"Discovered, currently not indexed": a diagnosis, in the order to check it

What the status says, and who said it

Three kinds of statement are involved, and they carry different weight. This table keeps them apart.

The two reported statements come from secondhand write-ups, not Google documentation, and the 2018 one concerned a site with an enormous page count. They are the best public evidence of what happens on real sites, so the checks below follow them. Treat the order as a diagnostic sequence, not a ranking of causes by frequency, which nobody outside Google has.

Check 0: is the URL just new?

New URLs wait. Google's own recrawl help says crawling "can take anywhere from a few days to a few weeks". If the page was published in the last couple of weeks and is in your sitemap, the right move is to look again in a few days. The Page indexing report only updates when Google crawls a page with known issues, so a stale count is not evidence of anything.

Check 1: is the server making Google back off?

Google's guide says the crawl limit goes up when response times, including time-to-first-byte, stay stable or improve, and down when the site slows. Measure it from outside a few times and compare:

curl -o /dev/null -s -w "TTFB: %{time_starttransfer}s  total: %{time_total}s\n" https://example.com/your-page/

Then look at the Crawl stats report in Search Console settings for response time and server errors. A host that cannot keep your app running will show slow first requests; the four measurements in the shared-hosting guide separate cold starts from CPU throttling from slow queries. Fixing a slow origin helps crawling and everything else.

Check 2: how does Google find these URLs?

Mueller named weak internal linking in 2018. A URL that appears only in a sitemap, with no page linking to it, tells Google it exists but not that it matters. For a static site you can count inbound links yourself. This script reads your HTML files and prints each page with the number of other pages linking to it, fewest first:

import pathlib, re, sys

root = pathlib.Path(sys.argv[1]).resolve()
pages = list(root.rglob('*.html'))
inbound = {p: 0 for p in pages}

for page in pages:
    html = page.read_text(encoding='utf-8', errors='ignore')
    for href in set(re.findall(r'href="([^"#?]+)', html)):
        if re.match(r'^(https?:|mailto:|tel:|//)', href):
            continue
        target = (page.parent / href).resolve()
        if target.is_dir():
            target = target / 'index.html'
        if target in inbound and target != page:
            inbound[target] += 1

for page, n in sorted(inbound.items(), key=lambda kv: kv[1]):
    print(n, page.relative_to(root))

For a CMS or a site you cannot read as files, a crawler gives the same answer. The SEO audit checklist lists Screaming Frog, whose free tier covers small sites.

The fix is not a link block on every page. It is a contextual link, in a sentence where a reader would want it, with descriptive anchor text, from pages that are themselves crawled often: the home page, section hubs and your most-linked guides.

Check 3: are there more URLs than pages?

Mueller's other 2018 example was accidental URL variations: parameters, upper and lower case, trailing slashes. Google's guide says the same from the other direction: eliminate duplicate content "to focus crawling on unique content rather than unique URLs". Open the Discovered list in the Page indexing report, export it, and group it by pattern. This counts URLs by first path segment and whether they carry a query string (adjust the URL column name to match your export):

import csv, collections, urllib.parse, sys

counts = collections.Counter()
for row in csv.DictReader(open(sys.argv[1], encoding='utf-8-sig')):
    u = urllib.parse.urlparse(row['URL'])
    first = u.path.strip('/').split('/')[0] or '(root)'
    counts[(first, 'with query' if u.query else 'clean')] += 1

for (first, kind), n in counts.most_common(15):
    print(n, first, kind)

A few hundred ?sort= or ?ref= variants of one page, or tag and archive pages that repeat each other, point to the cause. Fix them at the source with canonical tags and by not linking to the variants. Where several variants have to stay live, make the signals agree. If you decide some URLs should disappear, follow the removal order in the noindex vs disallow guide, not the reverse.

The Search Console API has no endpoint for this bulk report, which is why the export is manual. The Search Console extension can inspect single URLs, which is useful once you have a short list.

Check 4: are the pages worth crawling soon?

This is the uncomfortable one. "Not worth the effort, at least at the moment" was Mueller's phrase in 2018, and in 2026 he said these statuses can mean Google's systems are seriously worried about a site's quality, and that the response is to think about quality overall rather than fix URLs individually. Ask of each stuck page: would it earn a visit on its own? Is it distinct from a page you already have? Does it have a named author, dates and sources? The audit checklist has the trust checks. The usual action is to merge, improve or remove pages, not to resubmit them.

Help Google prioritise the pages you do want

• Link to them from pages Google already visits often, with anchor text that says what the page is about.
• Keep the sitemap honest. Google reads it regularly, so include the content you want crawled, leave out junk URLs, and set an accurate lastmod only when content really changed.
• Reduce the junk. Fewer duplicate and near-empty URLs make the real ones a larger share of what Google sees.
• Be patient after fixing. "Validate fix" typically takes up to about two weeks, can take much longer, and should not be clicked again until it succeeds or fails.

What does not work

• Pressing "Request indexing" repeatedly. Google says it won't speed things up and that there is a quota.
• Blocking or noindexing the list to make it shrink. That removes the pages you wanted indexed. If a URL is not wanted, noindex it first, as in the sequencing guide.
• Publishing more pages on the same pattern. If the stuck ones are thin, more of them adds to the problem.
• Treating it as a penalty. Google's description is of a crawl that has not happened yet, not of a manual action.

What this cannot tell you