How I Built an AI Visibility Analytics Tool in 4.5 Months as a Solo Founder (And Why It Matters Right Now)

I started building Loamly in August 2025 while working a full-time job at FELFEL in Zurich. Four and a half months later, I launched on Hacker News and Product Hunt on December 30-31. I got 2 free signups.

That's it. Two.

This is the unfiltered story of what building a SaaS solo actually looks like—not the highlight reel.

The Problem I Couldn't Ignore

In mid-2025, I noticed something strange in my analytics.

Traffic patterns were shifting. "Direct" traffic was growing faster than anything I was doing could explain. Branded searches increased without corresponding campaigns. Users mentioned finding me through "search" but couldn't recall specific queries.

The culprit became obvious: AI.

ChatGPT, Claude, Perplexity—these platforms were becoming primary discovery channels. But my analytics tools (GA4, Mixpanel, traditional stuff) were completely blind to it. They couldn't tell me:

• What AI platforms were saying about my brand
• How much traffic actually came from AI recommendations
• Whether AI mentions were positive, negative, or neutral
• How my AI visibility compared to competitors

I looked for tools that solved this. There wasn't anything that really did it well. So I started building.

Why I Chose to Build Nights and Weekends

I work a full-time job at FELFEL in Zurich. I'm not a VC-backed founder with runway. I'm not a rich person with connections. I don't know "10 CEOs to talk to this week."

What I have: nearly 10 years of work experience, understanding of systems and processes, CRM expertise, and the ability to build anything better than top engineers of the world. Plus, I have all the credits I want from Clay.com thanks to my job and signed up 20 email accounts across 4 different domains in Instantly.ai.

The decision framework:

1. Validate before committing full-time. I wanted to prove the problem was real and that I could solve it before leaving my job. Nights and weekends let me test the hypothesis without burning bridges.

2. Use constraints as advantages. Limited time forced ruthless prioritization. I couldn't build everything, so I had to build only what mattered. This prevented feature bloat.

3. Test distribution early. I could experiment with different channels (HN, PH, Indie Hackers, dev.to) without the pressure of needing immediate revenue to survive.

4. Learn while building. Every late-night coding session taught me something about the problem space. By the time I launched, I understood the domain deeply.

The reality: building nights and weekends means 15-20 hours per week of actual coding time. That's roughly 60-80 hours per month. Over 4.5 months, that's 270-360 hours total. Not much, but enough if you're focused.

The 4.5-Month Sprint: What Actually Happened

Month 1: Foundation (August 2025)

Critical early decision: Use a professional boilerplate.

I started with:

• Next.js 15 with App Router
• Supabase for auth and database
• Tailwind + shadcn/ui for design
• Vercel for deployment

The first month was all infrastructure. Auth flows, workspace setup, basic navigation. Nothing exciting. No AI detection yet.

Reality: Zero users, zero revenue, lots of boilerplate setup.

The unglamorous truth about building in public: most of the early work is plumbing. Nobody wants to read about setting up database migrations or configuring OAuth flows. But this foundation work is what makes everything else possible.

What I got right: Using a boilerplate saved me weeks. As Two Cents Software's solo founder guide puts it: "Combined with AI coding tools, this approach typically delivers a launched MVP in 8-10 weeks of part-time work."

What I got wrong: I over-engineered the database schema. Built for millions of users when I had zero. Should have started simpler.

Month 2: First AI Detection (September 2025)

This is where it got interesting.

I built the first version of AI traffic detection. The approach:

1. Parse incoming traffic referrers
2. Match against known AI platform patterns (chat.openai.com, perplexity.ai, etc.)
3. Store and visualize the data

The initial detection was basic—just looking for referrer strings. But it worked. I could see actual ChatGPT traffic hitting websites.

The first breakthrough: A test site I ran showed 3% of traffic coming from AI sources. Traffic that would have been invisible in traditional analytics.

3% might not sound like much. But for some sites, that's the difference between understanding your growth and being completely confused.

The first mistake: I over-engineered the dashboard. Spent two weeks building visualizations that nobody needed yet. Should have shipped faster.

Reality: Still just me testing on my own sites. No external users yet.

Month 3: The Pivot (October 2025)

October was humbling.

Beta users told me the traffic detection was interesting, but what they really wanted was to see what AI was saying about them. Not just how much traffic—but the actual content of AI responses.

This meant building something completely different: an AI querying system.

Instead of just tracking traffic:

• Query ChatGPT, Claude, Perplexity, and Gemini
• Parse their responses for brand mentions
• Track sentiment and positioning
• Compare against competitors

I essentially had to throw away a month of work and rebuild around the new core.

Lesson learned: "Solve a real problem" isn't enough. You need to solve the right version of the problem. My early users showed me I was solving for the symptom (traffic detection) when they wanted insight into the cause (AI visibility).

Reality: One person signed up in September. Still mostly building in isolation.

Month 4: The GEO Score (November 2025)

With the AI querying system working, I needed a way to communicate results clearly.

Enter the GEO Score.

The idea: a single number (0-100) that captures your Generative Engine Optimization performance. Like a PageSpeed score, but for AI visibility.

Building the GEO score meant:

• Defining what factors matter (content clarity, schema markup, E-E-A-T signals)
• Building scoring algorithms for each factor
• Weighting and combining into one metric
• Making it actionable with recommendations

The hardest part: Getting the weights right. Initial versions produced weird results—sites with obvious AI visibility issues scored high, while well-optimized sites scored low.

The fix came from analyzing actual AI citations. What patterns correlated with getting mentioned by ChatGPT? That data informed the scoring model.

Another mistake: I tried to make the score too comprehensive. First version had 15 factors. Trimmed it to 6 core factors that actually matter.

Reality: Second signup happened in November. Two users total. Still no revenue.

Month 5: Public Launch Prep (December 2025)

December was about polish. The product worked. Now it needed to look professional.

Tasks completed:

• Complete UI redesign with consistent design system
• Landing pages for SEO/GEO keywords
• Documentation site
• Pricing page
• Legal pages (privacy, terms, GDPR)
• Email system for reports

The PageSpeed grind: I spent an embarrassing amount of time optimizing PageSpeed scores. Target was 100/100. Achieved it. Worth it? Probably not, but it felt good.

SEO and GEO optimization for the marketing site itself became a mini-project:

• JSON-LD schema for Organization, WebApplication, FAQPage
• Comprehensive meta tags
• Sitemap with proper priorities
• robots.txt with AI crawler permissions

December launch:

• Launched on HN and Product Hunt on Dec 30-31
• Posted on Indie Hackers and dev.to
• Two more signups in December (4 total users)
• Zero paying customers yet

2 Signups on Launch Day: Learning to Handle Silence

I launched on Hacker News and Product Hunt on December 30-31, 2025.

I got 2 free signups.

That's it. Two.

What I expected: Based on other launch stories, I thought I'd get at least 50-100 signups. Maybe some paying customers. Maybe some press coverage.

What I got: Two signups. No paying customers. No press. Just silence.

How I felt: Honestly? Disappointed. I'd spent 4.5 months building this thing. Nights and weekends. Hundreds of hours. And the response was... crickets.

What I learned:

1. Launch day doesn't matter. The real work starts after launch. Those two signups became my first real users. I talked to them. Learned from them. Built features they actually needed.

2. Distribution is everything. I built a product in stealth mode for 4.5 months. I don't have a network. I don't know VCs or "10 CEOs to talk to this week." I work a full-time job. The playbooks that assume you have time and connections don't work for me.

3. The product works. The two paying customers I have now are actively using it. The /check tool generates real insights. The problem isn't product-market fit—it's that nobody knows Loamly exists.

4. Cold outreach is the answer. I have Clay.com credits from my day job and 20 email accounts warming up in Instantly.ai. The plan is to pre-generate AI visibility reports for prospects and email them. Show value first, then ask for the sale.

The conventional founder narrative of explosive early growth is statistically an outlier. Most successful products experience slow initial growth, plateau, then breakthrough only after substantial strategic adjustment or positioning refinement.

From 2 to 8 Users in Three Weeks: What Changed

Between December 31 and January 11, I went from 2 users to 8 users. Four more people signed up. Two of them converted to paid. That's a 25% conversion rate from signup to paid, which is good. But 8 total users after a launch is... humbling.

What actually worked:

1. Talking to the first two users. I asked them what they needed. Built features they requested. They became advocates. One of them brought a friend.

2. The /check tool. I built a free AI visibility checker that doesn't require signup. People use it, see value, then sign up. It's working as a lead magnet.

3. Blog content. I started writing about AI visibility, GEO, and attribution. Every blog post is also a distribution channel. People find the content, try the tool, then sign up.

4. Indie Hackers and dev.to posts. I wrote honest posts about building Loamly. Not polished. Not perfect. Just real. People resonated with the authenticity.

What didn't work:

1. Hacker News launch. Got minimal traction. Probably because I'm not well-known in the HN community. Should have engaged more before launching.

2. Product Hunt launch. Got 2 upvotes. Embarrassing, but honest. B2B tools don't do well on PH unless you have a network.

3. Twitter/X. I don't have a following. Tweets got zero engagement. Need to build audience first.

The key insight: Growth compounds when you optimize for quality users rather than quantity. Those 8 users are engaged. They're using the product. They're giving feedback. That's more valuable than 80 disengaged signups.

What This Means for Other Solo Founders

If you're considering building a SaaS solo, here's what I learned:

You can build sophisticated products in 4-5 months part-time. The constraint of limited time forces better decisions. You can't build everything, so you build only what matters.

Launch earlier than you think you should. I waited 4.5 months. Should have launched after 2 months with a basic version. The feedback would have saved me a month of building the wrong thing.

Talk to users even when it feels awkward. Those first two signups became my most valuable source of feedback. Every conversation taught me something.

Don't compare your Day 1 to someone else's Year 3. Other founders sharing $50k MRR screenshots while you're at $0? Their journey also took years. The highlights reel isn't the full story.

Distribution is harder than building. I spent 90% of time building, 10% on marketing. The successful solo founders do 50/50. I'm correcting this now.

Constraints are advantages. Having a full-time job means I don't need revenue to survive. That removes pressure and lets me focus on building something people actually want.

Why This Matters Right Now

The AI visibility problem is urgent. Companies are losing visibility into how customers discover them. Traditional analytics are blind to AI-driven traffic. Marketing teams can't optimize what they can't measure.

But here's what makes this moment special: we're at an inflection point.

In 2026, companies will realize that AI search visibility is not a supplementary channel but a primary channel. By 2027, companies will allocate marketing budgets based on visibility across five search channels (Google traditional, Google AI Overviews, ChatGPT, Perplexity, Claude) rather than treating Google organic as the primary channel.

The companies that established dominance in AI search in 2026 will maintain competitive advantage for years because LLM training and trust networks compound over time.

Building Loamly now—before the market fully understands the problem—gives me a head start. I'm not competing against established players. I'm defining the category.

The Honest Summary

Four and a half months in, here's where I stand:

• Product: Works. Genuinely useful. The /check tool is solid.
• Users: 8 total. 2 paying. Not enough.
• Revenue: Under $100. Not covering costs.
• Personal cost: Hundreds of hours, lots of 2 AM sessions.
• Biggest mistake: Building too long before launching.
• Outlook: I'm just getting started.

Building a SaaS solo while working full-time isn't romantic. It's coding until midnight, then waking up at 6 for your real job. It's your partner asking why you're still on the laptop. It's launching and hearing crickets.

But I built something. It works. Two people are paying for it. And I have a plan for the next phase: stop building in isolation and start putting Loamly in front of people who need it.

If you're considering the solo founder path, here's my honest advice: launch earlier than you think you should. Talk to people even when it feels awkward. And don't compare your Day 1 to someone else's Year 3.

---

Want to see what Loamly does? Try the free AI visibility check and get your GEO score. No signup required.