Shipping a digital product as a solo founder

I rebuilt the ToolGenX site in three weeks this spring. The old one was a Next.js app that sent every buy click to Gumroad, did about one verified sale per ninety days, and had no idea why. The new one is what you are reading now: nineteen products, direct payment, a refund flow that satisfies EU law, and a content layer that AI search engines can actually quote.

I want to write down the playbook while it is fresh, because the same handful of mistakes keeps showing up in other founders' Discord posts and the answers are not as complicated as we make them.

The four things you have to own end-to-end

You can outsource everything else. These four are the load-bearing pieces:

1. Distribution. One channel, repeated, until it works or until you can prove it never will.
2. Payment infrastructure. Stripe, Iyzico, or both. Not a marketplace redirect. Not "and we will figure out the cart later."
3. Legal and refund flow. Especially if you sell to EU customers. The EU Consumer Rights Directive Article 16(m) waiver is real and you must collect dual consent at checkout.
4. Content gravity. Three to ten posts that answer the exact questions your buyer types before they buy. Indexed by Google. Quotable by ChatGPT and Perplexity.

That is the whole list. Everything else — design polish, mascot, podcast appearances, founder story page — is optional. These four are not.

1. Pick one distribution channel and stay there for six weeks

Most solo founders fail at distribution because they try three channels half-heartedly instead of one channel fully. The "channel" is whatever you can show up on every weekday for six weeks without burning out:

• Long-form writing on your own site, syndicated to Medium and dev.to
• Short-form video (YouTube Shorts, Reels, TikTok)
• Long-form video (YouTube tutorials, in-depth product walkthroughs)
• Twitter/X writing, in public, with a real account
• Reddit posts in 2-3 relevant subs, no spam

That is the whole menu. Pick one. The choice matters less than the consistency.

Six weeks is the test. If you cannot do six weeks of two posts a week on your chosen channel, you do not have a distribution problem — you have a focus problem. Fix that first.

After six weeks you will know whether the channel is wrong, the message is wrong, or you just need more reps. Without those six weeks you cannot know any of those things, no matter what your analytics dashboard says.

2. Payment infrastructure: own the checkout

A Gumroad or Lemon Squeezy redirect is fine for the first three to ten sales because it lets you validate the price and the product without writing any code. After that crossover point — the moment you have proven there is a buyer — every additional sale on someone else's checkout costs you three things:

• Money. Marketplace fees are 5-10% per sale.
• Trust. Buyers wonder why your domain bounces them to a third party with a different visual identity.
• Signal. You see "1 sale" in your dashboard. You do not see who looked, who clicked buy, who got to step 2 and gave up, who used Apple Pay vs card. That signal compounds.

For ToolGenX I run Stripe Checkout and Iyzico Checkout Form in parallel — both gateways accept global Visa, Mastercard, and Amex, so the buyer picks at checkout. Stripe gives me Apple Pay, Google Pay, Stripe Tax, and a customer portal that handles refunds without me lifting a finger. Iyzico adds native Türkiye support — TRY pricing, taksit (installments), and an e-invoicing path — on top of the same global card coverage.

I also abstracted both behind a PaymentGateway interface in the codebase so I can add a third provider later without touching the order flow. That took half a day. The peace of mind is worth more than the half-day.

3. Legal and refund: the EU consent waiver is non-negotiable

If you sell digital goods to a single customer in the EU or UK, your refund flow must satisfy the Consumer Rights Directive Article 16(m). The plain-language version:

• Before download starts: the buyer has 14 days to cancel for any reason and get a full refund.
• After download starts: the right of withdrawal is waived — but only if the buyer expressly requested immediate access AND acknowledged that they lose the right to withdraw by starting the download.

This means two separate checkboxes at checkout, both unticked by default, with the exact text logged with the buyer's IP, user-agent, and timestamp. If you cannot produce that consent record when a buyer complains, the refund window stretches from 14 days to 14 days plus one year, and the burden of proof is on you.

I built this into ToolGenX before I built the actual checkout. It is not glamorous. It is the reason a credit-card chargeback does not turn into a regulatory complaint.

There is also a new EU mandate effective 19 June 2026: the Withdrawal Button. Every order eligible for cancellation must show a clearly-labelled button in the buyer's account that one-shot triggers the refund. I built that too. It costs me nothing because it auto-checks if the buyer downloaded anything before issuing the refund.

4. Content gravity that gets cited by AI

Search has bifurcated. There is the old Google ten-blue-links search, which is now sharing its result page with Google's AI Overviews. And there is the AI search layer — ChatGPT search, Perplexity, Gemini, Bing Copilot — that quotes from a much shorter list of sites.

Both layers reward the same thing: passages that can be quoted in isolation. A 40 to 60-word self-contained answer to a question you have phrased clearly. A specific number with a date next to it. A named author with a real social profile.

Research from Princeton, Georgia Tech, and IIT Delhi in 2024 found that this kind of structuring — Generative Engine Optimization or GEO — produces 30 to 115% higher visibility in AI-generated responses. The advantage is structural, not algorithmic. You either write quotable passages or you do not.

For my site:

• Every product page has a "Quick Answer" block under the hero.
• Every blog post — including this one — has a quickAnswer field in frontmatter and a FAQPage schema block at the end.
• The author entity is a real Person with a real sameAs list pointing at LinkedIn, X, YouTube, and Medium.
• robots.txt explicitly allows GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot, and Applebot-Extended.
• A public/llms.txt file at the site root gives AI crawlers a tour of the catalog without making them parse HTML.

None of this is glamorous either. It is the SEO equivalent of putting your name on the mailbox.

What I would do differently if I started over

I would write the three to ten cornerstone posts before building the site. Not after. Not in parallel. Before.

The posts force you to articulate who the buyer is and what they are trying to do. The site comes out cleaner, the product pages write themselves from the post outlines, and the SEO surface area lives on a content base that already passed the "would a real person care" filter.

I did it the other way for ToolGenX V1 — site first, posts later — and the result was a site full of generic SaaS copy that I had to humanize a year later. The cost of fixing that retroactively is roughly two to three times the cost of writing the posts first.

A workable order of operations

If you are starting fresh:

1. Week 0: Pick the one channel. Write three post outlines on it.
2. Week 1-2: Write the posts. Ship them to a one-page site that links to a Gumroad checkout.
3. Week 3-4: Watch the traffic. Adjust the offer. Add product two if you have signal.
4. Week 5-6: Same channel, more reps. By now you have either six sales or six weeks of evidence about why not.
5. Week 7+: If sales are real, rebuild the checkout on Stripe and own the trust signal. If sales are zero, change the offer, not the channel.

You do not need a new framework. You need to ship the boring version of all four pieces.

That is the playbook.

---

Written by İsmail Günaydın, founder of ToolGenX and ModernWebSEO. Solo software engineer and SEO strategist out of Istanbul. The thinking behind this post is implemented in the ToolGenX refund flow and product pages.