This atlas, measured on itself

What we are solving

Every guide promises results in a few weeks. Almost none of them show what the first days look like, so when your own project is silent you have nothing to compare that silence with.

This site went live on 10 August 2026. Here is everything it has two days later, and I will keep adding to this page monthly, including the months where nothing moved.

The first thing worth writing down arrived sooner than a month. So the tables below carry a second reading, taken on 13 August.

Steps

I did what the route tells you to do, on this site. In the order the route gives.

1. Deployed the site — 58 pages in English and Russian, 56 URLs in the sitemap.
2. Generated robots.txt and llms.txt at build time. Neither of them can drift away from the pages it describes.
3. Verified a domain property in Search Console. Google confirmed ownership through the DNS provider and never asked me for a TXT record.
4. Submitted the sitemap once, on 12 August at 10:20 UTC. Search Console read it two seconds later and reported 56 URLs.
5. Pushed all 46 canonical URLs through IndexNow — the whole site at the time — and the shared endpoint, Bing and Yandex accepted them. Three more pages shipped later that day, in both languages, and those I have not pushed.

Then I stopped and read what came back.

What did not work

• Expecting the sitemap to do more than it does. It was accepted in two seconds, which felt like progress. Two days later URL inspection called the home page unknown to Google. Submitted, discovered, crawled and indexed are four states, and both readings sit in the table below.
• Reading the access log as if every line were a visitor. The log holds 417 requests. Almost all of them are me and the two people I sent the link to.
• Counting bots with grep -o. It counts occurrences and I read them as requests. Some agents name themselves twice on one line: bingbot writes its token in the string and again in the bingbot.htm URL, so its count doubles.

This log survived it, because Googlebot says its name once beside a bot.html URL — but I had already published a doubled Bing figure elsewhere. Count lines, not matches.

grep -o 'Googlebot' /var/log/nginx/atlas.access.log | wc -l
grep -c 'Googlebot' /var/log/nginx/atlas.access.log

grep -c is the honest count. I had already published the doubled numbers on another page before I noticed.

• Publishing llms.txt because the checklist said so. Zero AI agents fetched it here. On the older project it was zero too, and of the 45 requests that file ever got, 44 were my own curl.

Verify

I read these numbers on 12 August 2026, two days after launch. I read them again on 13 August, three days in.

What	12 August	13 August
Impressions in Search Console	0	0
Clicks	0	0
Queries the site appears for	0	0
Home page status	URL is unknown to Google	Submitted and indexed
URL inspection verdict	NEUTRAL	PASS
Last crawl of the home page	never	2026-08-13T09:08:32Z

The performance figures cover 10 to 13 August. Impressions, clicks and queries are all still zero.

Between those two days the home page changed state: Google had never heard of that URL, and now it has crawled and indexed the page.

The site: query returns the home page, and Tools, Cases and Indexing with it:

Google results for site:atlas.smolevich.com, read on 13 August 2026. The home page, Cases and Indexing carry the date 1 day ago, the Tools page 7 hours ago.

So three days from launch this site is indexed and invisible at the same time: submitted, discovered, crawled and indexed are four different states, and passing indexed moved nothing in the performance report.

What it bought me is one thing: an impression is now possible at all, and on 12 August it was not.

The size of the site moved between two readings on 12 August, so both readings are here, and three pages shipped later the same day, in English and in Russian.

What	First reading	Read again after those pages
Pages published	58	66
URLs in the sitemap	56	66

The right-hand column is the live sitemap.xml and the page count of the same build — everything else above is the first reading, and I have not taken it again.

Requests by search and AI agents, counted one match per line:

Agent	Requests
Googlebot	6
PerplexityBot	1
GPTBot	1
ClaudeBot	1
ChatGPT-User	1

What Googlebot actually took: robots.txt three times, both sitemap files once each, and exactly one content page — /geo/citable-pages/.

That last line is the one I keep coming back to: Google has read the map of this site three times and fetched a single page off it. Nothing here is broken. Nothing here has happened yet either.

If your own project is two days old and quiet, this is what quiet looks like from the inside.

Next measurement in a month, on this page, whatever it says.