This week on Indicator
Craig shared 32 tips, tools, and techniques from this year’s Investigative Reporters and Editors Conference. They include how to surface sealed court cases, unmask anonymous companies, and trace crypto.
Alexios wrote about 20 doppelgänger domains buying Google Ads for queries about tickets for 15 top cultural attractions in order to fleece tourists. The ticket trap for the Louvre charges €76 for entry when the official price is just €22.
Alexios wrote about a clever travel scam: 20 doppelgänger domains are exploiting Google Ads to target tourists seeking tickets for 15 major cultural landmarks.. Case in point: a Louvre ticket trap charging €76 for an entry that should only cost €22.
Also: Indicator’s rough taxonomy of AI slop was featured in the proceedings from a Columbia Institute for Global Politics event on the topic.
Deception in the News
The Wetstraat 16 - 16 Rue de la Loi offices of the Prime Minister / BELGA PHOTO
📍 Flemish newspaper De Standaard listed the Belgian Prime Minister's residence on Airbnb, using the PM’s name as host. The rental was live for almost two months and received several booking requests (h/t Maarten Schenk).
📍 The British government is exploring “legislative options to require social media to make trustworthy news providers—which could include national and local news publishers and broadcasters—easily discoverable.” (It’s one of those ideas that seem great until you have to consistently and fairly define “trustworthy news providers.” — Alexios)
📍 Wikipedia’s long-estranged cofounder Larry Sanger was banned from the online encyclopedia for urging his off-platform followers to support a proposal for a project dedicated to “intellectual diversity.”
📍 Republican candidate for New York City Council Jonathan Rinaldi was arrested for allegedly forging endorsements from local powerbrokers. One of them was an AI-generated photo of a Democratic council member shaking hands with Rinaldi.
📍 An Italian AI chatbot marketed as an example of “digital sovereignty” lasted just a few days before being taken offline due to its outrageous responses. Among the reported goofs were recommending an AK-47 as a gift for a 5-year-old and claiming that Ryanair was a type of liquid yoghurt for airplanes.
📍 Also in Italy: The Five Star Movement posted AI-generated selfies of Prime Minister Giorgia Meloni posing with US President Donald Trump to illustrate commentary about a high-profile dispute between the two former allies. Pagella Politica called it a violation “in spirit” of the agreement it brokered among the country’s political parties to not publish deepfake videos of their adversaries.
📍 Two West Papuan activists told ABC News they have been targeted by deepfakes that misrepresent their positions.
📍 The Oversight Board overturned Meta’s decision to allow a seemingly AI-generated video that placed an Instagram user in a sexualized setting. The Board noted that “AI-generated impersonation is non-consensual by default and should be added to the set of signals the company uses to establish lack of consent.”
Tools & Tips
Matt Edmondson, a SANS instructor and founder of Argelius Labs, released “China Web OSINT Search Dorks,” a detailed guide to searching the Chinese web for people and companies of interest. Edmondson offers lots of examples, useful Chinese terms and phrases, advice for OPSEC, and more.
An excerpt:
Find the target's canonical Chinese name and high-confidence identifiers.
Search the name in Simplified Chinese, then test Traditional Chinese, English, romanized names, abbreviations, and former names.
Combine the target with a Chinese pivot term such as
招聘(recruiting/jobs),中标公告(award notice),行政处罚(administrative penalty), or简历(résumé/biography).Narrow the query with a source domain, document type, title term, or exact identifier.
Run the same concept across multiple search engines, platform-native searches, and official databases.
Treat search results as leads—not proof—and verify important claims against primary records.
📍 Pavel Bannikov created a free bookmarklet that integrates with the bulk option of Matt Wright’s MW Metadata YouTube tool. MW Metadata can extract a variety of data from a YouTube video, channel, or playlist. (The bulk option works for multiple videos.) Bannikov’s tool lets you quickly export metadata associated with a channel, playlist or set of videos to CSV, TSV, JSON, or Markdown.
📍 EmailCrawl is a new tool from Precious Vincent that extracts email addresses “from HTML content, meta tags, and mailto links.” It also performs validation and analysis.
📍 David Papava shared a simple OSINT Risk Evaluation Matrix that he says helps him “think through different types of risk: operational, legal, privacy, reputational, organizational, and ethical/professional.”
📍 The OSINT Insider published a new edition that looks at “the ways in which AI has transformed OSINT in practice, the future outlook for OSINT as an industry, and monitoring tools that elevate capabilities.”
📍 Arno Reuser wrote, “How to start an OSINT career.”
📍 Ionatan Waisgluss wrote, “Business Due Diligence: A Robust Starter Guide to Researching a Company.”
📍 Craig wrote a guest post for Wonder Tools, “A Digital Investigator’s Free Toolkit.”
📍 Benjamin Strick published a new video, “How I Track Ships in the Strait of Hormuz”:
Reports & Research
📍 A Wall Street Journal investigation found that prediction market Polymarket paid “dozens of mostly college-age creators […] to film themselves making fake trades and sometimes scoring fake wins.” The videos used fake websites or unrelated bets, and the company reportedly instructed influencers to not disclose they were paid.
📍 This year’s Digital News Report is out, and the proportion of respondents who agreed with the statement “Thinking about online news, I am concerned about what is real and what is fake on the internet" shifted slightly upwards to 62%. Only 11% disagreed.
📍 A study in PLOS Digital Health examined 971 of the most viewed TikTok videos related to sunscreen. While the researchers found that only a minority contained misinformation, those misleading videos averaged twice as many likes and shares as content promoting sunscreen use.
📍 NewsGuard found that far-right conspiracy theorist Laura Loomer saw her average engagement on X drop by 28% in the week after she renounced her former pro-Russian position. The company suggested that the drop could be a sign of prior inauthentic promotion by pro-Russian networks.
Want more studies on digital deception? Paid subscribers get access to our Academic Library with 75 categorized and summarized studies:
One More Thing
The @FacebookAIslop X account has been chronicling absurd examples of AI slop for a little more than two years. Here’s a recent example:
However, the account’s creator recently acknowledged they’re posting less frequently. It’s not from lack of interest, but because rapidly improving AI models make synthetic images “harder to notice and … to make fun of.”
Indicator is a reader-funded publication.
Please upgrade to access to all of our content, including our how-to guides and Academic Library, and to our live monthly workshops.
