The two answers come up about equally often. Across 35 readings of the seven addresses that are not the front page, search called them “discovered” 18 times and “unknown” 17 times, with 13 flips back and forth and zero readings of any indexed state at all. The front page, meanwhile, said the same thing one way in all five readings. There is a real boundary here and it is stable — but it runs between the front page and everything else, not between one step and two.
One note on how that was counted, because getting it wrong would have been easy and flattering. This article is a ninth address that did not exist when the table below was taken, and it wavers too — three flips of its own. Including it would have made the split an exact 20 and 20, which reads better and answers a different question: it is not a reread of the same pages. The figures above are the original eight. The ninth is reported here and kept out of them.
That leaves a shorter table, and this one I can stand behind:
| Clicks from the front page | Pages | Index |
|---|---|---|
| zero | one | in the index, five readings out of five |
| 1 and 2 | seven | not in the index, 35 readings out of 35 |
Two consequences. The claim that a second step is worse than a first has nothing behind it: on this measurement the two are indistinguishable, and the closing paragraph about articles falling off the front page should be read as “out of the index like everything else”, not “further out”. And the piece below is one long argument that a passing check is a claim until somebody reads it twice — which turned out to be true of its own table. I have left that table where it was, because a correction that deletes what it corrects is the same move under a nicer name.
What is unaffected: the distances themselves, which come from walking the live site and never touch Search Console; the fact that the one page linking to the second page of the index has never been fetched, which is read out of our own markup; and everything below about who links here, which came from a different field and a different call.
What I published
On 2026-09-08 this site had eight addresses in its sitemap, every one a real page a person can open. Search had fetched one page. Not rejected the rest, not judged them thin, not weighed them against anything — never asked for them at all. That is a different failure from the one I spent a week assuming, and it took a measurement to tell them apart. (The count is already out of date: this article is the ninth address.)
The measurement is two columns. For each address: how many clicks it takes to reach from the front page, and what Search Console says about it. I expected a smear — some pages doing better than others for reasons I would have to guess at. What came back has no exceptions in it at all.
| Clicks from the front page | Pages | What search says |
|---|---|---|
| zero | one | Submitted and indexed |
| one | six | Discovered — currently not indexed |
| two | one | URL is unknown to Google |
One step out and search knows the page exists but has never fetched it. Two steps out and it has never heard of the page. I want to be careful about what that does and does not establish: distance is the only column I have that varies with status, and within these eight addresses it varies with it perfectly. It is not a demonstration that length or age or topic play no part — I did not measure those, and with a single page at each end of the range I could not separate them from distance if I had.
The page two clicks away has everything except a reader
The single address at distance two is the second page of the blog index. It is worth naming what it already has, because the list is the whole argument: it sits in the sitemap with a current date, it answers 200 in 0.72 seconds, robots.txt allows it, and it names itself as its own canonical. Everything a checklist asks for is there.
Exactly one page on this site links to it — the blog index, two anchors and a rel=next, all from that one page. And the blog index is itself a page search has discovered and never fetched. So every road to it runs through a page nobody has driven down, and no amount of sitemap correctness moves that.
The front door has not opened since the first of the month
The front page is the one address that has been fetched, so it is the only door. Its last fetch reads 2026-09-01T08:23:31Z. I have now read that field ten times between 2026-09-07 and 2026-09-08 — eight of them inside 17 minutes, logged, because a single reading of a live system is an anecdote. The value never moved.
Which would be an ordinary story about a quiet site, except that four articles have gone up since that fetch, and each one adds a dated block to the front page. The door changes every time. Nobody has come back to look.
What brought anyone here in the first place
So I went and asked the other question, which I had been avoiding because I suspected the answer. How does search know this domain exists at all? The inspection names the referring pages, and for the front page it names two pages. The first is blog.barinov.pro — a subdomain of this same domain with no DNS record whatsoever: nothing resolves, nothing answers. The second is a company directory that returns 403 to us, so we have not read it and will not describe it.
For the articles themselves the field is empty. Not one of them has a referring page of any kind. And over the last 90 days, search reported 38 impressions and one click, on one page.
The queries are where I nearly wrote something false. The console named exactly one query — vladimir barinov, my own name — and the sentence that wants to be written is that this site appears only for the name of the person who runs it. Its own numbers refuse that. The named query accounts for four of the 38. The other 34 came from searches the console will not disclose, across 13 countries — which is not the shape of people looking up a name they already knew. What is true is narrower and duller: one query is all the console will name.
Then the check that made me rewrite this piece
Before publishing any of the above I went to fetch the outside pages this site itself names — the interviews and the profile links in its own markup — and looked for a link back. Of the four pages I tried, two refused to be read by anything automated: one answered 403, the other answered 202 with zero bytes in it. About those two I therefore state nothing at all. Not "no link" — unknown. A tool that cannot read a page has learned nothing about it, and the honest column is blank.
Of the two I could read, one carries an ordinary editorial link straight to this domain: weproject.media. The other names the person in its text and links nowhere.
What I actually take away
Publishing is not the same as existing, and I had been running the two together. Nothing about the pages was wrong. The dates, the canonicals, the structured data, the feed, the response times — all of it correct, all of it checked, none of it the constraint. The constraint is that almost nothing in the world points here, and the one internal path we do control gets thinner the further you walk.
The part in our hands is the second half of that, and it is unglamorous: a page that can only be reached through a page nobody reads is not published, whatever the sitemap says. The front page currently lists every article we have ever written, which is the only reason all of them sit one click away. That will not survive a daily cadence, and on the day it stops being true, every article that falls off the list arrives where the second page of the index is today: at distance two, where search has never heard of it.