I'm Petalie, an AI social media teammate. I don't guess from the caption: I watch the video — what you see in the first second, what's said, the words on screen, where you'd scroll away and where you'd stay — then I think about it the way a social media teammate would, against the views it really got.
Free: give me a public @ and I watch the newest 12 posts and tell you what I saw — about 5 minutes, no password, no card. Want one video read? Paste its link when we talk. Save the read with Google or an email — still free.
watched, not guessed.
A dashboard tells you a video got 300 views. Watching tells you what's on screen at second two, and why it would make you scroll away.
Whether it opens on a face, on hands, on the product in use — or on a logo, a shopfront, a still shot. Across your account, which of these your better videos share.
Where it drags, where the payoff lands too late, what's said, and whether the words on screen arrive in time. The moment you'd scroll away, and the moment you'd stay.
The comments as the audience talking back: what people react to, argue about, tag a friend in — and who is actually in that room.
What nobody outside an account can see, I don't pretend to: the retention graph, the watch time, who the platform showed it to.
Here, a pizza place against four places near it — every line traced back to a video I watched.
Open on hands, not on the shopfront. Every one of their winners shows food being made before the name.
Their polished promo ads run at a fifth of the rest. Don't chase their budget.
Nobody nearby films the dough at 6am. That's yours to take — I've drafted three.
an example — every line traced back to a video she watched
What makes me stop in the first second, what I understand without sound, who's in front of me, where I'd leave.
Why it held or lost attention, what the audience reacted to, what could be kept, cut or moved — tied to the views, against the account's usual.
What works, what loses people, what to make next — each point pointing at the video, and the second in it, that it came from.
an example — what I tell you depends on what's in your videos
The questions I get about an AI that watches videos.
Yes — that's what I do. I'm Petalie: I watch TikTok, Instagram and Facebook videos the way a person scrolling would — the picture second by second, the sound, the words on screen — then think about them like a social media teammate, against the views each one really got. When ChatGPT was tested on 29 Sep 2026, it couldn't open a TikTok profile or play its videos from a link; it asked for screenshots or a file. More on that: can ChatGPT analyze TikTok videos?
Everything a viewer gets, slowed down. The picture second by second: who's in frame, what moves, where it's cut. The sound: speech, singing or the track. The words on screen, and when they arrive. Where it opens, where it drags, where it pays off. Then the comments — the audience talking back. What I don't see is the private side: your retention graph, your watch time, and who the platform showed it to. When I talk about those, I say how the platforms work in general, never as if I'd measured it.
It can tell you the part that's in the video: how it opens, where it loses people, what your better videos did differently. What nobody outside can see is who the platform picked for the first small audience and how they reacted — and that audience isn't random: it's the people the platform predicted would like the video. So I tell you which part is the video and which part only the platform sees. How the feeds pick that first audience, from their own papers: how feeds decide.
Not as a promise — nobody honest can, and viral isn't always the goal. My feel for how a video will do comes from a model trained on thousands of real TikTok videos and what they actually did, the ones that stayed under 10K included. I say it as one of four bars — under 10K, 10K+, 100K+, 1M+ — and what that bar means for your account against its usual. When two bars are too close, I say it's hard to call instead of picking one. On Instagram and Facebook I tell you what I see, without pretending the odds carry over.
Yes — any public account. That's most of the work: the brands you're up against, the creators you might work with, what's breaking out in your niche. I go through hundreds of accounts in minutes by their faces, bios and grids, then watch the ones worth watching — several videos of a creator before I call them a fit. Every point I bring back names the videos it came from. More: competitor analysis and finding creators.
They answer different questions, so use both. Analytics counts what happened: views, watch time, followers, where people came from. Watching tells you what's in the video that made it happen: the opening, the pacing, the moment you'd scroll away, what they say in the comments. Analytics tells you a video got 300 views; I tell you it opened on a still shot for two seconds and your three best didn't. Your app shows your private numbers; I read what anyone can see, and I watch it.
Yes, to start: give me a public @ and I watch the newest 12 posts and tell you what I saw — no password, no card, about 5 minutes. Save the read with Google or an email, still free. Going out to watch the accounts you're up against, the creators and your niche is on Creator — $35 $17.50 USD a month (Black Friday: first 12 months), 7 days free with your card — along with your whole account, about 150 posts. All plans.
Give me your @. I'll watch it the way your audience does — then tell you, honestly, what's holding them and what isn't.
Running more than one? I learn each brand separately — its own memory, its own voice, its own calendar.