In November 2025, Reuters reporter Jeff Horwitz published an investigation built on internal Meta documents. The headline number: late in 2024, Meta projected that about 10% of its total annual revenue, roughly $16 billion, would come from ads for scams and banned goods.
- projected revenue from scam and banned-goods ads
- $16bn
- of Meta's total annual revenue, by the same estimate
- ~10%
- “higher-risk” scam ads shown a day, per a December 2024 document
- 15bn
Meta’s own systems flagged many of these ads: fake investment schemes, deepfaked endorsements, counterfeit goods. According to the documents, an advertiser was banned only when the system was at least 95% sure it was committing fraud. Below that line, suspected scammers paid higher ad rates and kept running.
Meta disputes the reading. Spokesman Andy Stone told Reuters the documents give a selective view, called the 10.1% estimate “rough and overly-inclusive”, and said the real figure was lower. He didn’t say what it was.
I run guwAPI, a small software studio in Bucharest. Numbers like these don’t make me think about ad markets. They make me think about what happens when a system is tuned for one metric and everything else becomes somebody else’s problem.
Optimise for volume, and trust pays the bill
Systems do what they are optimised for. When the target is volume and short-term revenue, quality, security and user safety turn into costs to keep down. Nobody has to decide to harm users; the incentive does it quietly.
Every engineering team meets a smaller version of the same trade-off: skip the validation, ship the shortcut, deal with it later.
Trust is cheap to lose and expensive to win back.
And once a platform feels unsafe, everyone who depends on it pays: honest advertisers, small businesses, users.
How we work at guwAPI
We build web applications, API integrations and internal systems for businesses. We are nowhere near Meta’s scale, but the principle doesn’t change with size: software should be built to last, built to be secure, and honest about what it does.
In practice, that comes down to three rules:
- Security and clean architecture aren’t extras. Data protection, strict validation and stability are not bolted on at the end. They are the foundation of everything we ship.
- No grey areas for sale. If a system produces a grey area, the fix is to close it, not to find a way to bill for it.
- Quality over volume. A solid system a business can rely on beats a fragile one shipped to tick a box.
What comes next
Users and businesses are tired of fraud, noise and flimsy apps. Over the next few years, features alone won’t separate good digital products from mediocre ones. Trust will.
Going digital should make things simpler and safer, not riskier. We would rather build systems that hold the weight of the businesses that run on them.