Meta and fraudulent ads: how incentives, automation and enforcement gaps collide

Reuters reports that internal documents show Meta earning very significant revenue from fraudulent ads running on Facebook and Instagram. That is not a surprise to anyone who has watched the platform from the outside, but the leak matters for one simple reason. It moves the debate from “scams are a side effect” to “scams are a visible revenue stream inside Meta’s own dashboards”. Once that is true, the real question is how hard Meta is willing to push against a line of business that makes money but damages user trust and attracts regulators.

How Meta’s ad machine creates space for scams

Meta has built one of the most efficient self service ad systems on the internet. Anyone can open an account, upload creatives, define some targeting and start spending. Automated review and machine learning models sit in front of the ad auction to enforce policy at scale. That is exactly the kind of system that scammers want to sit on top of.

Fraudulent advertisers get three advantages immediately.

Low setup cost – New ad accounts and pages are cheap and disposable. Burning an identity is just a cost of doing business.
Fast iteration – They can test lots of creatives and landing pages, see what passes review and what converts, then double down on the survivors.
Precise targeting – Meta’s tools for legitimate advertisers also let scammers find people interested in crypto, finance, side income, health products or other vulnerable niches.

Meta’s policy and review systems are designed to catch clear policy violations: obvious fake brands, crude phishing pages, illegal products. The problem space in between is much harder. Investment schemes that look almost legitimate, fake endorsements that mimic real content, deepfake “celebrity” pitches for financial products – all of that can be tuned to sit just inside the edge of what generic filters will flag.

The incentive problem inside Meta

The Reuters reporting suggests that Meta does not just know scam ads exist. It also knows how much revenue they generate. When that spend is large enough to show up as a line item in internal documents, the company’s incentive structure becomes as important as its technical controls.

In broad terms, Meta has three levers.

Turn the dial up on blocking – Aggressively block whole categories of high risk ads. That cuts scam exposure but also kills some genuine advertisers and takes a hit on revenue.
Rely on reactive clean up – Let the funnel stay relatively open, then remove campaigns after user reports, regulator complaints or media coverage. That preserves revenue but shifts more risk onto users.
Introduce heavy vetting for specific sectors – Build high touch onboarding for financial services, health claims and similar categories. That improves quality but is expensive and hard to roll out consistently in every country.

If internal dashboards treat all ad spend as equal, managers will always feel pressure to keep the funnel as open as possible. That is the core issue regulators pay attention to. They are less interested in whether Meta has a policy on paper, and more interested in whether the company is prepared to sacrifice meaningful ad revenue to enforce that policy in practice.

Regulators were already losing patience

Before these documents surfaced, regulators had already been pointing at Meta as a key vector for scams.

Financial regulators in the UK, EU and elsewhere have warned that a high proportion of investment scams and fake financial promotions now start life as online ads.
Consumer protection agencies have highlighted fake celebrity endorsements, deepfake videos promoting unreal returns, and cloned bank or broker sites driven by paid placement on social platforms.
The EU’s Digital Services Act (DSA) explicitly forces very large online platforms to identify and mitigate systemic risks, including deceptive advertising and fraud.

The Reuters report adds a blunt talking point to that stack. It suggests that Meta’s own internal documentation recognises fraudulent or borderline campaigns as a significant source of revenue. Once that is on the record, regulators can argue that the problem is not just technical difficulty, but also misaligned incentives.

Why fraudulent ads are hard to stamp out technically

Even if Meta had perfect internal incentives, this would still be a difficult problem. At scale, fraud and abuse are adversarial problems. The other side moves.

Adversarial creatives – Once scammers see which phrases or layouts get blocked, they adjust. They randomise text, lean more on images or video, mix languages and adopt legitimate sounding branding. Classifiers trained on last month’s scams quickly go stale.
Local rules and language – What is legal in one country might be illegal in another. Licensing rules differ. Disclosure requirements differ. A global filter that tries to encode all local nuance is bound to make mistakes at the edges.
Scale of review – Billions of impressions and millions of ads mean Meta cannot have a human look at every creative, landing page and funnel in depth. Automated review plus sampling is the only realistic operating model.
Attribution and tear down – When a campaign is flagged as fraudulent, tracing it back through ad accounts, payment methods, hosting and mule entities is complex. Scammers arrange their infrastructure to be disposable by design.

These constraints explain why scam ads are still present. They do not explain or excuse a situation where enforcement is kept weak enough that well known patterns of abuse remain highly profitable for long periods.

What a serious Meta response would look like

If Meta wants to convince regulators that it is not structurally dependent on scam derived revenue, it will need to do more than publish generic transparency reports. A credible response would have a few clear components.

Hard, published numbers – Regular data showing how many financial and scam ads are blocked or removed, broken down by region and category, with trends over time and independent audit where possible.
High risk category gating – Stronger onboarding for anything in investments, trading, crypto, unregulated health cures and similar segments. That means licence checks, identity checks and tighter rules on creative formats.
Visible revenue impact – Acknowledging that certain categories have seen large revenue drops because of policy changes, and treating that as a success metric in safety, not a failure in sales.
Better user protection paths – Clearer reporting tools, warning labels around high risk ad types, and visible support or cooperation with authorities when scams are confirmed.

The internal culture piece matters as much as any specific model or rule. If performance reviews and bonus structures focus on total ad spend and growth without discounting “toxic” revenue, the pressure to keep harmful but profitable campaigns running will always be there.

Why this story goes beyond Meta

Meta is the focus of the Reuters investigation, but the pattern applies to any large self service ad platform. Search engines, video platforms, app stores and other social networks all run similar funnels. The details differ, but the structural tension is the same.

If regulators decide that Meta has to move to a much more conservative stance on financial and high risk ads, that new baseline will bleed into expectations for other platforms. Possible outcomes include:

Common licensing checks for financial promotions across major ad platforms in a region.
Shared liability rules that treat platforms as co responsible for certain categories of misleading or fraudulent ads.
Requirements that ad risk and compliance teams sit outside direct revenue reporting lines, similar to how risk is handled in regulated finance.

In that scenario, the cost of running a self service ad business rises for everyone. Margins compress, but trust and regulatory stability improve over time. That is the trade regulators are trying to force.

My read on the Reuters documents

Assuming the Reuters story accurately reflects Meta’s internal documents, the key takeaway is simple. Fraudulent and borderline ads are not an incidental spillage. They are big enough to be seen and tracked internally. That changes how we should interpret Meta’s posture.

At this point, the interesting questions are not “does Meta have scam ads” or “are they hard to detect”. Those are settled. The questions that matter are:

How much ad revenue is Meta prepared to give up in order to materially reduce scam volume.
How quickly it is willing to introduce high friction onboarding for high risk categories, knowing that it will upset some legitimate advertisers.
How comfortable regulators are letting self policing run its course, versus writing hard rules that redefine what platforms are allowed to accept in the first place.

The answer will not come in one quarter. It will show up over a few years in how Meta’s ad mix, revenue growth, enforcement statistics and regulatory relationships evolve. For now, the documents that Reuters has surfaced make one thing clear. The era where platforms could treat scam driven ad revenue as an unfortunate but ignored byproduct is ending. The next phase is about how much pain they are willing to take in order to clean it up.