ChatGPT vs Claude vs Gemini: Which AI Writes the Best Instagram Reel Script?

If you're a content creator or you have something to market on Instagram, you've probably wondered: can AI actually write a good Reel script? Does it understand what makes a viral hook? I gave ChatGPT, Claude, and Gemini the same creative brief and found out.

The Experiment

I gave all three AIs a specific creative brief: a fitness account promoting a 30-day home workout challenge for beginners who keep quitting. Who can write the best script?

(Note: ChatGPT 5.2, Claude Sonnet 4.6, Gemini 3 Pro)

The Prompt

"Write an Instagram Reel script for a fitness account promoting a 30-day home workout challenge. Target audience is beginners who keep quitting. The first 3 seconds need a hook that stops the scroll. Keep it under 45 seconds. The goal is to get viewers to follow the account to join the challenge."

Scoring Criteria

I used four dimensions that reflect what actually makes a Reel work:

1. Hook Strength: would the first 3 seconds actually stop someone from scrolling?

   
   

2. Persuasion: does it make you want to follow and join the challenge?

   
   

3. Script Structure: does the content flow naturally? Does each beat build on the last? Is it tight, with no wasted time?

   
   

4. Human-sounding: does it sound like a real creator, or like AI wrote a script?

The Results

(Left to Right: ChatGPT, Claude, Gemini)

1. Hook Strength

Gemini — 4/5: Best hook. The hook is specific and immediately relatable. Makes the viewer think "that's me." Lost a point because it takes longer than 3 seconds to land. In a real Reel, you'd need to tighten it.

ChatGPT — 3/5: The hook is in the right direction and calls out the audience, but "Stop Scrolling" is overused. A question would have been stronger than a command.

Claude — 2/5: The hook is bold, but you have no idea "Stop Quitting" is about fitness. Could be about anything. The viewer has no reason to stay.

2. Persuasion

Gemini — 3/5: Identifies the problem well. It makes people feel understood rather than lectured. But it never explains what makes THIS particular challenge different from every other one people quit.

ChatGPT — 2/5: "What if you didn't need motivation… just 10 minutes a day?" sounds nice but is shallow. Every fitness account says this. The emotional hook is weak and doesn’t give enough reason to follow.

Claude — 2/5: "Most beginner programs want you in the gym 5 days a week." But is that really why people quit? The problem isn't clearly identified, so the solution doesn't land. Then it claims "thousands of people just like you finished this," which is fabricated social proof for a challenge that doesn't exist yet. Kills trust instantly.

3. Script Structure

Gemini — 5/5: Best flow. Problem → relatable scenes → reframe → solution → CTA. Each beat connects logically to the next. Visual suggestions like "failing a pushup, checking the fridge" are fun and easy to picture. Every section earns its time. Tightest script of the three.

ChatGPT — 2.5/5: First two beats work. Hook into relatable struggle is fine. But it falls apart after the mindset shift. The introduction, emotional hook, and CTA all feel disconnected, like separate ideas stitched together. Also not tight enough, the "emotional hook" scene adds length without adding value.

Claude — 2/5: Follows a marketing framework: problem → solution → social proof → CTA. Reads like a funnel, not a story. The fake social proof section wastes precious seconds on something that hurts credibility. The loosest script of the three, trying to cover too many persuasion tactics in 43 seconds.

4. Does It Sound Human?

Gemini — 3/5: "The couch wins" and "people who hate working out" feel like a real creator talking. But the closing lines slip into generic motivational poster language.

ChatGPT — 3/5: "Boom, you're done" is natural but the rest leans generic motivational language as well.

Claude — 2.5/5: "Just you and the floor" is a nice line. But the fake social proof and "prove yourself wrong" sound like ad copy. The overall tone feels like a brand, not a person.

Final Scores

1. Gemini: 15/20

2. ChatGPT: 10.5/20

3. Claude: 8.5/20

Verdict

Gemini wins. Best hook, most relatable, best flow. It understood the actual viewer better than the other two. But even Gemini's script wasn't something you could film as-is. The solution section needed more work.

ChatGPT was bland. The first two beats had potential, but the script fell apart halfway through. It stitched together motivational phrases without building a real arc.

Claude had the biggest miss. It went into marketing mode with bold statements, fabricated social proof, and sales tactics. It wrote an ad, not a Reel script. A real creator would never claim "thousands finished this" for a challenge that hasn't launched yet.

The Bigger Takeaway

None of them wrote a Reel script I'd post without heavy editing. The biggest shared weakness? They all described the challenge but none gave a compelling reason why THIS challenge would be different from every other one people quit. They can structure a script, but they can't create the core insight that makes someone actually care. That's still a human skill.

Practical Tip

AI can give you a decent first draft of a script structure, but don't expect it to nail the hook or the persuasion. Write your hook yourself. That's the part that matters most and the part AI is weakest at. Use AI to fill in the middle sections and tighten your wording.

Each AI approaches creative tasks very differently, which means each one will give you a completely different direction to work with. Try using Agora. You can ask all three at once and use their different angles to brainstorm a stronger script than any single AI would give you.