Using Internal Docs, Burner Accounts and Ad Library Research to Expose Meta's Scam Ad Business
Show & Tell from IndicatorShow Notes
In this episode of Show & Tell, hosts Craig Silverman and Alexios Mantzarlis sit down with Jeff Horwitz, investigative reporter at Reuters. Jeff shares how he showed that Meta has earned billions of dollars from online scam ads—and how he used the company’s own ad system to prove his point.
Jeff details how he combines leaked internal documents with hands-on experiments, including running his own "scam" ads to see how Meta’s AI tools would react. You’ll hear how Meta’s generative AI didn't just approve his deceptive ads—it offered suggestions to make them even sleazier.
What You'll Learn: His Approach: How Jeff uses document analysis, quantification methods, burner accounts, and internal sources to uncover stories. The Scam Ad Experiment: Jeff explains how he partnered with Meta's own "trusted" ad agencies in China to run ads that explicitly violated platform rules. AI-Enhanced Deception: Learn how Meta’s LLMs suggested high-pressure copy like "guaranteed profit" and "tired of living paycheck to paycheck" to improve scam performance. The "Black Box" Problem: Why even internal Meta staffers often rely on setting up "clean" test accounts to understand what their own algorithms are doing. Sourcing via LinkedIn: Jeff shares his "old school" but effective outreach tactics using LinkedIn Sales Navigator to find sources within big tech.
About Jeff Horwitz: Before joining Reuters, Jeff covered Meta for The Wall Street Journal where he broke major stories on child exploitation on the platform, Instagram's negative impact on teen mental health, facilitation of human trafficking, and the amplification of political violence. His Facebook Files investigation used internal Meta documents to expose systemic issues across the company's platforms.
Episode Highlights: [1:46] Introduction to Jeff Horwitz and his investigative work [4:47] How Jeff started investigating Meta's ad scam problem [7:37] The power of documents in investigative journalism [37:25] Meta's response to the scam ad reporting [39:30] Lightning round: Jeff's digital tools and investigative methods
More from Show & Tell from Indicator
In this episode of Show & Tell, we spoke to Shayan Sardarizadeh about his approach to debunking viral fakes for BBC Verify. "I want to check everything for AI," he told us. "I don't care where it comes from. If it comes from a national TV, if it comes from a politician posting it on their official TikTok, I don't trust anything." What You'll Learn: 📍Choosing where to look, and what to ignore when it comes to viral fakes 📍Why AI detectors are now an essential element of verification and why Shayan prefers the SynthID-based ones 📍How detectors can still fail, with a worked example from a viral image of Donald Trump 📍How reverse search is an essential complement to AI detectors and a mandatory starting point for any verification task Links mentioned in the episode 📍Shayan's thread on the viral Trump image: https://x.com/Shayan86/status/2100642920437108946 📍"AI generators are now required to offer detection tools. We tested them (and they need work)" (Indicator): https://indicator.media/p/ai-generators-are-now-required-to-offer-detection-tools-we-tested-them-and-they-need-work 📍You can follow Shayan on X at @Shayan86 and on LinkedIn: https://www.linkedin.com/in/shayan-sardarizadeh-0121555a/ About Shayan Sardarizadeh: Shayan is a senior journalist at BBC Verify covering disinformation, conspiracy theories and AI. About Show & Tell: Show & Tell from Indicator takes you inside digital investigations as leading journalists, OSINT investigators, threat analysts, and researchers walk through real cases they've cracked. Each episode features a recognized figure in the OSINT, digital investigations, or trust & safety community. They don't just tell us what they did — they show us. Hosted by Indicator co-founders Alexios Mantzarlis and Craig Silverman.
In this episode of Show & Tell, we speak to Renée DiResta about Grokipedia and the contestability of AI-mediated information sources. Grokipedia is worth studying not just as “a culture war artifact,” she said, but because even if Musk’s AI encyclopedia fades away, “something like it will exist at some point, and we may as well get in and figure out how to audit it now.”
In this episode of Show & Tell, hosts Craig Silverman and Alexios Mantzarlis sit down with Wall Street Journal reporters Caitlin Ostroff and Neil Mehta, who found that Polymarket paid creators to place fake bets on a faked version of its platform.
In this episode of Show & Tell, researcher Hal Triedman showed us how easy it can be to poison an AI research agent.
In this episode of Show & Tell, Craig and Alexios talk with Omer Benjakob, the disinformation and cyber reporter for Haaretz.