Crunching the data on AI slop

Today’s Social Media advertising rates are thawing:

↗️ META: $8.27 | ↗️ TIKTOK: $2.97 | ↗️ SNAPCHAT: $17.77

In this week’s edition: 

• 💼 LinkedIn tests pre-roll video ads

• Self-serve ads are coming to ChatGPT next month 🧠

• 📊 New data: not all AI slop is bad for ads

But first…

😬 Social media’s no good, very bad week in court

The Atlantic’s CEO Nicholas Thompson — of “most interesting thing in tech” fame — sums it up best: it was a terrible week for social media platforms in court. Here’s what we’re reading:

In a landmark California trial, a jury awarded $6 million—including $3 million in punitive damages against Meta and YouTube—to a 20-year-old plaintiff who argued that social media “contributed to health issues including anxiety, depression, and body dysmorphia.” (TikTok and Snap settled before the case went to trial.) The verdict is widely expected to open the floodgates to a wave of lawsuits against social media platforms. Some 3,000 such lawsuits are pending in California, according to the Wall Street Journal.

Meanwhile, a jury in New Mexico found that Meta “misled users about the safety of its platforms and enabled the sexual exploitation of young users” in a case brought by the state’s Attorney General, who said the verdict proved that “Meta executives knew their products harmed children…and lied to the public about what they knew.” In finding that Meta violated the state’s consumer protection laws, the jury found Meta liable for $375 million in damages—a number that could go up when the case reaches the penalty phase this summer. 

In Texas, a Federal court threw out a lawsuit brought by X against advertisers, claiming the World Federation of Advertisers had somehow broken the law by allegedly organizing a boycott of the app. The suit was widely seen as a threat to force advertisers back to the platform following years of brand-safety concerns. In dismissing X’s lawsuit, the judge found the company “failed to demonstrate that it had suffered any harm under federal antitrust laws.” 

The Inside Scoop🍦: Now trending in performance media

💼 LinkedIn is testing a new tool called Creator Discovery that brings influencer marketing into Campaign Manager, matching brands with vetted B2B creators for Thought Leader Ads. It will live under the Partnerships tab. The test launches in April, no word on when it will roll out widely. 

• At launch, the test will have between 500 and 1000 vetted creators

• At present, this is just a discovery tool: brands connect with creators via email, and fees are negotiated offline

🤳 Also being tested in LinkedIn: Instream Video Ads. Hyped as a “scalable way to place pre‑roll video ads alongside high‑quality, contextually relevant publisher and creator content,” the test will be limited to targeting content by category (including Technology, AI, Business, Finance, Leadership, Lifestyle), not individual creators and publishers.  

💅 Pro Tip: New benchmark comparisons are rolling into Campaign Manager: rolling out by the end of the month, you’ll be able to compare performance against prior time periods, peer benchmarks, or self‑benchmarks directly within Campaign Manager reporting.

📺 Google’s AI answers are increasingly sending users … back to Google sources, thereby taking traffic away from the open web. New research from SE Ranking looked at answers from Google’s AI Mode and found that Google.com is now the most-cited source across nearly every category. 

• Google accounts for ~17% of AI Mode citations, more than the next several domains combined 

• ~59% of Google citations now point to its own search results, not external sites 

• It’s not just AI Mode, it’s also AI Overviews: in an earlier study, data showed Google already dominating links there too (~40%+)

👽 Reddit is becoming the last stop before purchase and the source material for AI answers. According to Digiday, brands are starting to treat community conversation as both media channel and product insight engine—with real downstream impact on commerce and AI visibility.High-intent behavior is surging, with shopping conversations up ~40% YoY and users treating Reddit as a final “sanity check” before buying. And LLMs like ChatGPT increasingly pull from Reddit, turning threads into de facto product recommendations. 

💡 Our takeaway: Campaigns that amplify real user feedback are driving both engagement and sales, but require brands to let go of the script.

🧠 ChatGPT’s ad business is only six weeks old, but it has scaled to a run-rate of $100 million per year — while serving ads to only 20% of the eligible user base, according to SearchEngineLand.

• About 600 advertisers are currently beta-testing the ChatGPT platform; according to AdWeek, CPMs are around $60, with a minimum spend of $200,000. 

• Earlier this week, SearchEngineLand reported that ChatGPT would be opening to self-service access in April, but AdWeek reported yesterday that the beta is likely to extend for at least another month or two, with ChatGPT rolling out additional pilot programs outside the U.S.

🤖 Meanwhile, ChatGPT parent company OpenAI is shuttering Sora, the TikTok-for-AI-generated-video app it launched last year. Reports indicated that after an initial “holy crap” moment, the app became a ghost town. Meanwhile, Open AI has been retrenching for a relentlessly tough fight with competitors like Anthropic and Google.

👻 Snapchat is rolling out three new ad formats: Total Snap Takeovers, Offers, and expanded Dynamic Product Ads: 

• Total Snap Takeovers: a premium placement that guarantees the first ad slot in every major app surface (e.g., Stories, Spotlight, Chat)

• Offers: a new ad unit that embeds native, Snapchat-specific promotions and discounts directly within Snap Ads

• Dynamic Product Ads have been updated with expanded commerce formats including Multi-Segment DPA, vertical carousels, product-level video ads

🧪 By the numbers: An advertiser’s guide to AI slop

The industry has spent the better part of a year dunking on “AI slop”—and rightfully so. Turns out, though, some of it sells. New research from OM Media Trials and Zefr (which we got a peek at during this week’s 4As conference in Boston) suggests the right AI environments can actually improve how brands are perceived:

• Ads next to AI satire, art, and youth content = lift in brand perception (“refreshing,” “innovative”)

• Ads next to AI spam or misinformation = immediate downside (especially for high-trust categories)

• Labeling matters: transparency improved brand opinion for ~40% of respondents

Why do we care? By some estimates, 90% of web content will be AI-generated by 2030. So the question may become: which kind can you live with in your campaign targeting?

Thanks for reading! We’ll be back next week to fill your inbox with more tips, tricks, and treats from the addressable universe.

In the meantime, stay up to date with the latest performance marketing intelligence by following us on Linkedin, Instagram, and X.

This week’s edition of the Thread was created by Jess Dashner, Anoushka Madan, and Carly Carioli.