I built PayPerQR. Pushed it live. Bought a domain. Refreshed the dashboard.

Zero visits.

A week later, still zero. Two weeks, still zero. I started blaming the launch post, the positioning, the niche. Maybe QR codes were dead. Maybe nobody cared.

Then I built Toast Photos. Same thing happened. Zero. I told myself it was a marketing problem. Not enough promo. I needed more X posts, better hooks, a launch video, a Product Hunt day.

I was wrong. The problem wasn't marketing. The problem was that Google had no idea either site existed.

I had shipped two apps into a complete void, for months, and didn't realize until I finally sat down with Claude to figure out why my analytics were blank. Turns out SEO and analytics don't matter if you haven't done step zero. And step zero isn't in any of the content I'd been reading.

So this week I want to save you the two months I lost. Here's what I'd tell past me, what actually broke in my setup, and the exact AI prompts that walked me through the fix.

I'm not a marketer by training. I build things. So when people talked about SEO, I heard "write good content and Google figures it out."

That's half true. Google figures it out eventually, if it knows you exist in the first place.

Here's what I assumed:

Every piece of that was wrong.

Google doesn't find your site automatically. Google finds sites that other sites link to, plus sites that explicitly tell Google they exist. New indie domains with no backlinks fall into a third category: invisible. You can refresh your analytics all you want. There's literally nothing to measure, because Google's crawler hasn't been to your site. Which means none of your pages are in the index. Which means they can't rank for anything, because you're not a candidate for ranking in the first place.

No index. No rank. No traffic. No analytics signal. No clue what's wrong.

That's the trap. And almost nobody mentions it, because everyone writing about SEO assumes you already made it past step zero.

Quick detour because the terms get thrown around like they're obvious. They're not.

There are three things happening, in order.

Crawling. A bot shows up at your URL and downloads the HTML. That's it. The bot has been to your site.

Indexing. Google's systems read what the bot downloaded, decide it's worth storing, and add it to their giant searchable database. Now your page exists in Google's memory.

Ranking. When someone types a search, Google picks from indexed pages and orders them by relevance. This is the part everyone obsesses over. But it only applies to pages that got through the first two steps.

My apps were stuck at step zero. No bot had ever shown up. Nothing was in the index. Nothing was rankable. I was optimizing for step three while step one had never happened.

You need to tell Google about your site directly. That's what Google Search Console is for. It's not analytics. It's the front door where you hand Google a map of your site and say "please come read this."

I'd already been burned by this twice with PayPerQR and Toast. But I still had a third app, Distribution OS, sitting there with no Search Console setup, no sitemap, no robots.txt, and nothing telling Google it existed. So I sat down with Claude this week and walked through the whole setup live. That's what the screenshots below are from.

I'll be honest: I didn't want to read Google's documentation. I wanted someone to tell me what to click and what to paste. So I opened Claude, treated it like a patient engineer sitting next to me, and pasted my actual app into the conversation.

Here's exactly what happened.

This step is mostly UI clicks, not prompts. The flow is:

Mine went green instantly because Vercel's DNS propagates fast.

The first time I saw "Ownership verified," I actually said the word "oh" out loud. Because it was that easy, and I had spent two months not doing it.

The next thing Search Console wants is a sitemap. A sitemap is just a file that lists every URL on your site, in a format Google can read. Most frameworks can generate one for you. Mine didn't have one at all.

The prompt I typed:

Claude scanned my app routes and wrote both files. For Next.js 16 App Router, sitemaps go at src/app/sitemap.ts and robots go at src/app/robots.ts. Both are metadata routes — Vercel builds them automatically into /sitemap.xml and /robots.txt. No extra config.

Here's what the live sitemap looked like after the deploy — seven URLs covering all public pages, excluding the dashboard and API:

I submitted the URL in Search Console. First attempt showed a red error:

For about thirty seconds I thought something was broken in the sitemap I'd just shipped. Refreshed the dashboard. Different story:

Same submission. Two contradicting views. Turns out Search Console's outer list caches the first-attempt result, while the detail view shows the true state. The fetch actually worked on the retry a few seconds later, which is the normal behavior for a newly-submitted sitemap. Trust the detail view. If it says "processed successfully" with a discovered-pages count, you're good.

For the first time since buying the domain, Google had actually read my site. Seven pages discovered.

One more thing caught me. The robots.txt Claude generated was reasonable, but I had a rule I wanted to double-check.

The prompt I typed:

Claude spotted a subtle bug. I had written Disallow: /audit/*/ to block individual user audit result pages. Problem: that wildcard requires a trailing slash to match. Next.js App Router generates URLs like /audit/abc123 (no trailing slash) for dynamic routes. So the rule wasn't actually blocking what I thought it was blocking.

Fix: switch to an explicit-allow pattern.

Before:

After:

Two changes. The Disallow: /audit/ blocks the entire /audit/ tree, then two explicit Allow lines punch holes for the specific pages I actually want indexed (the main audit tool and the battle page). Claude also flagged that my Host: directive was Yandex-only; Google ignores it. Dropped it.

Rule of thumb I picked up from this: robots.txt is a suggestion to well-behaved crawlers. If you actually want to keep pages out of search, use a <meta name="robots" content="noindex"> tag server-side. That's enforced at the page level, not trust-based.

The last bug was the sneakiest. Some of my entry points were taking more hops than they needed to reach the canonical URL. Google handles up to about five hops without giving up, but every hop is a chance for the crawler to lose interest and bounce. Every flattened redirect is free crawl budget back.

The prompt I typed:

Claude traced every entry point and mapped the chain. Here's what came back:

Three of four were already clean. The only 2-hop entry was http://www.*, and it was 2 hops because Vercel's edge upgrades HTTP → HTTPS before running the domain rewrite. I could flatten the main www → apex path by explicitly configuring it in Vercel.

The fix — one config change: In the Vercel project → Domains → Add Existing → www.distributionos.dev. Then — and this matters — uncheck the default checkbox that says "Redirect distributionos.dev to www.distributionos.dev (Recommended)." Vercel's "recommended" default is the opposite of what I wanted. With that unchecked, select "Redirect to Another Domain" → target distributionos.dev → 308 Permanent Redirect. Save.

After that, https://www.distributionos.dev collapses to https://distributionos.dev in a single 308. The remaining 2-hop path (http://www.*) is unavoidable without turning off HTTPS enforcement, which isn't a trade I'd make.

Right now, 24 hours after everything was submitted, Search Console shows 7 pages discovered, crawl scheduled, indexing in progress. The real ranking numbers are a week or two away. But the plumbing is right for the first time in two months. No more shipping into a void.

Five things I wish someone had said to me before I shipped my first app.

Once your site is actually in the index, the next question is whether AI tools like ChatGPT, Claude, and Perplexity will cite you. Getting into Google is step one. Getting into the AI answer layer is step two, and it works a little differently.

That's what I’ll be testing next.

See you Tuesday, Hey

P.S. if you want a quick gut check, reply with your domain and I'll tell you if Google has indexed it. takes me about 30 seconds to look and it's the fastest win I can give you.

.