In my previous case study, I covered the first six months of launching my AI training platform and job board.
I launched aitrainer.work in January primarily as a job board, then scaled onto a multi-product utility platform for data annotators and job seekers interested in entering the AI training space. The first couple of months after launch, revenue had been rather unstable (as expected), oscillating between peanuts and a few hundred USD per month.
In the past month, I have made more than in the first six months combined. In fact, in August alone, revenue hit $4,733, outperforming my entire first six months ($2,863 combined) by 165%.
January 9.4k $709 Launched main site
February 7.0k $350 Got hit by Google update
March 7.4k $24 Referral lag hitting hard
April 9.4k $270 Launched talent pool, free academy
May 11.9k $150 Still in the pipeline gap
June 14.3k $1,360 Paid certification live
July 16.1k $1,222 SEO and proprietary data
August 32.8k $4,733 Referral backlog activated
The August spike came from a single one of my partner platforms that had never paid me before. After months of silence, they hired, onboarded and paid out a batch of ten referrals at once, netting me over $4k.
Most of those candidates were referred back in May. They sat in the onboarding pipeline for close to 90 days before getting assigned to billable client work. When the payouts landed, it confirmed the lag wasn't a dead end. The referral pipeline just takes months to spin up, and once it does, it compounds.
The certified exam I launched in June brings in a small, steady stream: $100-$160 a month. It's nice validation, but it doesn't move the needle by itself, yet it acts as a cushion and form of revenue diversification that's much needed at this stage.
Most of my engineering time in July and August went into programmatic SEO, structured data cleanup, and trying to get Google to index my pages.
Google still won't index me properly. About 8 organic sessions a day, even after two months of that work.
That's usually a death sentence for a pSEO site. Bing and the LLMs, ChatGPT, Claude, Perplexity, are carrying real traffic instead, without any ad spend.
There was a massive spike in traffic in the summer, but roughly half of it appears to be bots.
Still, I have new people joining my talent pool every day (close to cross 1000 registered users after just 4 months from launch) and the feedback left by real humans on my academy modules is overwhelmingly positive, which inspires me to keep going.
The Google indexing problem is still very much my priority, and hopefully the fixes I made in July are gonna take effect soon. Google SEO is notoriously slow, and I can see that people are actively searching for my brand, but I won't be able to unlock the real potential of this platform until organic SEO spikes once and for all.
Beyond that, the focus stays on what is already proven to work:
Double down on top converting partners: Prioritize referral programs with realistic client-matching timelines and strong payouts.
Refine core products: Iterate on the Academy and Platform Intelligence tools based on active user feedback, fix usability issues and bugs as soon as they are discovered.
Backlog and product expansion: I soft-launched a spinoff job board for remote async jobs aimed at introverts last week. It serves as a quick experiment to test whether a fresh domain runs into the same pSEO indexation wall as my main site, and because nothing screams sound decision-making like launching a side project while already working 100% capacity on your main one.
The 90-day lag is a useful distinction: it sounds less like a demand problem and more like a cash-conversion problem. I’d track each referral cohort from referral → onboarding → first billable assignment, then show partners the median days and drop-off at each step; that should make the pipeline easier to forecast and reveal where a small onboarding nudge can pull revenue forward.
The $4.7k month is strong evidence, but most of it came from a 90-day referral backlog. Have subsequent referrals started converting on a similar timeline, or is the repeatability of that channel still unproven?