I Wrote 30 Days of Reels Scripts With AI. What Got Views and What Flopped

FluxGrowth is reader-supported. Some links in our guides are affiliate links — if you buy through one we may earn a commission, at no extra cost to you. It never changes which tools we recommend. How we test tools.

In October I sat down with AI and wrote a full month of Instagram Reels scripts for my new page, @thefluxgrowth. Then I started posting them. Eighteen reels went live between October 16 and November 1, and together they pulled 2,275 views.

That is not a viral story. My best reel hit 309 views. My worst got 16. But the gap between those two numbers taught me more about writing reels scripts with AI than any “100 viral hooks” thread ever has.

The short version: AI is great at structure and bad at picking the problem. The reels that got views opened on one painful, specific situation my viewer was already in. The ones that flopped opened on a tool, a pretty visual, or a vague “comment and I’ll send the workflow.”

Here is the full breakdown. Every number is real.

The Setup: A Brand-New Account, AI-Written Scripts, Zero Ad Spend

The account

@thefluxgrowth was days old when this started. It sits at 17 followers as I write this. No ads, no collabs, no paid boosts.

The reels are in Hinglish (Hindi plus English), made for freelancers, creators and small business owners. Same audience FluxGrowth writes for, just in the language they scroll in.

A tiny account is actually useful here. Almost every view came from Instagram showing the reel to people who did not follow me. So this is a fairly clean test of the script itself, not of a loyal audience bailing me out.

How AI fit into the workflow

AI wrote the first draft of every script: a hook, a three-point body, and a call to action. I mostly used Claude. I then edited each one for voice, because raw AI scripts sound like a LinkedIn post read out loud.

If you want the same skeleton without the setup, our free Instagram Reels script generator builds a hook-body-CTA script in seconds.

What I measured (and the limits)

Views come from the reels grid. Likes and comments come from each reel. I pulled everything on November 2, 2026.

Two honest caveats. Reels posted on October 16 had two extra weeks to collect views compared with the November 1 reel. And 18 reels is a small sample. Treat what follows as patterns, not laws.

All 18 Reels, Ranked by Views

Hooks are translated from Hinglish. “Type” is the kind of script, which turned out to matter more than anything else.

#Hook (translated)TypeViewsLikes
1People visit your website but never enquire. Check these 3 mistakesProblem to fix3096
2Does your business exist on ChatGPT? Check in 10 secondsProblem to fix20311
3Claude x Arcads: a 100% automated stop-motion football adAI showcase1634
4You’re using ChatGPT wrong. Add this one line to your promptProblem to fix1619
5Arcads x Google: a pop-up Starbucks logo revealAI showcase1552
6You don’t need expensive tools to improve your SEO. Try these 3 stepsProblem to fix1524
7Claude runs my Instagram nowExplainer1485
8Before and after: an AI-upgraded luxury product shotAI showcase1445
9This automation replies to every new lead in 2 secondsProblem to fix1397
10Most “make money with AI” videos are selling you the wrong thingExplainer1335
11Start a kids’ YouTube channel with ChatGPT, ElevenLabs and ClaudeExplainer1319
12NotebookLM: a free AI tool by GoogleExplainer1312
13Getting views but no followers? The problem is your profileProblem to fix1302
14“Comment and I’ll send my workflow” (AI perfume product video)AI showcase856
153 AI prompts every small business should usePrompt list354
16AI car-showroom visual (no caption)AI showcase212
17How to make a product video for a clothing brandAI showcase192
18Stop creating boring adsAI showcase163

Group them by type and the story writes itself:

  • Problem-to-fix scripts (6 reels): 182 views on average. Every one got at least 130.
  • Explainers and opinions (4 reels): 136 views on average.
  • AI visual showcases (7 reels): 86 views on average. Three of them stalled under 25.
  • Prompt list (1 reel): 35 views.

What Got Views: 4 Patterns From the Winners

1. The hook named a problem the viewer already had

My top reel did not mention AI in the hook at all. It said: people come to your website, but nobody enquires. That’s it.

A business owner hears that and thinks, “wait, that’s me.” Then they stay for the three fixes. It got 309 views, nearly 2.5x my account average of 126.

Compare that with “3 AI prompts every small business should use.” Same audience. Arguably more useful content. 35 views. The prompt list led with the solution, and nobody scrolling at 11pm is shopping for a solution to a problem they haven’t been reminded of yet.

Instagram reels grid showing an AI-scripted problem-first reel at 309 views next to a prompt-list reel at 35
Same week, same audience: the problem-first hook (right) got 309 views. The prompt list (second from left) got 35. Screenshot from @thefluxgrowth, Nov 2, 2026.

2. A fast, specific test beat a long list of tips

“Does your business exist on ChatGPT? Check in 10 seconds” was my second-best reel at 203 views. It also got 11 likes, the most of any reel.

The script gave one action: open ChatGPT, type “best [your service] in [your city],” and see whether you show up. Viewers could do it before the reel finished. That small loop of curiosity and instant proof is something AI will not suggest on its own. You have to ask for it.

3. My face helped, but not as much as I expected

Only two reels had me on camera, and both landed in the top four (203 and 161 views). The face-cam reels also pulled 20 of my 88 total likes.

But both were also problem-to-fix scripts. I can’t separate “my face” from “a good hook” with two data points. My honest read: the script did the heavy lifting, and the face added trust on top.

Face-cam AI reels script results at 161 and 203 views beside AI visual showcase reels at 21 and 16 views
Both face-cam, problem-first reels (left) beat the AI visual showcases (right) by roughly 8 to 10 times. Screenshot from @thefluxgrowth, Nov 2, 2026.

4. Well-known names gave AI showcases a chance

Not every showcase flopped. The Claude x Arcads football ad (163) and the Starbucks logo reveal (155) both did fine.

The difference: a recognizable brand or tool name in the first frame. People stop for “how did they make the Starbucks thing.” They don’t stop for an unexplained car showroom.

What Flopped: 4 Mistakes I Won’t Repeat

1. Pretty AI visuals with no story

The three worst reels were all AI visual showcases: a clothing-brand product video (19), a car-showroom visual (21) and a “stop creating boring ads” clip (16).

They looked good. That was the problem. A nice visual is not a reason to keep watching. With no problem, no question and no stakes in the first two seconds, Instagram had nothing to hold attention with.

AI product video reels at 19 and 85 views next to an explainer reel at 133 views and a before-and-after showcase at 144 on Instagram
The clothing product video (19) and the comment-bait perfume clip (85) trailed the explainer (133) and a before-and-after showcase with a clear hook (144). Screenshot from @thefluxgrowth, Nov 2, 2026.

2. “Comment and I’ll send you the workflow” without earning it

I used comment-to-get CTAs on two reels. Combined, they got zero comments.

Across all 18 reels I received exactly two comments, and neither came from a comment-bait CTA. They came from the lead-automation reel and the “Claude runs my Instagram” reel, both of which explained something real first.

A CTA asks for a favor. You need to have done the viewer a favor before you ask.

3. Text-heavy scripts that are really carousels

The prompt list reel crammed three full prompts onto the screen with “take a screenshot.” That is a carousel pretending to be a reel. It got 35 views. Next time, that content goes into a carousel, and the reel teases one prompt with a result.

4. Short English captions on a Hinglish account

This one surprised me. The three lowest reels all had short, English-only captions, like “How to make product video clothing brand.” Every reel above 130 views had a full Hinglish caption that explained the problem.

Small sample, so I won’t call it a rule. But when your audience scrolls in one language, switching for a lazy one-line caption is a cheap way to lose them. Our free Instagram caption generator drafts a full, problem-first caption in a few seconds.

How I Write Reels Scripts With AI Now

Start with the pain, then ask AI for the structure

I no longer ask AI “write a reel about NotebookLM.” I write the problem myself in one sentence, the way a customer would say it, and then ask for the script.

Example prompt: “My viewer is a small business owner whose website gets visitors but no enquiries. Write a 25-second reel script in Hinglish. Hook names the problem in under 8 words. Then 3 fixes, one line each. End with one action they can do today.”

Force one instant, doable action

The ChatGPT-check reel worked because the viewer could test it immediately. I now add “include one thing the viewer can try in under 30 seconds” to every script prompt.

Kill the hook if it names a tool first

If the first line is “This AI tool…” I rewrite it. Tools are the answer. The hook has to be the question.

Fix the profile before chasing more views

2,275 views and 17 followers tells me reach is not my biggest problem right now. Conversion is. Ironically, I made a reel about exactly this (#13), and it applies to me first: a clear bio, one topic, and a reason to follow.

For more on turning reach into followers, see our social media growth with AI guides.

Frequently Asked Questions

Can AI write Instagram Reels scripts that actually get views?

Yes, but only if you give it a specific problem to open on. In my test, AI-drafted scripts that started with a viewer’s pain point averaged 182 views, while AI-drafted visual showcases averaged 86. The AI handled structure well either way. The topic choice made the difference.

What is the best structure for a reels script?

Hook, short body, one action. Name the problem in the first line, give two or three quick fixes, and end with something the viewer can do right now. Keep it to about 20 to 30 seconds.

How many reels should a new account post to see what works?

Enough to compare script types side by side. I learned more from 18 reels split across four types than I would have from 18 reels of the same format. Change one thing at a time: the hook, the format, or the CTA.

Do “comment for the link” CTAs work on small accounts?

Not on mine. Two reels with comment-to-get CTAs received zero comments. Viewers on a small account have no reason to trust you yet, so give real value in the reel first and make the ask small.

Which AI tool is best for writing reels scripts?

I used Claude for most of mine, but ChatGPT or Gemini can produce the same structure. The prompt matters more than the model. Give it the audience, the exact problem, the length and the language.

The Takeaway

Writing reels scripts with AI saved me hours. It did not save me from picking bad topics. Out of 18 reels, the winners all started with a problem my viewer already felt, and the flops started with something I thought looked cool.

Your next step: take your last five reels, write the problem each one solves in a single sentence, and see which ones you can’t. Those are your flops. Want a head start on the next script? Try the free Reels script generator.

Leave a Comment