Your next content idea might already be hiding in your best videos. Use ChatGPT to study your niche, find the posts that brought in followers, and turn those patterns into a batch of original scripts you can actually film.
- Study the hooks, topics and formats in your niche
- Find the videos that actually earned followers
- Turn your strongest patterns into 30 original scripts
- Get the full prompts, worked example and testing plan
- Keep improving with your own results
Start Here: Views Are Only Part of the Story
The plan · 4 steps
- 1Give ChatGPT your niche, audience and examples.
- 2Export your Instagram content data and check what it contains.
- 3Find your strongest follower-generating videos and study their scripts.
- 4Create 30 new scripts, film a small batch, and review the results.
Before you start
Tools
ChatGPT with file uploads available, plus access to your Instagram content insights or Meta Business Suite.
Materials
Your available content data and the transcripts or opening lines from your videos.
Time
Set aside about an hour for the first pass, then repeat the review as you publish.
A video can get a lot of attention without giving people a reason to follow you. This guide helps you find the topics and formats that do both.
A hook is the opening that earns attention. A format is the structure, such as a demonstration, comparison or short story. A topic is what the video is about. Your payoff is what the viewer learns or sees by the end.
You will build a niche playbook, a ranked list of your own videos, and 30 scripts based on the most useful patterns. The prompts work together in one conversation. Save the finished analysis and scripts outside the chat so you can use them again.
No prompt can promise a viral result. The advantage is making your next creative decision with evidence from your own audience.
Maverick's tip
If you are new and have no analytics, start with niche examples, use the beginner prompt later in this guide, and collect your own results as you go.
Give ChatGPT the Right Brief
Open ChatGPT and start a conversation. Replace the brackets, then paste this full setup prompt. You can give rough answers. A clear audience and a specific reason to follow you matter more than perfect wording.
Act as my viral content strategist for Instagram short-form video. My niche is [niche]. My audience is [who I help]. I want people to follow me for [specific ongoing value]. My usual videos are [length and format]. My tone is [description]. I can realistically film [number] videos per week with [equipment, locations and time]. Help me build a playbook from current niche examples and my own content data, then turn the strongest patterns into 30 original scripts. Work in stages: research examples, audit my data, identify follower-generating videos, analyze their hooks and structure, then write scripts. Do not skip straight to generic ideas. Use only material you can actually access. Tell me when you cannot read a link or view a video. Ask for a transcript, opening lines, screenshots or a description instead. Never invent metrics, quotes, personal experiences or results. Keep observed facts separate from explanations that are hypotheses. Do not call a new script proven just because its structure resembles an older winner. Start by asking up to five questions that would materially improve the strategy. Then give me a short list of the files and examples to provide next.
Give it something real to work with
For example: "I help busy parents make inexpensive dinners. Most videos are 30 to 45 seconds, filmed in my kitchen. I want followers who come back for practical weeknight meals." This gives the strategist a much clearer job than "make me famous."
Maverick's tip
Include three examples of your own writing or video transcripts when you want scripts that sound like you. Tell ChatGPT which phrases you would never say.
Study What Is Working in Your Niche
Gather a starting sample of 10 to 15 relevant videos from several creators. Include recent strong performers and a few typical posts from those same accounts. Write down the link, publication date, visible view count and date you checked it.
Paste transcripts or exact opening lines where possible. A link by itself may not give ChatGPT access to the video's words or visuals. Describe the first shot, on-screen text and final payoff if those are important.
Study the niche examples I provide below. If current web research is available, use it to add relevant recent examples and link to them. Label my supplied examples separately from anything you find. If you cannot access the material, ask me for it. For each example, record the creator, URL, date posted, date metrics were observed, visible metrics, exact hook when available, opening visual, topic, audience problem, format, payoff and call to action. Label missing information as unknown. Compare strong performers with typical posts from the same creators where data exists. Do not describe this sample as every top video in the niche, and do not infer competitors' follower conversions from views. Give me five recurring patterns, a short explanation of why each might work, and three original ways I could adapt each pattern for my audience. Separate observed patterns from hypotheses about causes. Preserve the structure without copying another creator's wording or personal story. Examples: [paste links, transcripts and notes]
The playbook you want back
Useful: "The opening shows the finished dinner, names a familiar time constraint, then proves the recipe in three visible steps." Too vague: "Use strong hooks and be engaging."
Maverick's tip
A huge view count on a huge account is not automatically an unusual result. Look for videos that stand out compared with that creator's other similar posts.
Get Your Instagram Data
Open Meta Business Suite and sign in with access to the Instagram account you want to analyze.
Export your video data
- Click Content in the left sidebar.
- Click Export data in the top-right corner.
- Choose Instagram, then select your Instagram account.
- Choose your date range. Go back to when you started the account if that history is available. Otherwise, use the earliest available date and note the coverage.
- Export the content data as a spreadsheet or CSV if offered. You want rows for individual videos. If you need multiple date ranges, keep each file so ChatGPT can check for duplicates.
- Open the downloaded file and check its column names and content rows. Save the original unchanged, along with the export date and reporting period. Keep Instagram data separate from Facebook data.
The date range for when posts were published may differ from the period over which their metrics were counted. Keep any description of that period with your file so ChatGPT can check that the numbers are comparable.
Useful fields to collect
| Field | Why it matters |
|---|---|
| Post ID or URL, publication date | Identifies the video and helps prevent duplicates. |
| Views, with the original metric label | Measures viewing activity using that report's definition. |
| Follows attributed to that video, if available | Lets you calculate recorded follows per view. |
| Reach, saves, shares, watch time, if available | Adds context without replacing follows. |
| Paid, boosted, trial or collaboration status | Helps separate content with different distribution. |
| Transcript, first line, opening visual | Lets ChatGPT explain the creative pattern. |
If follows are absent from the export, check the individual video's available insights. Record them only when shown, together with views and the observation date. Leave unavailable numbers blank. Do not divide account-wide follower growth among individual videos.
Maverick's tip
You do not need your password, direct messages or follower contact information for this analysis. Upload the content metrics and creative examples you actually need.
Check the File Before Ranking It
Attach the export in ChatGPT using its file upload control. CSV and Excel files are supported for data analysis. If you have several exports, attach them with clear filenames and explain the dates they cover.
Ask ChatGPT to inspect the files first. A neat-looking ranking is not useful if it mixes Facebook views with Instagram follows or treats missing data as zero.
Audit the attached Instagram content data before analyzing performance. List each file, its row count, columns, account/platform and apparent date coverage. Show how you map the original columns to post ID or URL, publication date, views, per-video follows, reach and any other available metrics. Do not silently rename reach, plays or impressions as views. Check for duplicate post IDs, overlapping exports, totals rows, missing values, zero denominators, number-format problems and mixed reporting periods. Keep a record of excluded rows. For duplicate snapshots, do not add their cumulative metrics together. Explain which snapshot you keep and why. Confirm whether follows are genuinely attributed to individual videos and whether views and follows use compatible periods and platform scope. If that is unclear, stop the conversion ranking and ask for the missing information. Keep missing values distinct from zero. Separate paid or boosted content, trial videos and other unlike groups when identified. Mark unknown status as unknown. Show five cleaned sample rows and summarize what analysis is and is not supported. Ask me to confirm any material assumptions before proceeding.
Check three rows yourself
Compare three videos with the original report. Confirm the IDs, views, follows and dates. If ChatGPT used the wrong column, correct it here before asking it to explain why a video succeeded.
Maverick's tip
Keep the original export and the cleaned version. A clean file should make the changes traceable, not hide them.
Find the Videos That Earned Followers
When comparable per-video follows and views are available, calculate follows divided by views, multiplied by 100. You can also express the same result as follows per 1,000 views.
A worked example
These numbers are fictional and show how the calculation works.
| Video | Views, follows and conversion |
|---|---|
| A: Broad entertaining clip | 1,000,000 views. 500 follows. 0.05%, or 0.5 follows per 1,000 views. |
| B: Useful niche tutorial | 100,000 views. 800 follows. 0.8%, or 8 follows per 1,000 views. |
| What the comparison shows | B has fewer views, but more recorded follows and a higher follows-per-view rate. |
This ratio describes the recorded metrics. It is not the percentage of unique viewers who followed, and it does not establish that the hook caused the result. Always keep total follows alongside the rate.
Using the audited data and confirmed column mapping, rank my videos by total attributed follows and by follows per 1,000 views. Also show views and conversion percentage. Use follows / views * 100 for the percentage and follows / views * 1000 for follows per 1,000 views. Show the calculation for three sample rows. Return N/A when a required metric is missing, zero-denominator or incompatible. Do not estimate missing follows. Compare videos with similar time to accumulate results. Where equal-age snapshots are unavailable, show age and the limitation. Keep paid, trial and unknown-status groups visible and separate where appropriate. Flag tiny samples and very recent posts. Propose a minimum-view threshold based on this account's distribution, explain it, and show how the ranking changes with and without it. Do not use a universal cutoff as a fact. Return the strongest videos, high-view/low-conversion videos, and promising lower-view videos. Include IDs or URLs, counts, rates, dates and caveats. Summarize topic patterns only when the titles or transcripts support them. Save the cleaned ranked table as a downloadable CSV if available.
Break Down the Creative Pattern
Give ChatGPT the transcripts from your strongest follower-generating videos and a few ordinary or weaker ones. Match every transcript to its video ID or URL. Include the opening visual and on-screen text. A spreadsheet title does not reveal the whole video.
Here are transcripts and visual notes matched to my ranked videos: [paste them]. Compare my strongest follower-generating videos with the comparison videos. For each, identify the exact spoken hook, opening visual, viewer problem, topic, promise, structure, demonstration or evidence, payoff, length and call to action. Explain what the viewer gets immediately and what reason the video gives them to follow for more. Label explanations as hypotheses, not proven causes. Do not invent retention data, audience psychology or visual details. Identify three to five repeatable combinations of topic, hook and format supported by multiple examples where possible. Show the source video IDs behind each. If only one video supports a pattern, say so. Create a one-page playbook with: keep doing, adapt, test, and stop assuming. End with ten original angles that fit my audience, my filming constraints and the strongest patterns. Avoid merely swapping nouns in the same script.
Turn a pattern into something new
Illustrative pattern: Show a useful result first, name the frustration, demonstrate three steps, then offer a specific reason to follow.
A cooking creator might turn that into three different ideas: a one-pan meal with minimal cleanup, lunch from leftover ingredients, and a freezer dinner for a busy evening. The structure is familiar, but each video solves a different problem.
Maverick's tip
Use comparisons to challenge the playbook. If the same hook appears in both strong and weak videos, it probably is not the whole explanation.
Write 30 Original Scripts
Run this after the audit and creative breakdown. The input is your evidence-backed playbook. The output is a batch of scripts to test, not a promise that the next 30 posts will repeat an old result.
Now write 30 complete, original Instagram video scripts using our playbook, my voice examples and my filming constraints. Allocate 15 scripts to our strongest supported patterns, 10 to nearby topics or formats, and 5 to new experiments. Treat this mix as a starting plan, not an algorithm rule. If we do not have enough evidence, say so and label more of the batch as exploratory. For each script include: number, working title, intended viewer, source pattern and source video ID where available, exact spoken hook, opening visual, full spoken body, on-screen text, demonstration or filming notes, specific payoff and a relevant call to action. Include an estimated runtime based on word count and state the speaking-rate assumption. Write in my natural voice. Make each script useful on its own. Use different problems, examples and payoffs. Do not copy competitors, invent my experiences, add unsupported statistics, or promise a guide, product or result I have not supplied. Mark factual claims that need checking before filming. Keep each script within my stated target length and practical to film. End with the five scripts you would film first and explain their evidence, relevance and production effort. If the response is too long, deliver scripts 1-10 in full, then continue 11-20 and 21-30 when I ask. Do not replace complete scripts with one-line ideas.
Read the first batch out loud
Check whether you would actually say the opening, whether the demonstration delivers the promise, and whether the ending gives the right person a reason to follow. Tighten the scripts before filming the whole batch.
Maverick's tip
You are borrowing the pattern from your own evidence. A fresh example, demonstration or useful result is what makes the next video worth watching.
Make the Scripts Sound Like You
Paste a script that feels stiff alongside two or three transcripts that sound natural. Keep any actual result or personal story accurate while changing the delivery.
Rewrite scripts [numbers] to sound more like these examples of my natural voice: [paste transcripts]. Keep the useful point, factual meaning and promised payoff. Make the opening specific and the body easy to say out loud. Remove generic hype, repeated transitions and unnecessary setup. Do not add personal experiences or results. Give me the final spoken script, a word count, three alternative hooks and simple filming notes. Label any factual claim I still need to verify. Preserve any wording I mark as exact.
Film a small batch first
Pick three to five scripts that are realistic to film now. Make sure your profile clearly tells a new visitor what you help with and what they will get by following. A video should fit the promise your account makes.
Try changing one major element at a time when you want a useful comparison. For example, keep the format similar while testing two distinct audience problems. Real posts still differ in timing, distribution and audience, so treat the result as directional evidence.
Keep a simple results log
Record the script number, pattern, post URL, publication date, date measured, views and per-video follows when available. Add filming effort and a short creative note. Choose consistent checkpoints you can maintain, such as seven days after posting, and compare like with like.
Maverick's tip
Do not discard an entire topic because of one weak video. Check the opening, demonstration, timing and distribution before deciding what to change.
Feed the Results Back In
Upload the new results alongside the earlier baseline. Keep the measurements comparable and ask ChatGPT to distinguish real evidence from a flattering explanation.
Review this new batch against our previous baseline: [attach or paste results]. Confirm the metric definitions, measurement windows and distribution groups are comparable. Rank by total attributed follows and follows per 1,000 views where supported. Show both, and flag missing data, small samples or immature posts. For each pattern, report the number of videos, total views and follows, median per-video rate, and aggregate rate calculated as total follows / total views. Do not confuse an unweighted average of percentages with the aggregate rate. Tell me what improved, what underperformed and what is still inconclusive. Do not claim causation from this observational comparison. Choose three patterns to continue, up to three to adapt, and one new experiment. Explain the evidence and give me the next five script briefs.
Save your playbook
Keep the original exports, cleaned rankings, transcript library, scripts and results log together. When you start a new chat, upload the relevant files and your latest one-page playbook instead of assuming ChatGPT remembers all the earlier data.
A useful rhythm
Research enough to find a good question. Make a small batch. Let the posts reach your chosen checkpoint. Review what the audience actually did, then write the next batch. You can work through 30 scripts over several weeks without making all 30 before learning anything.
If Your Data Is Missing or You Are New
You can still build a useful content plan. Keep the evidence label honest so a guess does not become a "proven" strategy after a few messages.
No per-video follows?
My file does not contain reliable per-video follows. Do not calculate or estimate follower conversion, and do not allocate account-wide growth to individual posts. Analyze the metrics actually available, such as views, reach, saves, shares or watch time. Keep their definitions distinct and label any ratios clearly. Explain what each metric can and cannot tell us. Identify promising content to inspect manually and give me a simple table for collecting per-video follows and comparable views if those become available. Call this a content-performance analysis, not a follower-conversion ranking.
No published videos yet?
I am starting without my own performance history. My niche is [niche], my audience is [audience], and I want to be followed for [ongoing value]. I can film [constraints]. Here are relevant niche examples I can provide: [examples]. Create nine original starter scripts across three clearly different topic-and-format hypotheses. Use three scripts per hypothesis. Include complete spoken copy, opening visual, payoff, filming notes and an appropriate call to action. Label all patterns as unvalidated for my account. Give me a simple results log and a plan to review comparable metrics after publishing. Do not invent evidence or predict guaranteed growth.
Maverick's tip
Account-wide follower growth is useful context. It cannot tell you exactly which individual video earned each follower.
Common Mistakes and Easy Fixes
| Mistake | Better approach |
|---|---|
| Asking ChatGPT to study videos it cannot access | Provide transcripts, opening lines and visual notes. Require it to identify gaps. |
| Exporting only an account summary | Use per-video content rows for a per-video ranking. |
| Assuming the export covers the account's full history | Record the actual available dates and limitations. |
| Treating missing follows as zero | Leave them missing and use the fallback analysis. |
| Mixing cumulative snapshots or adding duplicates | Keep one compatible snapshot per video and a clear audit trail. |
| Comparing paid, trial and ordinary organic posts together | Separate known distribution groups and flag unknown status. |
| Picking the highest percentage from a tiny sample | Show counts, total follows and sensitivity to a sensible threshold. |
| Comparing yesterday's post with a year-old video | Prefer equal-age checkpoints and disclose unequal exposure time. |
| Calling a new script proven | Tie it to its source pattern and test the actual new video. |
| Copying a hook without delivering its payoff | Make the demonstration or useful answer fulfill the opening promise. |
If an upload or export will not cooperate
Open the spreadsheet yourself and confirm the file is readable. If needed, provide a smaller relevant file or paste a small table with the original headers. When the export is unavailable, manually record the metrics shown for a manageable set of videos. Mark the sample as partial.
If the account's settings or labels differ from this guide, ask for help using a screenshot with private details removed. Do not let ChatGPT guess a click path from memory.
Quick Reference
| What you want | What to do |
|---|---|
| A useful strategy brief | Step 1: give your niche, audience, voice and filming constraints. |
| Current niche patterns | Step 2: supply examples and demand accessible evidence. |
| Your actual performance history | Step 3: export content rows and record the coverage. |
| Numbers you can trust | Step 4: audit columns, duplicates, scope and missing values. |
| Videos that earned followers | Step 5: compare total follows and follows per 1,000 views. |
| A repeatable creative playbook | Step 6: connect the ranking to real transcripts and visuals. |
| 30 complete scripts | Step 7: request full scripts tied to the supported patterns. |
| Natural delivery | Step 8: revise against your own voice examples. |
| A smarter next batch | Step 9: compare consistent checkpoints and update the playbook. |
The formula: Per-video follows / comparable views x 100 = conversion percentage. Multiply by 1,000 instead for follows per 1,000 views. Missing follows means no follower-conversion calculation.
The deliverables: A niche playbook, cleaned ranking, transcript breakdown, 30-script batch and results log. Save them so your next review starts with evidence.
Maverick's tip
Do this today. Start one ChatGPT conversation with the setup prompt. Export the Instagram content data you can access, audit it, and identify three promising videos. Add their transcripts, create your first script batch, and choose one useful video to film.
Find the pattern. Make something useful. Give people a reason to come back.