If you took Pinterest, mashed it together with everything annoying about Threads, and sprinkled generative AI prompts on top — that’d pretty much sum up the newly launched Meta AI site’s social feed.
So far, prompting AI chatbots — those are the questions or requests you make — has primarily been a private affair. You pull up ChatGPT, Claude, or Gemini, type in your prompt, and whatever it spits out is for your eyes only — unless you take a screenshot and terrorize the world by posting your AI experiments online. But not with the Meta AI site. Here, you can share your AI results with just two clicks.
The result is a fascinating microcosm of the human-AI experience and, specifically, how so few people know what to do with generative AI. The irony is that Meta VP of product Connor Hayes told The Verge the company added the whole social aspect to show AI newbies what they can use it for.
This almost makes AI worth it to a certain subset of irony poisoned, terminally online people. Like me. Screenshot: Meta AI
Scroll the feed, and you’ll find an odd assortment of Pinterest-like cards. The vast majority are experiments with image generation, some are simple queries, and a few feature folks experimenting with AI gotchas (e.g., how many letter Rs are there in the word strawberry?). Browse for a bit, and you’ll find the feed refresh and resurface the same few posts over and over again — much like how in the early days of Threads it would just repopulate old content. I’ve seen this user’s posts featuring food-themed fashion shows at least a dozen times. Occasionally, you find something that makes you chuckle — like this image prompt of Gary Vee yelling at an elderly man that he’s “still got time to make it on Meta AI.”
But taken altogether, the social feed feels more like a poster for all the complaints people have about AI.
Take this prompt asking Meta AI to imagine a room — any kind of room — without a clown in it. The final result is a picture of a deranged clown sitting on a living room couch. Or this one for a Jackson Pollack style illustration of cherry blossoms that is decidedly nothing reminiscent of Jackson Pollack. Or this conversation where a user asks Meta AI to help them figure out a healthy snack that won’t spike glucose. The answer isn’t wrong, but it’s also easily searchable on Google. It’s not something that proves AI searches are inherently better. The same goes for the AI images.
The comment section nails it. “Narrator: They put a clown in it.” Screenshot: Meta AI
For every Gary Vee gem, there are a dozen random AI landscapes that probably sounded cooler in the user’s head. I’ve lost count of how many posts I’ve seen asking Meta AI to make some iteration of the papal conclave. If you were trying to sell a skeptic on the power of AI, I’m not sure generating an image of a mashed potato mattress is a convincing example. After several days of browsing, I don’t think I’ve come away with new ideas of how to prompt AI either.
The public nature of the feed can feel creepy, too.
Wading into the feed can sometimes feel like eavesdropping on thoughts you weren’t meant to see or hear. This prompt, for example, feels an awful lot like I’ve wandered into a therapy session where I watch someone convince Meta AI to validate their decision to dabble with Bitcoin.
There’s also this extremely detailed image request for a “sultry Asian beauty exud[ing] bad girl energy at night” and the thirteen user attempts to get it just right. It makes you wonder how much of the feed was accidentally shared. Probably very little; you have to deliberately hit the share button, which triggers a large window alerting you that you’re about to post everything publicly. You then have to actually click a separate ‘Post’ button. Which, given some of these posts, it also feels weird that people want me to see this.
I don’t think this is the sort of prompt a person would normally make public, but it’s actually hard to accidentally share on Meta AI’s site. Screenshot: Meta AI
That you can comment on people’s results adds another interesting dynamic. Most comments I’ve seen are nice. Some are funny. But again, so far, chatting with AI has largely been a private affair. Would you prompt differently if you knew that other people would eventually see? I probably would.
Meta is just the first to add a social feed to its chatbot. OpenAI is reportedly working on its own version for ChatGPT. And Elon Musk’s Grok chatbot is now available to all X users, whether you like it or not. You can see the logic to it. Creating something shareable inevitably leads to viral trends that, may in turn, encourage people to see what AI is all about.
The trick is giving the average person — not first adopters, not tech evangelists — enough reason to stick around. I’m a naturally curious person who has fun poking at AI chatbots. Sifting through the Meta AI feed, I can find plenty of things to gawp at or pique my interest.
My mother-in-law? I’m not convinced a steady stream of surreal living rooms made of candy is what’ll finally convince her to give AI a go.
