I Sold AI Prompts on Gumroad for 90 Days: Real Numbers
I Sold AI Prompts on Gumroad for 90 Days — Here's What Actually Made Money
Ninety days ago I uploaded a Gumroad product with zero followers, zero email list, and zero idea whether anyone would pay for a text file full of prompts. I want to walk you through exactly what happened — the sales, the refunds, the pricing mistakes, and the one change that actually moved the needle — because most "I made $X selling prompts" posts skip the boring, useful parts.
This isn't a highlight reel. It's the log.
Why I Started Selling AI Prompts on Gumroad
I'd been using AI tools daily to run every part of my online business — content, SEO research, customer replies, product descriptions, even debugging code I didn't fully understand. Somewhere around month four I noticed I had a folder of prompts I kept reusing and refining. Not clever one-liners, but structured, multi-step prompts that consistently gave me usable output on the first or second try.
A friend asked me to send him a few. Then someone in a small online business group asked the same thing. That's usually the signal worth testing: when people are asking for something you already have, there's a product hiding in there.
Gumroad made sense for three reasons. No inventory, no shipping, and a checkout flow simple enough that someone could buy on their phone in under a minute. If you're in the US, UK, Canada, Australia, or New Zealand, Gumroad payouts land in your bank account with minimal friction — no waiting on international wire transfers or dealing with a marketplace that takes 30-40% like some course platforms do.
I decided on a simple offer: a organized prompt vault, not a random dump. Prompts sorted by use case — starting a business, growing one, automating parts of it — so a buyer could open the file and immediately find something relevant to what they were doing that day, instead of scrolling through 200 unrelated lines.
The First 30 Days — What Actually Happened
I'll be honest about the start: it was slow, and the first version of the product wasn't good enough to justify what I charged for it.
What I priced it at (and why that was wrong)
I launched at $37. That number came from watching a few successful creators price their prompt packs in that range and assuming I could match it. Nobody told me the price needs to match the perceived depth of the product, not the category average.
Here's what the first two weeks looked like:
- 4 sales
- 2 refund requests
- $0 in repeat customers (obviously, 30 days in)
Both refund requests said some version of the same thing: "this feels like it's not worth $37 compared to what's out there." That stung, but it was accurate. My first version had maybe 400 prompts, loosely organized, with almost no explanation of how to actually use them. I'd made the mistake every beginner makes — treating quantity as the pitch instead of usefulness.
I dropped the price to $19 and left it there while I rebuilt the product. Sales ticked up slightly — 6 sales in the next 10 days — but the refund pattern told me price wasn't the real problem. Structure was.
What Changed in Days 30–60
I rebuilt the vault from scratch over about a week. Instead of a flat list, I organized it into three clear categories that matched actual stages of running an online business: starting, growing, and automating. Each prompt got a one-line note explaining when to use it and what kind of output to expect.
I also added something I hadn't thought to include the first time: example outputs. Not screenshots, just short samples showing what a good response to that prompt actually looks like. That one change did more for conversions than anything else I tried.
By day 45 I raised the price back up — this time to $27, landing between my two earlier price points, but now backed by a product that actually held together. The results were different:
- 14 sales in the next 15 days
- 1 refund request (and it was a legitimate mismatch — the buyer wanted image-generation prompts specifically, which I didn't cover)
- First returning customer, who came back to buy a second, smaller product I'd listed
The refund rate alone tells the story. Same general product category, better structure, roughly a third of the refund rate.
I also started paying attention to where traffic was actually coming from, and this part matters if you're doing something similar. My Google Search Console data showed a slow but real increase in impressions and clicks for blog content around "AI prompts" and "digital products to sell" — not viral, not overnight, but a steady climb week over week. Meanwhile my GA4 traffic dashboard was full of noise: sessions from Singapore and China with 10-20 second engagement times and zero conversions. If I'd been optimizing based on raw pageviews instead of Search Console clicks and actual Gumroad sales, I would have drawn completely wrong conclusions about what was working.
That's worth repeating for anyone tracking their own numbers: pageviews are not customers. Watch what converts, not what's loud.
The Real Numbers After 90 Days
Here's the full breakdown, no rounding up:
- Total sales: 31
- Total refunds: 3
- Net revenue: roughly $650 after Gumroad's fee
- Average order value: $21
- Repeat customers: 4
- Traffic sources for sales: direct link shares (Telegram, a couple of Instagram DMs) accounted for about 60% of sales; blog content and search accounted for the rest
That's not retire-on-a-beach money. It's also not nothing for something built with zero ad spend, zero paid promotion, and maybe 15-20 hours of actual product work spread across three months. What it proved was the model works at a small scale, which matters more than the dollar figure — because small scale that works is scale you can repeat and expand.
The clearest lesson from the numbers: my second version of the product, priced lower than my original mistake but higher than my panic price, sold at roughly triple the rate of version one with a third of the refunds. Fixing the actual product mattered more than fixing the price.
What Actually Sells (and What Doesn't)
A few patterns showed up clearly enough that I trust them:
Organized beats abundant. Nobody wants 1,000 loose prompts. They want 150 prompts they can actually find and use. Every buyer message that praised the product mentioned the categories, not the quantity.
Examples convert better than descriptions. Telling someone "this prompt writes great product descriptions" is weaker than showing them one paragraph of what that actually looks like. It removes the guesswork before they buy.
Niche-specific outperforms general. My best-selling individual section wasn't the broad "starting a business" category — it was the automation section, aimed at people who already had something running and wanted to save time on repetitive tasks. Buyers with an existing business in mind knew exactly what problem they were solving.
Refunds are data, not failure. Every refund request in the first 30 days pointed at the same weakness. That's a gift, even though it doesn't feel like one when it happens.
Direct relationships outsell cold search traffic, at least early on. Nearly two-thirds of my sales came from people I had some existing connection with — a Telegram community, a comment thread, a DM. Search traffic is compounding and worth building, but it's slow. If you're starting from zero, don't wait on SEO alone to pay your bills.
If You're Thinking About Doing This Too
A few honest notes if you're considering building something similar:
Don't launch your first version at a premium price. Start lower, get real feedback fast, and raise the price once the product earns it. The math on a $37 product with a 50% refund rate is worse than a $19 product nobody asks to return.
Build for a specific stage of the buyer's journey, not a general audience. "Prompts for business" is vague. "Prompts for automating the parts of your business you're already tired of doing manually" is a decision someone can make in ten seconds.
Track the metric that pays your bills, not the one that looks impressive on a dashboard. Sales and Search Console clicks told me the truth. GA4 pageviews, inflated by bot traffic from regions with no buying intent, told me nothing useful.
And treat your first 90 days as research, not a verdict. My first two weeks looked like a failed idea. By week twelve, the same idea — rebuilt once — was working. The gap between those two outcomes wasn't luck. It was paying attention to what the refunds and the silence were actually telling me.
If you want the exact vault this whole experiment was built around — the rebuilt version, organized by stage, with example outputs included — it's the same one I sell today: 10,000 AI Prompts to Start, Grow & Automate an Online Business. It's the product these 90 days were spent getting right, and it's priced based on what actually earned trust, not what I guessed on day one.
Internal links to include when publishing:
Reminders before publishing:
- Add descriptive alt text to every image you upload in Blogger (can't be set from here)
- Double-check your Blogger theme for duplicate canonical tags if your SEO audit flagged that before