The intersection of social media and AI has led to epic failures from OpenAI and Meta, and it's worth looking at why. How do companies with such massive reach and successful products have no idea what people actually want?

OpenAI's Sora app is the most prominent example. One early sign of how it was going was when the daughter of Dr. Martin Luther King Jr. had to publicly tell them, "Please stop."

Deepfakes were part of the problem. Even if OpenAI wound up preventing users from putting new words in Dr. King's mouth, you could create deepfakes of many others. I created ads of OpenAI CEO Sam Altman endorsing AI Marketers Guild. They weren’t that deep, but they were fake.

That points to an even bigger issue. Most of the content was boring. Maybe at first it was amusing the first time you saw a colleague as Batman, or Abraham Lincoln riding a Segway. By the third or fifth one you saw from a friend who was way too proud of their creativity (that friend might have been me), you were ready to block them. 

So how could Meta watch one of its biggest rivals commit all these unforced errors and go, "We should do something like that”?

This month, Meta killed an AI feature on Instagram where others could remix your posts with AI. Apparently, it’s a bad idea to opt everyone into that by default.

Here's how Meta described the feature when it launched: "Meta is also launching the ability to @mention friends’ public Instagram accounts in Meta AI and generate creative AI images featuring them, such as personalized birthday cards, group trip memes, or playful edits between friends. It's an upgrade to what people can create with AI features at Meta, making it more personal, fun, and social."

I somehow missed all the coverage of its launch. But the backlash was as swift as it was effective. The first time I learned about it was from a Threads post from Kara Swisher, who was teaching followers how to disable the feature. Many others published guides, and before most Instagram users could follow the instructions, the feature was dead. 

It took OpenAI 208 days to kill the Sora app. Meta killed that Instagram AI feature in just three.

How did Meta wind up launching a feature that sounds comically bad, even with the best spin from its undoubtedly talented marketing comms team? 

Is there no one around to tell folks like Mark Zuckerberg that such ideas are terrible? At what point do you become rich enough that you only surround yourself with sycophants? I wouldn't mind being that kind of rich, but if I can be slightly less rich to avoid the sycophancy, I'm willing to go right up to that limit. 

Granted, why even pay for sycophants? Who needs that when you can ask AI —perhaps the algorithms you've created — if your ideas are good? 

AI will be sure to tell you that you're the fairest one of all. The fairy tale got it right. 

I tested this theory. In one of my Gemini accounts, I asked it to come up with the worst ideas I could vibe code. And it did, including a "LinkedIn Thought-Leader Auto-Responder" that it said was so bad that “you will lose all your professional connections within 48 hours." 

I went into a different Gemini account so it wouldn't know my history, and it's one that knows way less about me. I asked Gemini what it thought of this idea. It said, and I'm quoting it verbatim, "Build it immediately." It did see the potential for this to be performance art, but it was still convincing me, in great detail, to build the exact thing that the other Gemini instance said would destroy my professional reputation. 

So, maybe that's how it happens.

The bright side? People are speaking out. The novelty of AI wearing off helps our judgment. And the next time one of these tech giants rolls out a tone-deaf feature or product, they might break Meta's record for how quickly it's killed. 

— David Berkowitz, Chief Community Officer, Marketecture Media

Who: CMOs, Agency Leaders, Brand Marketers, Regulatory Compliance Officers

What: The ANA, 4As, and IAB teamed up to form a joint governing body focused on artificial intelligence across advertising. The council will establish shared industry definitions, retail media standards, and workforce development frameworks for brands and agencies.

Why it matters: Marketers are getting unified guardrails for AI adoption and vendor evaluation instead of navigating fragmented guidelines from competing trade groups.

(ANA)

Who: Insights, Agency Strategy, Brand Planning, CMOs

What: Agencies are increasingly combining synthetic AI audiences with traditional consumer research instead of replacing human respondents outright. The goal is to speed up testing while keeping findings grounded in real customer behavior.

Why it matters: This feels like where AI research is actually landing, making teams faster instead of trying to replace traditional research.

(Digiday)

Who: Chief Creative Officers, Brand Strategists, Social Media Directors, Content Leads

What: A global study of 400 marketing leaders found that while 88% of marketers have increased creative output using AI, only 45% see a meaningful improvement in quality. The research highlights that static demographic inputs are bottlenecking AI tools and recommends feeding real-time community signals into AI creative briefs.

Why it matters: Scaling ad volume with AI yields diminishing returns unless teams upgrade their prompt inputs with actual community and behavioral data.

(Social Media Today)

Liking AI Brief? The Refresh is a weekly recap of all the most important advertising news, written by AdTechGod.

Thanks for reading this week’s AI BRIEF by Marketecture. 

Marketecture Media is committed to bringing you the latest news, opinions, and community updates. 

If you are interested in advertising on this newsletter, our podcasts, The Advertising Forum, or in our Slack community, you can contact us here.