question is whether the refund channel is a permanent feature of Google's business model or a temporary gap. fraud0's entire revenue model depends on two conditions: Google's filters remain imperfect enough that there's money to recover, and Google continues to honor the Click Tracker certification and the refund policy. Either of those could change.
Google has the unilateral power to revoke the certification, tighten the refund criteria, or build its own native refund automation - which would eliminate the need for third-party intermediaries. The company has also shown no interest in admitting its filters miss a material fraction of invalid clicks. That's not a competitive blind spot; it's a brand and trust calculation.
On the other hand, building better internal detection has a cost. Outsourcing the refund process to a certified partner who takes the operational burden - and who's incentivized to find recoverable spend because it pays them - is a rational outsourcing decision. The one-third success fee also means Google's actual refund exposure is lower than the headline recovery numbers suggest. If a claim yields $100 in credited recoveries that pass Google's independent verification, fraud0's one-third commission comes to roughly $33 and the advertiser keeps about $67. But the advertiser would never have recovered the full $100 on their own - maybe $15 through a manual claim that had a 10% chance of approval and a 30-50% grant rate. So Google may end up issuing more in credits than before, but it's paying less in customer goodwill.
The market context
The ad fraud detection market was valued at approximately $2.48 billion in 2025, with projections of $15.3 billion by 2034. Digital ad fraud losses are estimated to exceed $100 billion globally in 2026. fraud0 says it has already audited more than $2 billion in digital advertising spend.
Both companies are private. FouAnalytics doesn't disclose revenue. fraud0 raised $6.55 million in total funding, with its most recent round in 2023 led by Signals Venture Capital. The company employs roughly 27 people and is headquartered in Munich.
What the machine tells you
This is not a story about AI being new. It's a story about someone building a service around a gap in a platform's billing policy, getting certified to access it, and taking a third of what comes through.
The economics are straightforward for the participants. The advertiser gets money they wouldn't have recovered. fraud0 earns a commission that scales with recovery volume. FouAnalytics gets a distribution partnership that deepens its value proposition. Google gets reduced customer friction and a lower-cost refund pipeline.
The uncertainty sits at the classification boundary: how long does the Google certification last, and what happens if Google closes the gap it built? As long as Google's filters are imperfect and the refund channel exists, the machine runs. The question isn't whether fraud0 is clever - it is. The question is whether it's a durable business or a licensed workaround.
The refund channel belongs to Google. fraud0 is just the plumber.