A major perk for @indicator.media members is our monthly workshop. This Friday, @craigsilverman.bsky.social will walk folks through how to look for valuable documents hiding in open air, the topic of our latest guide. Upgrade to paid now to attend the workshop!

Today on @indicator.media's (free) Briefing: A twinkle of optimism on the AI slop front as TikTok and Google roll out tools to filter and detect AI content. I remain bearish about generative AI and info quality. But I’ll celebrate developments that increase user agency and platform transparency.

TikTok & Instagram have been deluged by thousands of videos from female students that praise Jenni, an AI study app. They say it's better than ChatGPT. They say it can improve your grades. What don't they say? That they were paid to promote Jenni.

Dishonest advertising meets AI grift in @craigsilverman.bsky.social's latest on @indicator.media:

Now that is an interesting finding.

(The bots are also leaning heavily on Wikipedia, fact-checkers and the media -- those very institutions Musk, Zuck, and co. think will be supplanted by Notes). Read, like, subscribe for all the good stuff:

New on @indicator.media: The single most active contributor to Community Notes since Sep. 1 is Zesty Walnut Grackle. Zesty is an AI bot, one of eight currently active on X after the platform onboarded them in a bid to boost flagging participation.

Check out the preprint behind our blog from last Thursday:

This newsletter pulls a few things together, including my Atlantic essay on working the referees, and @mantzarlis.com’s fantastic new analysis on what exactly is in Grokipedia.

Really crazy story from @mantzarlis.com for INDiCATOR about “Grokipedia” and the sources it’s citing. In addition to crazier stuff: “Grokipedia largely emulates Wikipedia in over half of its entries, seemingly content to transcribe pages verbatim…” indicator.media/p/grokipedia...

Good work here. Not unexpected, but very useful to have it clearly laid out.

So here's what we have: An AI-powered encyclopedia that heavily leans on a volunteer effort that Musk has repeatedly attacked, which selectively edits a subset of topics to recast based with an undisclosed set of instructions. For those pages, citations to conspiracy sites and hate groups go up.

After a week of reviewing the site-wide data and many entries in detail, I think Grokipedia has made significant editorial decisions around sourcing and topic treatment that lend credence to @matteowong.bsky.social's argument in The Atlantic that it is “the next step in Musk’s propaganda machine.”

NEW on @indicator.media: A first *full-scale* comparison of Grokipedia v Wikipedia. Last week the awesome @harold.bsky.social rocked up to my desk bearing gifts. Hal had collected almost all 900K Grokipedia entries and compared them to their Wikipedia equivalents for text and citation similarity.

Sensitive, confidential & interesting docs are sitting out in the open on the internet. Do you know how to find them? My new @indicator.media Guide dives into Google dorks & tools that can uncover publicly available docs on websites and in unsecured buckets: indicator.media/p/guide-to-h... #osint

So fun to see @indicator.media’s audit of AI labels being cited in places as different as Private Eye and The Hindu (The piece: indicator.media/p/tech-platf...)

"The question, however, is whether Grokipedia can succeed not as an alternative to Wikipedia but as a means to delegitimize it. 'Flooding the zone with sh*t' as a strategy doesn’t necessarily seek to replace institutions of collective knowledge generation; it seeks to deface them"

Today on @indicator.media: I speak to the lawyer representing the New Jersey teen suing AI nudifier ClothOff. We spoke about the broad coalition of lawyers fighting to shut down these tools of online abuse, the difficulty of getting their owners served, and the claims his client is pursuing.

Not all misinformation requires sophisticated tech or AI. For @craigsilverman.bsky.social's @indicator.media, I deep-dived into these accounts generated millions of views—and potentially handsome profits—with a little clever editing and an understanding of how strong emotions drive clicks.

🔎 Negli ultimi due anni le principali aziende tecnologiche hanno promesso di aiutare gli utenti a identificare i contenuti generati con intelligenza artificiale. Un'indagine di @indicator.media ha dimostrato però che le cose non sono andate come annunciato

What an incredible evening at the #CJFJTalks event The New Age of Digital Deception! A huge thank you to @craigsilverman.bsky.social of @indicator.media and CTV News’ Carolyn Jarvis for their eye-opening insights on navigating our increasingly complex media landscape. /1

TONIGHT: Don’t miss our CJF J-Talk, in partnership with @mediasmarts.bsky.social , featuring @craigsilverman.bsky.social of @indicator.media and CTV National Anchor Carolyn Jarvis on misinformation, modern verification tools and Silverman’s vision for building resilient information ecosystems. /1

We published an investigation from @kcinbk.bsky.social into YouTube vids that falsely claim to show police racially profiling and arresting a person that turns out to to be an FBI agent. They've earned over 300 million views. In response, YouTube banned 13 channels: indicator.media/p/youtube-co...

This is most literally the LOWEST HANGING FRUIT of AI detection: read (or maintain) the already-available metadata. Alexios' investigation shows that most of the platforms can't even do that.

Over the course of three weeks, Indicator published 516 posts containing AI images and videos across Instagram, LinkedIn, Pinterest, TikTok, and YouTube. Only 169 posts, or just over 30% of the total, were correctly labeled as AI-generated.

Super important work.

Shallow impersonations of Telemundo, CNN, Fox News and other media got more than 200 million views posting fake ICE arrests and bombastic AI slop about geopolitics | @indicator.media

An audit by @indicator.media suggests not only are major platforms failing to label AI images and videos as promised, some of them (Google, Facebook, Instagram) aren't even labelling stuff generated *by their own AI tools*, which they can be 100% sure are AI. It's almost like they don't give a shit.

Today on @indicator.media: A first-of-its-kind audit of AI labels on major social platforms.

Media Literacy Week is next week, and so is the CJF J-Talk The New Age of Digital Deception, in partnership with @mediasmarts.bsky.social ! Join Carolyn Jarvis as she sits with @craigsilverman.bsky.social of @indicator.media to explore his frontline experience tracking misinformation campaigns. /1

101 TikTok accounts masquerading as real news spread misinformation to millions - from @mantzarlis.com indicator.media/p/101-tiktok...

Our website is experiencing some issues and members may be unable to login. We are looking into it, thanks for your patience!

I'm looking for world's baddest researchers and experts on AI watermarks, metadata, and labeling for a big audit that's going to be published on @indicator.media on Wednesday. Who should I chat to?

Sometimes misinformation is rather blunt and unsophisticated but still a genuine hoax. 100+ fake accounts pretending to be real news found on TikTok. Piece by @mantzarlis.com indicator.media/p/101-tiktok...

"But sophistication wasn’t the point. In an ocean of content, you just need a way to make someone slow down for the three seconds necessary to get a potentially monetizable view. A recognizable logo helps with that goal." indicator.media/p/101-tiktok...

“101 accounts impersonating Telemundo, Televisa, CNN and other media outlets were spreading clickbaity falsehoods to millions on TikTok [..] The 101 accounts documented in this article had amassed a cumulative 5.5 million followers.” indicator.media/p/101-tiktok...

It will be interesting to see how the new owners of TikTok will handle scams like these.

We love our members and it’s gratifying to know that people in different fields see value in what we’re doing at @indicator.media

Bad content drives out good.

Actual (dis)honest-to-god fake news, from the looks of it commercial spammers partying like it is 2016, only on Tiktok.

NEW on @indicator.media: I found 101 accounts impersonating Telemundo, Televisa, CNN and other media outlets spreading clickbaity falsehoods to millions on TikTok. The company took them all down after I reached out.

I'm so proud that @indicator.media can serve a wide range of members: investigators, journalists, non profit leaders, researchers and tech workers. @craigsilverman.bsky.social and I wanted Indicator to be a resource for people who give a s**t about our online spaces. Turns out, a lot of people do.

Read what some of our readers have to say about Indicator: indicator.media/upgrade

The new edition of @indicator.media is out, on the topic of ad libraries, and featuring our take on recent changes in the EU. indicator.media/p/the-indica...

So, do users care about getting their notes considered 'helpful'? Cautiously, I'd say yes: Through a regression discontinuity analysis, we found that having your first note rated 'helpful' by the algorithm has a positive (5%) impact on going on to write a second note.

Today on Indicator: Is the community behind X's Community Notes cooked?

There are lots of ways to make money on social networks, but creators are increasingly being paid directly by platforms. Meta, TikTok, Snap, & X have programs that pay people based on engagement. My latest piece shows how to identify monetized accounts:

In today's @indicator.media briefing: AI Slop on Sora, on Trump's Truth Social account and in Amazon biographies. Plus: A US company hired by Israel promising "a bot-based program" to flood the zone with Israeli MFA talking points and Silicon Valley's Australian lobby group fiddles with definitions

I also tested to see whether the labels actually triggered as promised on one of my posts. They very often did not. Even just saving an AI-generated video on my laptop before uploading it to a different platform stripped the metadata labels rely on. Cropping did that too.

New on @indicator.media: Almost all major platforms have promised to label AI content on their services, but it can be hard to keep track of who said they do what and how. I collected information for 8 different platforms in one nifty table, that we will keep updating.