Two months ago I launched AJIEZ, a site with 106 free browser-based tools — PDF merging, mortgage calculators, a background remover that runs its ML model in your tab. No accounts, no uploads; everything happens client-side.
The build was the easy part. The humbling part was Google. A brand-new domain starts at zero: not ranked, not crawled, not indexed. Fourteen days after launch, 69% of my English pages were in the index. This post is the honest log of what moved that number — and what turned out to be a waste of time.
Where the clock started
The site went live on September 6, 2026 with a sitemap of a few hundred URLs (342 by September 20 — tool pages plus bilingual guides). Six days in, on September 12, Google Search Console reported 46 of 256 submitted URLs indexed, about 18%. Sitemap submitted, robots.txt clean, internal linking reasonable. And then… nothing much. Crawling a new domain is not a priority for anyone but you.
What didn't work: the Indexing API
The Google Indexing API looks like the obvious lever. I tested it three times. Every call returned 200 OK. Every call was silently ignored — not an error, not a quota message, just no effect on indexing. (This matches Google's own documentation: the API is intended for job postings and livestreams. It just fails quietly for everything else.)
If you're tempted: don't. That dead end cost me two evenings.
What worked: nightly submissions, until the wall
The unglamorous thing that actually worked was submitting URLs through Search Console's URL Inspection — by hand at first, then automated with a headless-browser job that runs at 2 AM, opens the inspection tool, and submits "discovered, not yet indexed" pages until Google stops accepting.
Running that job daily is how I found the quota, empirically: on September 19 it submitted 69 URLs successfully and the 70th was refused. Not documented anywhere I could find; the ceiling seems to depend on page state ("already discovered" pages are cheap) and the trust level of a young property. Sixty-nine a day is plenty when your backlog is three hundred pages.
The result: 18% on September 12 → 69% of English pages by September 20 (186 of 342 URLs overall counting the Chinese half, which lags at about 40% — more on that another day).
The queue matters as much as the volume
Sixty-nine submissions a day is a budget, so order matters. I split the sitemap by language and pushed the English half first, the Chinese half second — English is where the search demand and the tool keywords are for this site, and the Chinese pages are zero-maintenance long-tail that can wait their turn without hurting anything. Within English, tool pages outranked guide pages, because a tool page that gets indexed can rank for its query next week, while a guide needs links and age to compete.
If your sitemap has a "half that matters," submit that half first. The quota doesn't care about your priorities, so you have to.
The other battle: knowing which visitors are real
Here's the part nobody warns you about. A new domain's analytics are a bot festival. Before any human arrives, you'll see dozens of "visits": WordPress-scanners probing /wp-login.php, fake Chrome user-agents from datacenter IPs, and crawler volleys hitting five pages in sixty seconds.
So I built filters, and two of them did most of the work:
- The five-pages-in-60-seconds rule. No human reads five distinct pages of a tools site in a minute. Anything matching that pattern gets counted as a crawler, whatever its user-agent claims.
- The owner cookie. I marked my own devices once. From then on, my own testing stopped inflating numbers — on one loud debugging day, "traffic" was 1,392 visits, of which 4 were strangers.
After filtering, the honest baseline was 3–31 real visits per day (measured September 20). That's the number I make decisions on. The fake number would have made me think launch week was a triumph.
The Bing lesson: submissions aren't the bottleneck everywhere
I pushed every URL to Bing via IndexNow from day one. Twelve days later (as of September 23): Bing had crawled 45 pages, indexed exactly one, and its own API reported zero inbound links. Bing indexes what it trusts, and trust on a fresh domain comes from outside links. Google responded to submission + content; Bing is waiting for backlinks — and DuckDuckGo follows Bing, so that lane unlocks the same way.
The first real signals
Two moments made the grind feel real, both around September 20: the first person who found the site via Google search and came back a day later, and a mention in a developer newsletter's GitHub issues that brought a wave of actual humans — more convincing than any bot-swarm milestone.
What I'd do differently
Three things, in order of how much time they cost me:
- Start URL Inspection submissions on day one. I treated the sitemap as the submission and waited. The sitemap is a map, not a request; the inspection tool is the request. A week of passive waiting was a week of quota unused.
- Skip the Indexing API experiment entirely. The docs technically tell you it's for jobs and livestreams; I talked myself into believing "200 OK" meant something. It doesn't.
- Build the analytics filter before launch week, not after. For several days my mental model of the site's traction was almost entirely fictional — scanner traffic wearing a visitor costume. Ten minutes with a cookie and a pacing rule would have fixed it on day one.
The meta-lesson: on a new domain, indexing and truth are both manual projects. Google won't hurry, and your dashboard won't self-correct. Both needed deliberate, slightly tedious systems — a 2 AM submitter and a noise filter — before the numbers meant anything at all.
Takeaways
- Indexing API: dead for normal sites. Three tests, silent no-ops.
- URL Inspection submissions work, and the practical ceiling for a young property was 69/day in my test — enough to clear a small sitemap in under a week.
- Filter your analytics before reading them. Five-pages-a-minute and an owner-cookie cover most of the noise; a "spike" that isn't humans will mislead your roadmap.
- Google indexes on submission; Bing waits for links. Same effort, different currencies.
- Give crawlers something worth keeping. 106 real tools and 55 guides is why submitted pages stayed indexed instead of bouncing back out.
The site is ajiez.top — the full tool index is one page, and the guides hub has the longer how-to writeups. A Product Hunt launch is coming on October 14; if the cold-start log was useful, that's where the story continues.
Top comments (0)