The task
You have a paid offer concept (a $79 course, a $6 eBook, a $14 Notion template pack). You need a structure that justifies the price, that you can actually ship without 6 months of work, and that nets out positive after the platform takes its cut. This guide gives you the prompts to make AI design that skeleton, plus the current (June 2026) fee math so the structure you pick is the one you can afford to sell.
TL;DR
- AI is good at structure, not voice. Hand it the buyer’s before/after state and 5 real search questions; it returns a defensible module/chapter map. You write the lessons.
- Three product types follow three different structure logics: courses sell on promise, eBooks on linear narrative, templates on visible utility. Do not reuse one skeleton for all three.
- Platform fees decide your floor price. Gumroad takes 10% + $0.50 per sale (12.9% + $0.80 with processing); Notion Marketplace takes 10% + $0.40; Teachable Basic adds a 5% transaction fee on top of its subscription. A $6 eBook on Gumroad nets you about $4.72.
- For the AI step, use a 1M-token model (Claude Opus 4.7 or Gemini 3.1 Pro) so it can hold your whole brief plus competitor outlines in one pass.
When this is the right job for AI
- You know the buyer and their before/after state.
- You can hand AI 3-5 specific buyer questions (“how do I price the first product?”) that the content should answer.
- You want a structural skeleton, not the copy. Writing the modules is still your job.
If you cannot name the buyer’s before/after in one sentence each, stop and do that first; the model will invent a generic buyer and you will get a generic outline.
Which model to use (June 2026)
This task is structure-heavy and benefits from a long context window so the model can hold your brief, your competitor’s table of contents, and your past content in one prompt.
| Model | Context window | Best for this task | Cost |
|---|---|---|---|
| Claude Opus 4.7 | 1M tokens | Strongest at constraint-following (“6 modules max, cut don’t split”) and scope warnings | Pro $20/mo; API $5 / $25 per 1M tok |
| Gemini 3.1 Pro | 1M tokens (Google AI Pro $19.99/mo) | Paste 3-4 competitor outlines, ask it to find the gap | API $2 / $12 per 1M tok |
| GPT-5.5 (Thinking) | ~320 pages in ChatGPT Plus; full 1M only on $200 Pro | Quick single-product outlines | Plus $20/mo |
For a single offer, any Pro-tier plan works. If you are pasting several long competitor outlines as reference, Claude Opus 4.7 or Gemini 3.1 Pro hold them without truncation; the in-app ChatGPT Plus window caps near 320 pages, so trim references first. See our ChatGPT vs Claude vs Gemini comparison for the full breakdown.
What to feed the AI
- Offer: format, price, length (e.g. “6-module course, 4 weeks, $79”)
- Buyer: who they are, where they are now, where they want to be in 8 weeks
- 5 specific pain questions the buyer is currently searching for
- Constraint: how long YOU can spend producing (e.g. “I have 3 weekends”)
Copy-ready prompt
You are designing the structure for a paid offer.
Offer: 6-module video course, $79, ~4 hours total runtime.
Buyer: solo creators on Xiaohongshu with 1k-5k followers stuck under 10k.
Where they are now: posting 2-3x/week with inconsistent voice and no offer ladder.
Where they want to be in 8 weeks: 10k+ followers and one paid offer live.
Current search questions:
- How do I find a niche I can actually monetize?
- How do I write hooks that don't feel like every other creator?
- What's the smallest paid product I can ship in 4 weeks?
- How do I price something nobody has bought from me before?
- What do I do with people who unsubscribe?
Constraint: I have 3 weekends to produce.
Output:
1. The promise the sales page can make (one sentence).
2. 6 modules, each with: one outcome, three lessons, the one deliverable the buyer ships at the end.
3. The two modules I can build first (highest leverage to validate the offer).
4. The single objection most likely to kill conversion, and where in the structure to address it.
5. A "scope warning" if my 3-weekend constraint can't actually produce 4 hours.Sample output structure
- Promise: “In 4 weeks you will have a recognizable voice, your first paid offer live, and one weekly content engine you can keep running.”
- Modules:
- M1 Niche-pricing fit: outcome = one defended niche. Lessons: niche tests, demand signals, pricing floor. Deliverable: 1-page niche brief.
- M2 Voice: outcome = a 5-line voice doc. Lessons: voice audit, anti-voice list, voice training prompt. Deliverable: voice doc.
- M3 Hook system: outcome = 30 hooks ready to deploy.
- M4 Offer ladder: outcome = a 3-tier offer (free → $19 → $79).
- M5 Launch loop: outcome = first paid offer live.
- M6 Retention loop: outcome = a 90-day plan.
- First to build: M1 + M4 — without these, nothing else converts.
- Top objection: “I haven’t built an audience yet.” Address in M1 with a demand-validation lesson before any niche choice.
- Scope warning: 4 hours over 3 weekends = 80 min recorded per weekend, before edits. Realistic if you outline everything before recording; tight if you don’t.
Price-floor reality check: net out the platform fee first
Before you lock a price, run the structure against the cut your platform takes. As of June 2026:
| Platform | Fee per sale | $6 eBook nets | $79 course nets |
|---|---|---|---|
| Gumroad (direct link) | 10% + $0.50 (12.9% + $0.80 with Stripe/PayPal processing) | ~$4.72 | ~$68.01 |
| Gumroad Discover marketplace | 30% | ~$4.20 | ~$55.30 |
| Notion Marketplace | 10% + $0.40 | ~$5.00 | n/a (templates only) |
| Teachable Basic | 5% transaction fee + $39/mo subscription | n/a | ~$75.05 before the monthly fee |
Sources: Gumroad pricing, Notion Marketplace seller terms. Thinkific and Teachable’s higher tiers charge 0% transaction fees but a higher monthly subscription, so they only win above a few thousand dollars/month in sales.
Why this belongs in the structure step: a $6 eBook on Gumroad Discover nets ~$4.20, which means a refund-heavy chapter design (no sample, weak Ch 1) costs you real money per refund — Gumroad keeps its 10% + $0.50 even on refunds. Cheap products need the lowest-friction structure precisely because the margin is thin.
Ebook worked example: “Build the Pomodoro Habit in 30 Days”
Same prompt template, swapped for a $6 eBook — 9 chapters, ~28,000 words. Ebook structure is a chapter chain, not a module set. Readers go Ch 1 → Ch 9 linearly; they do not “pick one chapter to watch” the way they do with a course.
Offer: ebook, $6, 9 chapters, ~28k words.
Buyer: knowledge workers who tried Pomodoro and quit after 3 days, want a real focus habit.
Where they are now: 3 Pomodoro apps installed, abandoned each after 2-3 days.
Where they want to be in 30 days: 4 stable Pomodoros daily, 21-day unbroken streak.AI’s 9-chapter skeleton (after editing):
- Ch 1 Why you quit last time: four failure modes (wrong cadence, noisy environment, missing reward, self-blame). This chapter is the hook — put it in the sample.
- Ch 2-6 Five-day practical (one chapter per day): Day 1 calibrate cadence / Day 2 design environment / Day 3 handle interruptions / Day 4 work with procrastination thoughts / Day 5 weekly review. Each chapter = one exercise card + one pass/fail criterion.
- Ch 7 The second-week trap: “I’ve done 5 days, I deserve a break” — the highest-save-rate chapter.
- Ch 8 Advanced stacks: Pomodoro x time-blocking, x task list, x weekly review — for readers past Day 14.
- Ch 9 What happens after Day 30: habit-locked-in checklist + when to stop using Pomodoro.
Sample strategy: release Ch 1 + Ch 3 in full (~4,000 words) as a downloadable PDF on the landing page. The TOC itself sits mid-page as the sales tool — nine chapter lines, one-line abstract each, reader decides from the TOC.
Why ebook = chapter logic, not module logic: course modules can be consumed out of order, abandoned mid-way (“I just needed M1”); ebook readers expect linear narrative, Ch 4 assumes you read Ch 1-3. That’s why the sample chapters are 1 + 3 (build trust), not a random pick.
Notion template-pack worked example: “Freelancer Finance Template”
$14, 5 interlinked databases + 1 dashboard. Template structure is visible utility, not narrative. Buyers see the value in the preview image — they do not read it.
Offer: Notion template pack, $14, 5 databases + 1 dashboard + 1 README.
Buyer: freelancers $50k-$300k/yr, spending the month before tax season digging receipts out of email.
Where they are now: invoices scattered across email, bank app, paper folders; tax prep takes 2 weeks.
Where they want to be in 1 week: all income/expense auto-categorized, tax export done in 1 hour.AI’s 5+1 structure (after editing):
- DB 1 Income: client, project, invoice #, amount, paid date, tax category.
- DB 2 Expenses: category, client attribution, receipt image, pre-tax/post-tax flag.
- DB 3 Clients: contact info, contract link, lifetime revenue, last touch.
- DB 4 Invoices: status (draft / sent / overdue / paid), linked client, linked project.
- DB 5 Tax buckets: quarterly withholding, annual deduction categories, cash-reserve target.
- Dashboard: month-to-date revenue vs target / outstanding AR / current-quarter tax bucket / top 5 clients / 30-day cashflow chart.
Preview gallery strategy: lead with the dashboard (one image shows the whole value), then show the 5 DB details. Flip the order — DB details first — and the buyer can’t connect “how does this help me.” The dashboard is the pitch.
README is part of the product, not docs: include (1) the 3-step copy-to-your-workspace flow, (2) the 6 most common customization scenarios with how-to-edit (“how do I add a currency field”), (3) integration notes with invoice-PDF generators. About 70% of template refunds come from “couldn’t install”; a real README saves half of them.
Marketplace note: if you list on the official Notion Marketplace you must onboard through Stripe and clear the creator waitlist; payouts run biweekly on Thursdays with a $20 minimum balance. Most template sellers cross-list on Gumroad for the direct link and reserve the Marketplace listing for discovery.
Why template = visible utility, not narrative: courses and ebooks sell on promise (“in 30 days you will…”); templates sell on preview image (“I see the dashboard, I can use it now”). So template landing pages need less copy and more screenshots — 8+ images, high resolution, hover-to-zoom.
How to refine the AI’s output
- AI gives 12 modules → cap it: “6 modules max. Cut, do not split.”
- Outcomes are vague → require “each outcome must be a noun the buyer can hold up after the module.”
- Skipping objection handling → “name the objection that kills the most conversions and where in the structure to address it.”
- Promise sounds like a coach’s slogan → “the promise must be testable in 4 weeks.”
- The model pads with filler lessons → “every lesson must change what the buyer does on Monday; delete any lesson that’s just background.”
Common mistakes
- Designing the modules before the promise. The promise constrains the modules — not the other way around.
- 8 modules at $79. Buyers don’t finish 8 modules; they remember “I didn’t finish” and ask for refunds.
- No deliverable per module. Without a deliverable, modules feel like content, not training.
- Putting the objection-handling at the end. By then, the buyer already churned.
- Pricing before checking the fee. A $5 eBook on Gumroad Discover nets ~$3.50 after the 30% cut — sometimes a $9 price clears more.
FAQ
- Course vs eBook vs Notion template? Course = highest price ($79+) and highest production cost. eBook = lowest price ($6-15) and fastest to ship. Template = low cost AND high retention if genuinely useful. AI can compare all three by promise: feed it the same buyer and ask which format the before/after state best supports.
- Should AI write the actual lessons? Outline plus a rough first draft, yes. Final copy is yours — the voice is the moat, and a marketplace full of AI-flat eBooks is exactly where readers refund fastest.
- Which AI model should I use for this? For a single offer, any $20 Pro-tier plan (ChatGPT Plus, Claude Pro, Google AI Pro) is fine. If you’re pasting several competitor outlines as reference, pick a 1M-token model — Claude Opus 4.7 or Gemini 3.1 Pro — so nothing gets truncated.
- What about cohort-based courses? Same structure prompt; add “and a live element per module” as a constraint. AI will redistribute scope so the recorded part shrinks and the live calls carry the rest.
- How do platform fees change my price floor? Gumroad takes 10% + $0.50 per sale (12.9% + $0.80 with payment processing) as of June 2026; Notion Marketplace takes 10% + $0.40. Multiply your target take-home by roughly 1.15 to set the list price, then round up to a clean number.
Related
- AI digital product sales page
- AI newsletter lead magnet
- AI course sales page
- AI personal brand voice design
- Comment-to-Content Prompts: Turn Replies Into Posts
Tags: #AI writing #Creator #Creator monetization #Digital product #Course