Thank you to our paying members! You make our work possible. Remember that our first members-only workshop is today, June 27, at 12 pm ET. Reply to this email if you need the link.
Our weekly Briefing is free, but if you’re not a member you should upgrade to access all of our reporting, resources, and the monthly workshop. We’re offering a limited time 20% launch discount that ends in just a few days. If you upgrade now you can join today’s workshop or access the recording.
This week on Indicator
Alexios wrote about the extensive ad campaign run by the Israeli government in five European countries to promote its justification for bombing Iran.
Craig published the Indicator guide to connecting websites together using OSINT tools and methods. It’s 5,000 words of detailed instructions for free (and paid) tools that help you link sites together in order to uncover ownership, connections, and funding.
Deception in the News
Adieu ClaimReview. Google announced in a relatively muted fashion that it was killing off support for ClaimReview, the structured data that powered its fact-checking features in Search and News. This is directionally consistent — though far less dramatic in tone and impact — with Mark Zuckerberg’s decision to terminate Meta’s US fact-checking program earlier this year.
ClaimReview served to inform Google’s crawlers of the claim and rating contained in a fact-checking article. The information was highlighted to users in different ways across News, Search, Images, and YouTube.
I worked extensively on ClaimReview while at Google, including by driving the launch of a dedicated section for COVID-19 fact checks in Google News, and publishing information on the billions of impressions fact checking snippets got on Search. Google hasn’t launched new fact-checking features in a while, leaving them out of About this Result on Search and showing them sparingly on YouTube.
A Google spokesperson told me that removing ClaimReview and six other structured data types has no impact on the ranking or the traffic of the affected webpages.
In some ways, the advent of LLM-powered Search made this transition inevitable. Google doesn’t need, and users may not want, structured data that comes in categories like “claimReviewed” and “reviewRating."
On the other hand, the company has repeatedly made ClaimReview the centerpiece of its commitment to the fact-checking ecosystem. In communicating its actions to combat disinformation to the European Union last year, Google noted that fact check snippets were seen 120 million times in the first half of 2024 across the 27 Member States. And users claim they seek fact-checking websites much more than AI chatbots when trying to verify the news, according to the Reuters Institute.
It is also notable that in discontinuing ClaimReview, the company made no mention of replacing it with a fact-checking feature suited to the AI age.
Fact-checkers meeting this week for their annual conference tell me they were unaware of Google’s plans but not surprised by the decision. Bill Adair of Duke University (and a godfather of ClaimReview) struck a “we don’t need ‘em” tone, posting that the benefits of structuring fact checks go beyond what Google does with them. “There are many products, including some that haven’t been invented yet, that will take advantage of our ClaimReview database,” he said.
Indeed, ClaimReview has helped inform a pretty rich corpus of academic research by providing a global database of fact checks. Google will also continue to maintain its Fact Check Explorer.
Other fact-checkers were less equanimous. Carlos Hernández-Echevarría of Maldita called the move “as reckless as it gets.” Andrew Dudfield of Full Fact wasn’t angry but he was disappointed, writing that “anyone looking for good information will now have a more difficult time: quite the opposite outcome of Google’s stated intention to simplify search results.”
— Alexios
In other news this week:
📍 Bellingcat obtained a Stripe screenshot that suggests nudifier Crush AI earned at least $45,000 from its abusive services. Read more of our coverage of Crush AI here.
📍 Elon Musk got in a fight with Grok, his AI chatbot, and threatened to retrain it on a “rewritten” version of “the entire corpus of human knowledge, adding missing information and deleting errors.”
📍 Meta’s Oversight Board overturned a company decision not to label a likely AI-generated audio clip involving Kurdish political leaders Bafel and Qubad Talabani. The board said it was concerned that “despite the increasing prevalence of manipulated content across formats, Meta’s enforcement of its manipulated media policy is inconsistent.”
📍 Videos falsely claiming Italian Prime Minister Giorgia Meloni has announced a partnership with the far-right AfD party in Germany are doing numbers on TikTok, YouTube and Instagram. Their audio track is likely AI-generated, reports Correctiv.
📍 What To Fix published an investigation that found Facebook pages affiliated with Russia Today, Sputnik and other EU-sanctioned entities were accepted into Meta’s revenue-sharing program. Some are still active, and potentially receiving payouts, as recently as this month. “Our findings call for urgent scrutiny into Meta’s revenue redistribution partnerships as well as more regulatory oversight of social media monetization systems and governance practices,” What To Fix said.
📍 The Foundation for Defense of Democracies (FDD) found a large Persian-language Telegram channel is instructing people on “how to use AI and create fake accounts on X where they pose as Israelis and post demoralizing messages in Hebrew.”
Tools & Tips
📍 The Data and Research Center added geocoded addresses to OpenAleph, an open source platform for storing and analyzing data. (Run a search here.) “With geocoded addresses now built into OpenAleph, you can map entities, spot who shares an address, explore what’s nearby, and uncover links that aren’t immediately obvious,” read an announcement from DARC. You can learn more in this blog post.
📍 Rhino User Checker is a username enumeration tool. It’s nice that the results page includes profile pics. (Via Cyber Detective)
📍 Airwars, which runs the Open Source Munitions Portal and investigates civilian harm in conflicts, recently launched a new website. Details here.
📍 Vantage is an open source tool that “enables users to reconstruct spatial contexts of images and videos by projecting them onto 3D environments.”
📍 At the Investigative Reporters & Editors conference last weekend, Bloomberg reporters Diana Dombrowski, Umar Farooq, and Gary Harki presented, “Turn your questions into data with an LLM.” You can download the PowerPoint slides here.
📍 Paul Bradshaw maintains a helpful list of resources and tools for journalists related to generative AI.
Events & Learning
📍 Last call for the first members-only Indicator workshop! Members can reply to this email to receive the Meet link. Or upgrade your subscription and join us today at noon US Eastern time, when Craig will walk you through applying the Actor, Behavior, Content framework to analyze online reviews for signs of inauthenticity. He’ll also show how to use the free Instant Data Scraper browser extension to grab data from online platforms like Facebook and YouTube, and demo some additional tools.
Reports & Research
📍 DFRLab found that Grok made a right mess when asked to verify a video that falsely claimed to show Iranian missiles had destroyed Tel Aviv’s Ben Gurion airport.
📍 In the International Journal of Communication, three scholars argue that the proliferation of anti-misinformation laws around the world between 2010 and 2022 stemmed from a combination of factors that included “the desire for governments to control the flow of information and high-profile revelations around platform governance.”
📍 A systematic review of studies about climate misinformation by the International Panel on the Information Environment concluded that “coordinated misinformation campaigns actively shape climate narratives.”
📍 Lupa published a report on digital scams in Brazil informed by the 142 incidents it has covered since 2020.
📍 Operation Overload, a Russian-aligned influence operation that seeks amplification by volunteering its false content for fact-checking, has expanded to Bluesky and TikTok, according to CheckFirst.
Want more studies on digital deception? Paid subscribers get access to our Academic Library.
One Fun Thing
John Oliver covered AI slop on Last Week Tonight. It got wonderfully silly.
Indicator is a reader-funded publication. Thank you, members!
If you’re not a member, you should upgrade to take advantage of our limited time 20% launch discount. You get access to all of our content, including our how-to guides and Academic Library, and to our live monthly workshops.
