A shop we work on publishes in five languages and has 6,534 URLs in its sitemap. Search Console showed traffic arriving at a few dozen of them. The obvious question, which page of the other six thousand is working, had no obvious answer.
Why the coverage report is not the answer
Search Console's coverage report is a sample and a summary. It tells you that a number of pages are "Discovered, currently not indexed" and gives you a few examples. It does not hand you the list, and the examples change between visits. For a site with a handful of pages that is fine. For a sitemap with thousands, it is a statistic about a problem rather than the problem.
The URL Inspection API is different. You give it one URL and Google answers for that URL: is it indexed, when was it last crawled, which canonical did Google choose, was it fetched at all. The quota is 2,000 URLs a day per property, which is the reason nobody does this by hand.
What came back
We inspected 2,181 of those URLs between 12 and 18 September 2026:
| State | Pages |
|---|---|
| Unknown to Google | 1,401 |
| Indexed | 731 |
| Excluded by noindex | 40 |
| Crawled, not indexed | 8 |
| Redirect | 1 |
Sixty-four percent had never been fetched. Not penalised, not judged thin, not competing with a better page. Simply never seen, while sitting in a sitemap that had been submitted months earlier.
Unknown is a different problem from not indexed
This distinction is the whole point, and it is the one most tools blur.
A page that was crawled and not indexed has been judged. Google fetched it, read it, and decided it was not worth keeping. The fix is editorial: make the page better, or accept it.
A page that is unknown has not been judged at all. Google never arrived. The fix is mechanical: link to it from a page Google already crawls, check it is in a sitemap that is reachable and submitted, and make sure nothing in robots.txt stands in the way.
Those two need opposite work, and a tool that files them under one heading sends you rewriting pages that nobody has read.
We got this wrong ourselves, briefly. Our own detector counted pages whose fetch state Google reported as unknown as fetch errors, which put 1,399 of them in a bucket labelled "Google could not fetch this". They were not errors; Google had simply never tried. One class now says what it means, and the fix text with it.
What we did with the list
Once you have the list, the work sorts itself. Pages nobody links to got links from the pages that do get crawled. Pages that duplicated a better page were pointed at it. The rest went into the sitemap check: submitted, reachable, and no rule blocking the crawler.
Then the same inspection runs again, and the number moves or it does not. That is the part a score out of 100 cannot give you.
If you want the same list for your own site, the coverage report here runs on any public sitemap without an account.