AI's Influencer Problem - by Sarah Gibbons

Future Problems

SubscribeSign in

AI's Influencer Problem<br>The major AI labs have a rare chance to shape AI's adoption and meaning, and most are spending it on outdated marketing.

Sarah Gibbons<br>Apr 21, 2026

20

Share

Last Tuesday, Reese Witherspoon showed up on my Instagram feed telling me “it’s time we learn about AI before it gets too far ahead of us.” She shared how she was with her own book club and only 3 of 10 women use AI (and of those three, only one felt like they actually knew what they were doing). She framed AI literacy as future-proofing.<br>The backlash was immediate. Roxane Gay commented: “Absolutely not.” Writers and illustrators pointed out that the AI Reese was endorsing is trained on the unauthorized work of authors (the people Reese’s famous Book Club was built to serve). One writer called generative AI as an act of feminism “vile.”

Reese's reel on my feed last Tuesday, with 167K likes and 13.4K comments.<br>Last Thursday, Ashley Striblet (of Early Insights Club) posted a screenshot of a different influencer moment: Carly Weinstein’s “dinner party with @claude.ai.” A quintessential influencer brand dinner: place cards, conversation starters, and vibey lighting. Paraphrasing Ashley: it would be nice to see AI labs make meaningful changes to how they develop AI, rather than manufacture consent through influencers. I’d seen Carly’s post earlier and had the same feeling.

From Carly's story. Also at Claude's brand dinner: Deepa ('helping you get happy, healthy, hot, and rich'), Brittany Leighball ('a real-life rom-com'), Rawan ('career, travel, personal finance').<br>Back in February, Microsoft Copilot took a group of New York fashion influencers to the Super Bowl on an influencer brand trip (they had nosebleed seats). The Copilot use cases that came out of the trip, documented on TikTok, were “hey copilot how do you cure a hangover“ and “how to get a stain out”. To recap, a multi-trillion-dollar company sent a powerful, emerging technology to one of the largest cultural events in the country…to ask about hangover cures.

Greta's TikToks from the Copilot-sponsored Super Bowl weekend.<br>Short-form social content skews heavily to Gen Z, yet Gen Z is the demographic the most skeptical of AI. Gallup’s February survey of 14–29-year-olds shows excitement down 14 points year over year (to 22%), anger up 9 (to 31%), and only 18% feeling hopeful. Among employed Gen Z, 48% say the risks of AI at work outweigh the benefits—more than three times the 15% who say the reverse. The lower the use and trust, the higher the hostility.<br>An aside, I actually don’t disagree with Reese’s main point. A Federal Reserve Bank of NY study found that 50% of men used generative AI in the past year, compared with about 33% of women. The gap carries into the workplace: LeanIn.Org’s 2026 survey found 33% of men use AI daily or constantly at work, versus 27% of women. And a meta-analysis of 18 studies found a roughly 25% gender gap in adoption overall. AI’s value can help its users succeed at work, which maps to pay, promotions, and raises.<br>It’s easy to see all this as outside noise to our product/ux/strategy bubble. However, what the average person does with AI (or refuses to) is what steers the money. That money decides what this technology becomes, and who it ends up serving. My goal is just to noodle out loud and sketch what alternative could look like.

Thanks for reading Future Problems! Subscribe for free to receive new posts.

Subscribe

This Matters, and It Could Look Different

The influencer playbook we know today was built to sell goods. It emerged from beauty and fashion in the mid-2010s, designed around products you can hold up to a camera, style into an apartment, and wish for from a distance. AI is shaped differently. Its value lives entirely in what a human does with it. It cannot be held up to the camera. So far, the major AI labs have grabbed onto the old playbooks anyway.<br>The underlying challenge, for design and marketing alike, is that AI is a product with a virtually endless set of uses, for nearly anyone, across an enormous range of fluency. It’s hard to market a technology clearly when you haven’t decided who you’re designing it for.<br>Claude is what I use most, so the rest of this leans toward Anthropic, though most of it could be said of any majorAI lab. Anthropic has positioned itself, in my eyes, as the mainstream “thinking person’s AI.” Here’s what marketing could look like if it took the technology’s own promise—that AI can help humans do more intrinsically human things—as its core thread.<br>Idea 1: Fund domain experts, not influencers.

There is a population of people in this country whose authority is unimpeachable inside their own corner of the world, even if it’s entirely unknown outside it. For example, a labor lawyer well-known in her region, a history teacher in his twenty-fifth year teaching, or a social worker with a strong reputation. These people have...