Our weekly Briefing is free, but you should upgrade to access all of our reporting, resources, and a monthly workshop.
This is your last chance to get 20% off all Indicator memberships. Join before midnight tonight!
This week on Indicator
Alexios found that TikTok content farms have published conspiracy theories and fictional AI-generated tributes about Charlie Kirk. Just 53 videos got 32 million views on the platform. TikTok deleted most of them after Indicator reached out. (Related: Lead Stories found synthetic songs for Kirk on YouTube.)
Craig wrote the Indicator Guide to tools for capturing webpages and social media content. He evaluated 11 tools you can use to capture webpages, social media accounts, and other digital assets as you browse. Paid members also get access to the beta of a promising new tool in development.

Meta slowly spins up a Community Notes model that X seems to be abandoning
Last week, Meta CISO Guy Rosen finally posted an update about the company’s Community Notes program, which launched in March.
Rosen said that 70,000 users have contributed more than 15,000 notes to the crowdsourced context labeling feature. Of those, about 6% have been published. That means the notes were found to be helpful by enough users with a range of different viewpoints in order to be publicly attached to a post on Facebook, Instagram, or Threads.
It’s hard to know what to make of the figures. In the same six-month period, 325,000 notes were proposed on X, 8.7% of which were published. The comparison isn’t entirely fair: X’s program also started small with just ~1,000 contributors authoring about 9,000 notes in the first six months of its existence. So there’s time for Meta’s program to grow.
This is the first time that Meta has shared meaningful data about Community Notes since its launch. The fact that it came from a senior executive suggests that the company remains committed to the feature.
Still, neither Meta nor TikTok — which said in July that its crowdsourced labeling feature called Footnotes has 80,000 contributors — provide anywhere near the same level of transparency as X, which has opened up basically everything.
Thanks to three Indicator readers in the Meta program, on Thursday I collected screenshots of 24 posts that users had either requested or proposed notes on. (If you’re in the program and are willing to be contacted in the future, let me know here).
Compared to what I saw three months ago, the notes being proposed in the “Write” tab for Instagram users appear to be less random and more fact-checkable. Facebook, on the other hand, continues to ask contributors to write notes on mugshots and petty social media beefs.

Overall the posts were more frequently about politics and current events than the last time I checked. Posts from the White House and claims about Charlie Kirk and Jimmy Kimmel dominated the sample. This could just be a consequence of the centrality of such topics in the American information space right now.
The notes themselves were a mixed bag. A handful correctly pointed to relevant fact checks, while others added unhelpful opinions.

None of the people who sent me screenshots have seen a published note on their main feed, except for Alex Mahadevan of Poynter, who saw one because it was his own. With just 900 published notes to date across three social networks, this isn’t terribly surprising.
In all, Meta’s Community Notes is still a relatively limited exercise given its central role in Mark Zuckerberg’s January capitulation to the Trump administration.
Ironically, Meta is slowly spinning up a replica of a program that appears to be sputtering on X. As shown in the below charts, the number of notes authored and published have been on a gentle decline on Elon Musk’s platform, and its main architects are signaling they want AI to play an ever larger role. — Alexios

Deception in the News
📍Politically-charged AI-generated parody appears to be a growing genre around the globe. An unlabeled video that showed fictitious counter protestors making outrageous claims at a far-right rally in London got 1.8 million views on X. And Zvezda, a TV network run by Russia’s defense ministry, launched an entire program featuring sloppy deepfakes of world leaders. (h/t GDELT and 404 Media)
📍 Australians were targeted with scam ads for sky lantern festivals that never took place. The country’s Consumer Protection commissioner said the events were fake and and that the associated websites were set up to collect payment details and personal information.
📍 BOOM revealed that YouTube creators are making money with deceptive videos about Trump’s tariffs on India, including “deepfake videos and AI voice clones impersonating international public figures who appear to be praising Modi’s statesmanship and projecting India’s diplomatic clout internationally.” YouTube suspended 20 channels flagged in the report.
📍 An AI-enabled podcast company is reportedly churning out 3,000 episodes per week. “We believe that in the near future half the people on the planet will be AI, and we are the company that’s bringing those people to life,” CEO Jeanine Wright told the Hollywood Reporter.
Tools & Tips

📍 Boudjenane Soufiane highlighted how manhole.co.il can help in OSINT investigations. As of this writing, the site says it has details about “2,952 different [manhole] covers, from 76 countries and 1,039 cities.” Soufiane noted that the site’s search function “lets you filter by city, country, category, manufacturer, shape, or even keywords that appear in the description or comments.”
📍 Pavel Bannikov is an editor with Provereno Media, an Estonian-based fact checking outlet. He created two useful, free bookmarklets:
EXIF Data Extractor “scans all images on the current webpage and allows you to check their EXIF metadata.”
Scam Surface Mapper “analyzes web pages to identify and visualize potential scam infrastructure by mapping outgoing links, analyzing suspicious domains, tracking parameters, and presenting findings in an interactive graph interface.”
You can also read our recent guide to using bookmarklets to reveal hidden online profile info and speed up OSINT investigations.
📍 PreverseTube is a free archiving tool for YouTube videos and channels. Along with archiving the video, it saves “descriptions, titles and other interesting metadata.” You can also search within its database of archived videos and channels. (via Cyber Detective)
📍 YouTube Video Finder is another great place to search for archived videos. It searches across 12 services. (via Logan Woodward)
Events & Learning
📍 The Centre for Investigative Journalism is holding a four-part workshop, “Climate Investigations Course: ECOSINT.” It starts Sept. 23 and has a sliding fee scale.
📍 Bellingcat is holding a four-hour workshop on Sept 30, ”Command Line Tools for Open Source Researchers (for Absolute Beginners).” It costs €350.
📍 Media & Learning is holding a free webinar on Sept 24, “Persuasion by design: understanding Influence(rs).”(via the EU Disinfo Lab newsletter)
Reports & Research
📍 Reuters found that major AI chatbots can be cajoled into drafting phishing emails. The news org tested the messages’ efficacy on 108 senior citizen volunteers. 11% of targets fell for them.
📍 A new paper in the Journal of Online Trust & Safety concludes that in the context of the US 2020 election, “exposure to untrustworthy websites is associated with the likelihood of believing that the certified election winner did not win by 4.2%.”
📍 Open Measures tracked the evolution of the “Pallywood” and “Gazawood” narratives that claim the evidence of death and devastation in Gaza has been manufactured or overhyped. The claims peaked at the beginning of the Israeli invasion of Gaza and again in July 2025; the locus shifted from 4chan and Truth Social to TikTok and Bluesky, according to the report.
📍 The Foundation for Defense of Democracies said it uncovered a likely Chinese intelligence operation that created fake consulting companies in order to recruit new assets.
📍 Participants of an online focus group conducted by the Digital Democracy Institute of the Americas said they often turn to Google Search and ChatGPT to fact-check information they see online.
Want more studies on digital deception? Paid subscribers get access to our Academic Library with 55 categorized and summarized studies:
One Thing
Italian news magazine L’Espresso posted an AI-generated image of the bullets engraved by accused Charlie Kirk assassin Tyler Robinson. The outlet disclosed the use of AI in the post but not on the image itself, which we should expect to get reused out of context. (We added a “Made with AI” label to the image just to be safe.)

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.