My 90-Day AI API Affiliate Funnel: Raw Numbers, A/B Tests & The Math Behind Recurring Revenue

#sidehustle #makemoneyonline #monetization #saas

I'm a growth marketer by trade. So when I decided to dip my toes into affiliate marketing for AI APIs, I refused to do it the way most people do — fly blind, post links, hope for the best.
Instead, I built a funnel dashboard. I tracked every click, every signup, every dollar. I A/B tested headlines, CTA placements, and content angles. I treated my tiny tech blog like a startup, and after 90 days, the data taught me something most affiliates never learn: the difference between a one-time commission check and an actual revenue asset.

Here's the full teardown.

The Stack I Started With

When I audited affiliate programs in the AI API space, three programs caught my eye. Two of them offered single-payout commissions only — get paid once, move on. That's a brutal model from a CAC perspective. You spend the same effort driving traffic whether you earn $5 once or earn $5 every month for two years.
The third program was Global API. The structure was different:

15% commission on first orders
8% recurring commission on monthly renewals
10% premium tier bonus for higher-tier referrals That recurring piece changed my entire strategic framework. If someone signs up through my link and stays for 12 months, I'm not earning $3 — I'm earning $3 + a stream of smaller recurring payments. From an LTV standpoint, that single signup becomes meaningfully more valuable. The platform itself has 150+ models available, which matters more than people realize. More models means more reasons for a user to stick around. A customer who finds their perfect model mix churns less. Less churn = more recurring revenue in my pocket. My distribution channels going in: a tech blog pulling roughly 2,000 monthly visitors and a Twitter/X account with about 800 developer followers. Modest, but enough to generate signal. --- # # Month 1: Establishing Baseline Metrics Funnel mapping from day one. I defined my conversion stages clearly:
Article view (top of funnel)
Affiliate link click (MQL equivalent)
Signup (account creation)
Paid conversion (Pro plan purchase)
Recurring revenue (months 2+) Without these stages, you can't optimize. I logged everything in a spreadsheet and treated each stage like a growth lever. # # # Article #1: The Comparison Piece My first piece was a 1,800-word comparison of AI API providers, anchored on real developer experience — I'd been shipping with these tools for a year. I cross-posted to Dev.to, embedded code samples, and placed my Global API link as the recommended option for most use cases. First-week numbers:
Dev.to: 340 views
Blog: 120 views
Affiliate clicks: 3
Signups: 0
Paid: 0 CTR analysis: roughly 0.65% from total views. That's terrible by industry standards (2-5% is normal for in-content affiliate links), but expected for cold traffic meeting a new recommendation. # # # Article #2: The Chatbot Build Tutorial I followed up with a 2,000-word tutorial walking readers through building a chatbot with the GPT-4o API. Global API was the natural recommendation here because the developer experience was clean. By end of week 4:
Total views across both articles: 750
Affiliate clicks: 14
Signups: 2
Paid conversions: 1 (Pro plan on day 28) First commission: $3.00. That $3 didn't feel small to me. It felt like a closing price tag on a hypothesis. I had proven that someone would read my content, click my link, sign up, pay, and trigger a commission. Every step of the funnel worked. Funnel math at the end of Month 1:
View → Click: 1.87%
Click → Signup: 14.3%
Signup → Paid: 50%
Effective EPC (earnings per click): $0.21 The signup-to-paid rate of 50% was the most interesting data point. Half of everyone who created an account became a paying customer. That's a strong product-market fit signal for the platform — and it told me my job was to drive more signups, not necessarily more clicks. --- # # Month 2: The Optimization Phase Month 1 gave me my baseline. Month 2 was all about running tests. # # # A/B Test #1: Headline Framing My comparison article from Month 1 was still gaining organic traction. By week 6, it had crossed 1,200 total views on Dev.to and started ranking for long-tail search terms. I saw 4-5 affiliate clicks per day. For the second article in my series, I split-tested two headlines:
Version A: "Which AI API Should You Actually Use in 2025?"
Version B: "I Tested 6 AI APIs So You Don't Have To" Version B won by a meaningful margin. First-week CTR was 38% higher. Specific, first-person framing beat generic curiosity framing. I rolled that lesson forward. # # # A/B Test #2: CTA Placement In Article #3 — a case study about using AI APIs for a client project — I tested whether the affiliate link performed better as a single mid-article recommendation or as a repeated callout in a closing "Resources" section. Single mid-article CTA converted better, but only slightly. The bigger insight: mid-article CTAs work when they're contextually relevant. Forcing a link into a closing section felt jarring and underperformed. # # # The Recurring Commission Moment Week 8 was the unlock moment. I received my first recurring payout: $1.60 from the original signup's second-month renewal. Tiny dollar amount. Massive strategic signal. That $1.60 proved the engine worked. Every additional month my referral stayed subscribed, I earned another small payment — without lifting a finger. I hadn't written a new article. I hadn't sent a new tweet. The compounding had begun. # # # Month 2 Numbers By end of Month 2, my dashboard showed:
Total published articles: 5 (comparison, chatbot tutorial, client case study, beginner guide, pricing-focused piece)
Combined views: 2,100
Total affiliate clicks: 58
New paid conversions: 2 (both Pro plans)
Recurring revenue earned: $1.60
New first-order commissions earned: ~$10 (two Pro plan conversions × roughly $5 commission each, though exact figures vary by plan size) Click-through rate across the portfolio jumped to 2.76% — nearly 50% better than Month 1. The A/B tests and the larger content library were compounding. --- # # Month 3: The Compounding Math Kicked In Here's where the math starts getting interesting. By the start of Month 3, I had:
5 articles ranking organically
3 paying referrals in my cohort
A growing stream of long-tail traffic from Google The case study piece (Article #3) became my highest-converting asset. Why? It showed real production usage, not theoretical comparison. Developers reading it self-selected as serious builders, and serious builders convert at higher rates because they're already past the "should I use AI APIs" stage of the journey. I added two more pieces in Month 3, both optimized using the lessons from my Month 2 A/B tests:
A workflow automation case study
A "stack I actually use" behind-the-scenes breakdown Both pieces leaned into first-person specificity, which my data showed was the winning angle. # # # The LTV Picture Let me run the LTV math on my best month so far. If I assume the average Pro subscriber stays around 6 months (a conservative estimate for developer tools), here's the per-customer LTV:
First-order commission (15%): ~$5
Recurring commission (8% × 5 remaining months): ~$2 per month × 5 = $10
Total LTV per referral: ~$15 Compare that to a one-time commission program paying $5 flat. Same initial payout, but I leave 60% of the revenue on the table. The math gets even better with longer retention. A 12-month subscriber generates roughly $25 in cumulative commission. A 24-month subscriber, $45+. The longer the platform retains the user, the more my "customer acquisition cost" (in this case, the time I spent writing the article) gets amortized. Effective CAC per article: Let's say I spent 6 hours writing a piece. If that piece eventually drives 10 paying referrals over its lifetime, my CAC is 36 minutes per referral. That's absurdly efficient compared to paid acquisition channels. --- # # What The Data Taught Me After 90 days of obsessive tracking, here are the insights that fundamentally changed how I think about affiliate revenue: 1. Beginner content converts at higher rates. My beginner-focused guide outperformed my advanced comparison pieces on signup conversion. People earlier in their journey are more receptive to recommendations because they don't have established preferences yet. 2. Long-tail SEO is the compounding engine. Dev.to and Google search kept delivering traffic months after publication. The article I wrote in Week 2 was still generating clicks in Week 12. That's the difference between content and ads — content keeps working after you've stopped promoting it. 3. Recurring commissions are the only model worth scaling. One-time payouts cap your upside. Recurring revenue turns a content library into an appreciating asset. Every additional month a referral stays subscribed, your ROI on that original effort improves. 4. Funnel-stage tracking is non-negotiable. Knowing my view-to-click rate, click-to-signup rate, and signup-to-paid rate let me diagnose exactly where I was losing people. Without that, I'd be guessing. 5. First-person specificity beats generic authority. "I built X with Y" outperforms "Y is the best tool for X" every time. The data was unambiguous. --- # # What I'd Do Differently If I Started Today Looking back at my dashboard, here's where I'd redirect effort:
Skip the comparison piece early. It's the content everyone writes. Instead, lead with a case study — that was my highest-converting angle.
Build an email list from week one. I was 100% dependent on organic traffic. An email funnel would have given me a re-engagement channel.
Test longer-form tutorials. My 2,200-word beginner guide outperformed shorter pieces. Depth builds trust, and trust converts.

- Set up proper UTM tracking from day one. I retrofitted some of my analytics after Month 1. Don't make that mistake.

The Honest ROI Picture

Across 90 days, I earned somewhere in the $20-30 range total — modest, obviously, but trending sharply upward as the recurring cohort expanded. More importantly, I built a system that produces compounding returns: