Meta profits from scams.
Brzoska says: prove it.

The internet's advertising model is straightforward.

The advertiser pays. The platform shows the ad. The platform earns money.

Things get more interesting when the advertiser is a scammer.

They still pay. The platform still takes the money. But the user at the other end may be watching a deepfake of a famous person promoting a “sure-fire investment” — something with about as much connection to investing as my final_final2 folder has to version control.

That is exactly the sort of case that sparked Rafał Brzoska's dispute with Meta.

It began with a “Brzoska” who was not Brzoska

In 2024, investment ads using the likenesses of Rafał Brzoska and his wife, Omena Mensah, appeared on Facebook and Instagram. One video supposedly showed a deepfake Brzoska promoting a platform that would let Polish people earn money. Other ads featured fabricated stories about Mensah, including claims that she had been assaulted, arrested or had died.

Brzoska reported the matter to Meta and later to Poland's Personal Data Protection Office (UODO). In August 2024, the president of UODO ordered Meta Platforms Ireland to temporarily stop displaying fake ads using his personal data in Poland. Meta challenged the decision but withdrew its complaints in March 2025.

That did not end the dispute.

In March 2026, Warsaw's Court of Appeal, considering interim relief in the Brzoska and Mensah case, addressed a more significant question: is Meta merely a passive “place” where someone posts an ad?

According to the reasoning reported by XYZ, the court found that the platform plays an active role in advertising: Meta accepts the order, collects payment, provides the targeting system and optimises distribution. That matters because it challenges the simple defence that “we only host the content”.

Then came a figure: $16 billion

This is where we should slow down, because headline figures can quickly take on a life of their own.

In November 2025, Reuters published a major investigation based on internal Meta documents from its finance, safety, engineering and lobbying departments.

According to those documents, at the end of 2024 Meta estimated that around 10.1% of its total annual revenue could come from ads linked to scams and other prohibited categories. Reuters calculated that as approximately $16 billion.

There is a crucial detail here, though.

$16 billion does not mean “$16 billion from scams alone”.

That broad category also included illegal gambling, prohibited goods and other ads that breached the platform's rules.

However, the same Reuters report gave a second, more specific figure: according to another internal document, ads described as “higher risk” scam ads— those showing clear signs of fraud — were expected to generate around $7 billion in annual revenue.

That is still an absurd amount of money.

But the difference between $7 billion and $16 billion is large enough that we should not lump them together.

Where did Reuters get the data?

Not from a user survey, nor by counting strange ads on someone's own feed.

Reuters examined a set of unpublished documents produced within Meta between 2021 and 2025. According to the report, one document from December 2024 estimated that users of Meta's platforms see an average of 15 billion high-risk ads a day.

The documents also showed that the automated system was supposed to block an advertiser when it assessed the likelihood of fraud at no less than 95%. At lower confidence levels, suspicious advertisers could face higher prices in the ad auction — known as penalty bids.

Reuters also described internal limits on the impact of anti-fraud measures on revenue. A document from early 2025 said that the team checking suspicious advertisers was to operate within “revenue guardrails” equivalent to around 0.15% of Meta's revenue in the first half of the year.

Does that sound bad? Yes.

But fairness requires the other half of the story.

Meta says the figures are being presented without context

Meta disputed Reuters' interpretation of the documents. Company spokesperson Andy Stone said the 10.1% estimate was “rough and overly inclusive” and covered many legitimate ads. The company said subsequent calculations produced a lower figure, but did not give Reuters an updated number.

Meta also says that penalty bids were not a way to “earn more from scammers”, but a way to make suspicious advertisers lose more auctions. According to the company, tests of the approach led to fewer scam reports and a small fall in advertising revenue.

In March 2026, Meta said that across 2025 it had removed more than 159 million scam adsglobally, with 92% removed before a user reported them.

In response to the current Polish debate, the company published further figures on 28 August: between July 2025 and June 2026, it said it had removed 137,000 fraudulent adsin Poland, with more than 88% removed proactively. According to Meta, the number of scam-ad reports in Poland fell between July 2024 and June 2026 by 83%.

What was reporting a scam like for an ordinary user?

Here we have a Polish test by CERT Polska.

From January to November 2024, experts used an ordinary Facebook account to report 122 adsthat they considered fraudulent.

• 10 were removed,
• in 106 cases, the report was closed with the message “We didn't remove the ad”,
• in 6 cases, there was no feedback.

This is not evidence of the current effectiveness of Facebook's entire moderation system. It was a limited test, from a specific period, using a specific method.

But it illustrates very well why people become frustrated when reporting an ad that looks like a scam from a mile away.

PLN 20 million a day in Poland? Time to apply the brakes

In recent days, Rafał Brzoska has put forward a much stronger figure: according to his team's estimates, Meta could be earning as much as PLN 20 million a day.

Multiplied by a year, that gives around PLN 7.3 billion.

And here we have a problem.

In February 2026, Juniper Research, in a study commissioned by Revolut, estimated that all the social platforms it analysed combined earned around €172 million, or approximately PLN 725 million.

That is roughly ten times less than the annualised version of Brzoska's current estimate — and it covers not just Meta, but also TikTok, Snapchat, X, LinkedIn and the other platforms analysed.

So no: at present we cannot honestly present “Meta earns PLN 20 million a day from scams in Poland” as an established fact.

Brzoska describes it as a preliminary estimate and has announced a report that will explain the methodology. Until we see it, PLN 20 million a day remains his team's claim.

He may be using a different definition of a scam, a different period, different sources or a much broader estimation method. For now, we simply do not know.

That is why it is worth distinguishing a strong headline from solid evidence.

What do we know from the Juniper Research study?

The Juniper Research study is also an estimate, not access to the platforms' accounts. It was commissioned by Revolut, an organisation with a strong interest in payment fraud. Its authors used data required under the EU's DSA, their own models and other sources, including the Reuters investigation.

According to that study, in 2025:

• Europeans were shown around 993 billion scam ads,
• the platforms were estimated to have earned around £3.8 billion / €4.4 billion in revenue from them,
• in Poland, fraudulent ads were estimated to have generated around 37 billion impressions,
• they accounted for more than 6% of all ad impressions on Polish social media,
• the platforms' revenue from those ads in Poland was estimated at around €172 million.

These figures are much better documented than the current “20 million a day”, but they are still a research model, not Meta's financial report.

Enter ScamWatch

The new ScamWatch was developed as part of Rafał Brzoska's 150% initiative. Technically, the idea is more interesting than simply “report an ad”.

A user submits a suspicious ad. The report is to be verified with help from experts and AI. The system analyses the content, image, domain, infrastructure and links to other campaigns, among other things.

So the aim is not just to say:

“this ad looks suspicious”.

It is about connecting multiple ads to a single operation, preserving evidence, passing it to the platform and — the most interesting part for me — measuring the response time after a report.

If the system starts collecting enough data, the discussion could move from:

“I feel Facebook doesn't respond”

to:

“we reported X ads, Y were removed, and the median response time is Z”.

That is when things get much more interesting.

You do not need a conspiracy theory to see a conflict of interest

In my view, this is the most important part of the story.

There is no need to imagine someone in Menlo Park sitting beside a red button labelled “KEEP THE SCAM — IT MAKES MONEY”.

The business model creates the tension itself:

That alone does not prove bad faith.

But if the company's internal documents really contain “revenue guardrails”, discussions about how enforcement affects financial forecasts, and arrangements where a suspicious advertiser pays more to enter an auction instead of being automatically removed, it is hard to pretend that economics is not part of the problem.

Meta responds that scams are also bad for its business because they damage the trust of users and legitimate advertisers. That makes sense too.

Both things can be true at once.

A scam can bring the platform short-term revenue while damaging its ecosystem in the long term.

What about the “150%” idea?

Brzoska's initiative proposes that a platform which earned money from displaying a particular scam ad should pay a penalty of 150% of the revenue it earned from that ad, with the money going into a compensation fund.

Should precisely that mechanism become law? That is a separate debate: it would require a definition of a scam, an appeals process, liability for misclassification and a method for calculating revenue from a particular ad.

But the underlying idea is very clear:

make it more financially worthwhile to stop a scam than to display it.

In that respect, the idea gets much closer to the heart of the problem than another campaign telling people to “be careful — there are scammers online”.

ScamWatch will become interesting when it produces data

For now, ScamWatch is a new tool and a statement of how it intends to operate.

The most interesting stage is only just beginning.

If, in a few weeks, we see a public record of reports, related campaigns, response times and details of what was actually removed, we will have something far more valuable than another exchange of PR statements.

We will have data.

Then we can return to that PLN 20 million a day and see whether the methodology holds up as well as the headline.

The biggest problem with scams is not a shortage of strong opinions. We have plenty of those. It is a shortage of data that neither side can conveniently interpret to suit itself.

If ScamWatch supplies that data, then whatever happens next in the Brzoska versus Meta dispute, it will be a useful piece of code.

And I have a soft spot for situations where the answer to a problem is:

“right, let's build a tool and check”. :D