Why I'm sharing it
This method came from a social post. The author wasn't named in the copy I saved. It describes a loop for Reels that treats each batch as a small experiment, and that's what caught my attention.
In clinical data work you try something on a small group, read the results and only then decide what to change. Instagram's trial Reels let you run the same kind of check on a video before your followers ever see it.
I don't show my face or my home on camera, so I adapted the steps to work with text on screen, a voiceover and everyday clips.
What makes it work
You start from proof. Before writing anything, you find Reels in your niche that beat their creator's usual numbers and work out why people stopped scrolling. You borrow the shape of the hook and leave the words behind.
The AI gets real context: your audience, your offer, where viewers should go next, captions that worked and the clips you already have. A thin brief gets you generic videos.
You ask for finished pieces. A chat is good at ideas. An agent that can work with files on your computer, such as Claude Code, can go further and turn those ideas into files you can edit. The original post was cut off at this step, so I'm describing the idea rather than the author's exact setup.
Trial posts do the judging. Instagram's trial option shows a Reel to people who don't follow you yet. After about a day, you hand the analytics back to the AI and the next batch starts from what the numbers said.
The prompts
Step 1: Study the hooks that already work
Here are [10 to 15] Reels from my niche that got far more views than their creators usually get. For each, I've written the hook shown on screen and the view count: [your list]. For each one, tell me why people likely stopped scrolling: did it name a problem, surprise them or make them curious? Then turn each into a fill-in-the-blank pattern I can reuse without copying their words.
Step 2: Give the AI the full picture
Before you write anything, here is my context. Audience: [who they are and what they keep asking about] What I sell: [your offer] Where viewers should go after watching: [the next step you want] Captions from Reels that did well in my niche: [3 to 5 captions] Clips I have: [describe 5 to 10 clips from your camera roll] I don't appear on camera and I don't show my home, so plan around these clips, text on screen and voiceover. Tell me what you understood, then ask me about anything missing, one question at a time.
Step 3: Make a batch, not a list
Using the hook patterns and my context, write [5] Reels ready to produce. For each one, give me the hook on screen, a voiceover under [40] seconds, which of my clips goes where and for how long, the text overlays and the caption. If you can work with files, save each Reel as its own file in [folder] so I can open them in my video editor.
Step 4: Learn from the trial
Here are screenshots of my Reels analytics from the last 24 hours. Some topics did better than others. Tell me what the stronger Reels have in common and what the weaker ones have in common. Then give me 20 new angles on similar problems for my next batch.
How to use it
Do the first step by hand. Scroll your niche, save 10 to 15 standouts with the hook and the view count, and only then open the AI.
Keep your clips simple: hands at work, a walk, a coffee, street shots. Check the background of anything you film indoors.
Turn on the trial option before you post each Reel. Wait about 24 hours, run step 4 and start the next batch from its answer. Then repeat the loop: batch, trial, analytics, next batch.
Adapted from a social post whose author I couldn't identify. If you know who wrote it, tell me and I'll add their name.