Googlebot fetched our blog index after 58 days and again 19.6 hours later: 1 of the 14 posts it links is indexed 6 days on, 13 have 0 Googlebot fetches in 7 days, and PerplexityBot fetched 8 of the 14

What was shipped, and what was written down before it shipped

By 2026-08-27, 14 of our blog posts, published 2026-08-19 to 2026-08-27, had sat in Search Console as "Discovered, currently not indexed" or "URL is unknown to Google" with no Googlebot fetch on our ledger. Our earlier test of the Request indexing button is in the 2026-08-28 post; the button is not a lever we plan on. On 2026-08-28 we shipped one variable instead: a rebuilt /news index grouped by month, with rel="next" and rel="prev" links on every post, so that each of the 14 is reachable from a page Google already indexes. The 14 URLs were pinned in a fixture that day. The read was pre-registered for 2026-09-11 with one rule: at least 7 of 14 indexed means the index works; 0 of 14 means internal linking is not the constraint; anything between is inconclusive and is reported as such. One instrument makes the read every day: it inspects each URL through the Search Console API and counts Googlebot rows on our server ledger for the same path over the trailing 7 days.

The index page itself: two channels, one second

The index page had not been crawled by Google since 2026-07-06 18:12:19 UTC (the lastCrawlTime in our 2026-08-23 sweep of every URL) and had 0 Googlebot rows on the ledger from the ledger's birth on 2026-08-17 to 2026-09-01, in 22 rows from all crawlers. Then:

The two 2026-09-03 rows are 49 ms apart, the shape of a desktop and a smartphone Googlebot pair. Search Console still reported the 2026-09-02 crawl at the read; the API's lastCrawlTime lags the fetch itself, which the ledger sees at once. Rows on /news by crawler in the same 7 days:

The 14 linked posts, seven daily reads

Each row is one run of the instrument at about 16:40 UTC. "Flips" counts URLs whose Search Console state differed from the previous day's read. The one URL that went in did so on read 3 (2026-08-30); its lastCrawlTime is 2026-08-30 04:29:48 UTC and it carries the only Googlebot rows on the denominator.

45 state changes across 6 transitions on 14 URLs, and 1 of them is a change in indexation. The rest is the API alternating between "Discovered" and "unknown" for the same never-fetched URL; every not-indexed URL has flipped at least twice. The per-URL view, with the state on each read as a letter (D discovered, U unknown, I indexed):

Where Googlebot went instead

Googlebot is not absent from the site. In the 24 hours to the read it made 38 requests to 18 distinct paths, 2 of them the index page and 0 on the 14 posts the index links. In 7 days: 280 requests to 60 paths, 2 on the denominator (both on the one indexed post), 0 on the five posts published since 2026-09-01. The 92 requests to /news posts covered 36 distinct posts, published 2026-07-09 to 2026-08-22; the newest of them is 12 days older than the newest post on the site (published 2026-09-03).

Who did fetch the 14 posts in the same 7 days

8 of the 14 were fetched by at least one AI-engine crawler (PerplexityBot 8, OAI-SearchBot 3, ClaudeBot 1, GPTBot 0). These are re-fetches: our 2026-09-01 read found ClaudeBot, OAI-SearchBot and PerplexityBot had each fetched all 17 of the posts published 2026-08-24 to 08-31 within days of publish. The Google side of the same posts has not moved: 13 of 14 with 0 Googlebot rows in 7 days.

The five newest posts, for the record

Outside the pre-registered denominator, the posts published 2026-09-01 to 09-03 were inspected in the same run: 4 read "Discovered, currently not indexed", 1 reads "URL is unknown to Google", and all five have 0 Googlebot rows in 7 days. They will be their own series; they are not added to the 14.

What this means if you care about AI visibility

• An index page being crawled and a linked post being crawled are different events. Googlebot read our index three times in 20 hours and, by the read, had followed none of the 13 links to posts it had never fetched. If a client's blog is "Discovered, not indexed", a crawl of the blog index does not by itself change that; check the posts, not the index.
• Search Console's coverage state is not a daily signal. The same never-fetched URL read "Discovered" and "unknown" on alternate days, 45 flips in six days across 14 URLs. Count indexed against not indexed, and read lastCrawlTime; anything finer is noise.
• The AI engines' crawlers do not wait for Google. PerplexityBot re-fetched 8 of these 14 posts in the same week that Googlebot fetched 1. Whether an engine then names your business when a buyer asks is the count our free 60-second check reads per engine; being crawled is where it starts, not where it ends.

Limits

• One site, 14 URLs, one week after one change. The read is a case, not a rate.
• The ledger records requests whose user agent declares a crawler; it does not verify the address. Googlebot rows on it are self-declared. The one place it is checked against Google's own record, the index page's crawl, the two agree to the second.
• URL Inspection has a daily quota, so the instrument reads the 14 URLs plus 11 control URLs once a day at about the same hour; a fetch and an indexation between two reads show up on the next one, not at the minute.
• Search Console's lastCrawlTime lagged the ledger by at least the 6 hours between the 2026-09-03 fetches and the read; the ledger sees the request, the API reports Google's record of it later.
• Age is counted from the publish date in our registry, not from Google's first discovery, which we cannot see.

The next reading, pre-committed

On 2026-09-11 the same instrument runs on the same 14 URLs and the verdict is scored by the rule written on 2026-08-28: at least 7 indexed, the index works; 0, internal linking is not the constraint; between, inconclusive. It is published either way. Two lines are falsifiable before then: if Googlebot rows appear on any of the 13 zero-row posts within 7 days of the 2026-09-03 index fetches, they are named in the 09-11 post with their timestamps; if none do, the sentence "Googlebot read the index three times and followed none of its links to unfetched posts" is restated at 14 days.