Six weeks ago Google had indexed 4 of my pages. It just hit 115. I changed almost nothing

Rakesh Kumar Maity — Tue, 04 Aug 2026 07:04:00 +0000

115. Six weeks ago I wrote here that Google had indexed 4 of my 122 pages, and I argued the problem wasn't technical, it was a crawl-budget gate on a new domain with no backlinks. A few commenters told me I was coping and should go fix my site. So here's the follow-up, with numbers, including the part I got wrong. The site is tamethebot.com, a free library of job-specific AI prompts, and here's the original post if you want the before.

Search Console, then and now:

	June 24	July 24
Indexed	4	115
Not indexed	122	14
of which "Discovered - currently not indexed"	119	0 (marked "Passed")

The "Discovered" bucket, the one I argued was a budget decision and not a quality one, went to zero. Google ran a validation pass on it and marked it Passed. If those pages had been thin enough to reject on quality, they'd have moved to the other bucket, "Crawled - currently not indexed." They didn't. They just got crawled and kept.

What I changed between the two screenshots

Almost nothing on the site.

I didn't rewrite the templates. I kept shipping roughly one page a day, same as before. The one real content experiment I ran was rebuilding a single prompt page to be genuinely deep: a worked example, outbound citations, a proper FAQ. Over the following month that page earned 8 impressions. The site averaged about 8.6 impressions per page. The deep rewrite landed at dead average. Depth was not the lever.

The thing that changed was off the site. I published that June post. It carried two dofollow links back to my homepage and sitemap, and it was, a little embarrassingly, my only real backlink. Then I waited. On a young domain, "wait" is an actual strategy, because the budget gate opens once there's a reason to spend the budget.

Can I prove the backlink did it? No. One site, an n of 1, and the domain also aged six weeks and kept getting recrawled. Age, the link, and normal crawl recovery are tangled together and I can't cleanly separate them. But the direction matches what I predicted, and no content change is even a candidate for the cause, because I didn't make one that moved anything.

The one thing I was flat wrong about

My June post had a confident line in it: "nothing was broken." One thing was. My IndexNow deploy hook, the bit that's supposed to ping search engines the moment a deploy goes live, had been silently failing for weeks. It compared the deploy's environment_url (always a *.vercel.app preview URL) against my production domain, so the check was always false and the ping never fired. I assumed it worked because I'd written it and it didn't throw an error. It just quietly did nothing, on every deploy, for a month.

Two honest caveats on that. IndexNow only pings Bing and Yandex, never Google, so this bug is not why Google woke up. And "code that runs without errors and also without effect" is the most humbling kind of bug there is, because nothing ever tells you.

Indexed is not the same as visited

The part I actually care about is whether indexing turns into people. GA4, last 30 days: Google organic sessions up about 163%. That's off a tiny base, so hold the champagne, but it's the indexing recovery becoming real visits instead of just green bars in a dashboard.

One caveat I'd have completely missed a month ago: a chunk of my "active users" are not users. The top cities in my analytics were Prineville, Boardman, and Council Bluffs. Those aren't markets, they're datacenter towns, the AWS and Google server farms. Some of my hard-won traffic is the crawlers I was so happy to finally attract. Real human traffic is smaller than the raw number, and pretending otherwise would make this post a lie.

The plot twist

The same 115 pages Google's indexer just accepted and stamped "Passed," Google's AdSense rejected. Twice. Same content, same site, "low value content."

I don't think either half of Google is wrong. The indexer asks "is this a real page a searcher might plausibly want," and for a free library of job-specific prompts, mostly yes. AdSense asks "is this a site with enough original depth and traffic to build an ad business on top of," and for a six-week-old templated library with one backlink, fairly no. Getting indexed was never the same bar as getting monetized. Month two is where that gap stopped being a theory and started being my P&L.

The lever that finally moved my indexing was a backlink I earned by writing a post about having no backlinks. I don't have a tidy lesson to close on. I have a Search Console that says Passed, an AdSense that says no, and a plan to keep writing in public until both of them say yes.

Google indexed 4 of my 122 pages. The problem wasn't technical

Rakesh Kumar Maity — Thu, 25 Jun 2026 07:17:01 +0000

Four. Out of 122 pages in my sitemap, Google has indexed four. Not crawled, indexed. The other 119 sit in a Search Console status I'd never read closely before: "Discovered - currently not indexed."

I run a small Next.js site, a free library of tested AI prompts (https://tamethebot.com). About 90 prompt pages, 15 profession pages, a few guides. It's six weeks old. When I saw "4 indexed" I assumed something was broken and went hunting for the bug. Most of what I checked turned out fine, and the real answer wasn't a bug at all.

The sitemap was fine

My first guess was a broken sitemap. App Router builds one from sitemap.ts, so I checked what Google actually receives:

curl -sI https://www.tamethebot.com/sitemap.xml | grep -i content-type

content-type: application/xml

Status 200, correct content type, 122 URLs, each with its own lastmod. Search Console even said "Sitemap processed successfully." Not it.

robots.txt had a real problem (just not this one)

This part is worth stealing. My robots.ts was doing this:

disallow: ["/api/", "/_next/", "/admin/"],

Blocking /_next/ stops Googlebot from fetching the CSS and JS it needs to render the page, which Google's own docs tell you not to do. I removed it. But it wasn't my indexing problem, because my content is server-rendered into the HTML, not painted on by client JS. That sent me to the next check.

The pages weren't thin, and weren't client-rendered

Next theory: maybe the pages rendered client-side and Googlebot saw an empty shell. They don't. Every prompt page ships about 1,000 words of real HTML, statically generated with generateStaticParams. I also measured whether the pages were near-duplicates of each other, since "templated" was the fear I couldn't shake. Two unrelated prompt pages shared about 11% of their text. Not duplicates.

So the sitemap was clean, the rendering was clean, the content was unique, the internal links were there, and the homepage was indexed. Nothing was broken. That's when the exact wording of the status finally clicked.

"Discovered" is not "Crawled"

Google has two different "not indexed" buckets, and they mean opposite things.

"Crawled - currently not indexed" means Google fetched the page and decided not to index it. A quality judgment.

"Discovered - currently not indexed" means Google knows the URL exists but hasn't fetched it yet. A crawl-budget decision.

I was in the second bucket. Google wasn't rejecting my content. It hadn't spent the budget to look. On a new domain with almost no backlinks, that's the default, not a malfunction. The clearest proof: my homepage is indexed and links straight to 6 prompt pages, and even those 6 hadn't been crawled. A clean path, still skipped.

What I took away from it

You can't request your way out of a budget gate. I submitted 15 pages by hand through URL Inspection and nothing moved. The real lever turned out to have nothing to do with code.

Authority comes before page count. Shipping 90 pages before earning a single backlink asks Google to spend crawl budget it has no reason to spend.

Read the literal status string. One word, "Discovered" versus "Crawled," saved me from a week of pointless technical fixes.

A correct technical setup is the floor, not a growth lever. My sitemap and rendering being perfect bought me nothing on their own.

I spent two days auditing my own machinery, and the machine was never the problem. The quiet wasn't Google turning me down. It was Google saying it hadn't looked yet, and I hadn't given it a reason to.