We started an Instagram account with only AI-made reels: week 1 results
In October 2026 we opened a new Instagram account, @bakevideo, and posted only reels made in BakeVideos from one-paragraph ideas, with no filming and no editing app. In their first two days, on an account with almost no followers, our first two reels reached about 2,700 and 1,200 views, with 108 and 52 likes, a like rate of about 4% of views. That is not viral. It does show that an AI-made reel can hold strangers' attention, and it gives us a baseline to measure against. This page is the log; we will update it as the experiment runs.
Why we are doing this
We make a tool that turns an idea into a short video. The honest question every creator asks us is "will these actually get views?" A demo on our own site cannot answer that. A public account that anyone can check can. So the rules are simple:
- Every reel is made in BakeVideos, from a written idea, exactly as a paying creator would make it.
- No paid promotion, no bought followers, no engagement groups.
- We publish the numbers, including the reels that flop.
The numbers so far
Counted on 21 October 2026, two days after the first reel went up (views from Instagram Insights; likes and comments are public on each reel):
| Reel | Posted | Views | Likes | Like rate |
|---|---|---|---|---|
| If Pokémon Became Real Tomorrow (62 s) | 19 Oct | ~2,700 | 108 | 4.0% |
| Could You Survive 24 Hours With Charizard? (81 s) | 20 Oct | ~1,200 | 52 | 4.3% |
Both are animated stories: every scene AI-made in one look, with a voice, captions and music, from a one-paragraph idea.
How to read this. Views on a brand-new account come almost entirely from people who do not follow it, because that is how Instagram tests a reel: it shows it to a small group of non-followers first and widens the audience when they respond (Instagram, April 2024). So these views are strangers choosing to keep watching. A like rate around 4% is the encouraging part; a few thousand views is not a breakout, and we will not call it one.
What we think worked
- A "what if" premise. Both reels ask the viewer to imagine something impossible happening to them. The question is the hook, and it is answered only at the end, which keeps people watching.
- Characters people already love. This is the big caveat. Both reels used Pokémon, a franchise with a huge fan base, and fans watch anything about it. We cannot yet tell how much of the result came from the format and how much from the franchise.
- A whole story, not a list. Each reel moves through a day or a timeline (the first hour, the first week), so there is always a "what happens next".
What we are changing in week 2
The franchise question matters for anyone reading this. Your own channel should not depend on characters someone else owns: the rights holder can have the videos taken down, and a channel built on another company's characters is hard to monetise. So week 2 tests the same format with original characters:
- Could you survive 24 hours with a baby dragon as your pet?
- What if dinosaurs came back tomorrow?
- Could you survive 24 hours inside a giant's house?
Each link opens the studio with that idea filled in, so you can run the same test on your own account. If the original-character versions land near the same like rate, the format is doing the work. If they fall well short, the franchise was.
Alongside that we will post one reel a day for 30 days, the same video on Instagram, YouTube Shorts and TikTok, and log for each one: the idea as typed, the kind of video, its length, and at 48 hours and 7 days its views, likes, shares, saves, follows and the share of viewers who did not follow us. Shares and watch time matter more to Instagram than likes (see how the Reels algorithm works), so those are the numbers we will judge by.
What this does not prove (yet)
- Two reels is a sample of two. The full 30-day write-up will report the median reel, not only the best one.
- Views are not a business. We will also report how many followers and profile visits the reels earn.
- Every account is different. Your niche, language and posting time will change the numbers. The method is what transfers.
FAQ
Can AI-generated reels get views on Instagram? Yes. On a new account with almost no followers, our first AI-made reels reached about 2,700 and 1,200 views in their first days, shown mostly to non-followers. Views depend on whether strangers keep watching and share, not on how the video was made.
Is a 4% like rate good for Instagram Reels? Likes divided by views around 4% means a meaningful share of strangers who saw the reel chose to react. Treat it as a baseline for your own account rather than a universal benchmark, and track shares and watch time alongside it.
Does Instagram show AI reels to fewer people? Instagram's published recommendation rules focus on originality: accounts that repost other people's content without changing it lose recommendations. Reels you make yourself are original content. Instagram may add its "AI info" label to AI-made media.
Can I use famous characters like Pokémon in my reels? They are owned by their rights holders, who can have videos taken down, and they make a channel hard to monetise. Use original characters for anything you want to keep: BakeVideos keeps one consistent cast across every scene.
How long will this experiment run? Thirty days of daily posts on Instagram, YouTube Shorts and TikTok. We will update this page each week and publish the full results with every reel's numbers at the end.
Related
- How the Instagram Reels algorithm works
- What if videos: how to make them with AI
- Faceless Instagram Reels maker
- AI story video generator
- Consistent characters in AI videos