The usual answer is that a noindexed page still gets crawled, so the links on it still count, at least for a while. It’s repeated constantly and it’s genuinely hard to check. This test sets out how I’m checking it on my own sites, what I expect to happen, and how you could run the same thing.
The question
Does a page carrying a noindex directive still pass value through the links on it? Specifically, can a brand new destination page, whose only inbound link comes from a noindexed source page, get discovered, indexed and ranked for a term nothing else on the web uses.
Why this matters
If noindexed pages pass nothing, then large noindexed sections are dead weight in your internal linking, and any page relying on them for discovery is effectively orphaned. If they do pass value, a noindex is a display decision rather than a structural one. That changes how you handle filtered pages, thin archives and staging sections.
The method
Two pages on separate sites I control. The source page carries a noindex directive and one outbound link. The destination page is brand new, linked from nowhere else, and contains a single invented term that returns no results in Google today.
Then I wait, checking three things weekly in Search Console and by site search: whether the destination is discovered, whether it is indexed, and whether it ranks for the invented term.
| Element | Setting |
|---|---|
| Source page | Live, crawlable, noindex via meta robots, one outbound link |
| Destination page | New URL, no other inbound links, no sitemap entry, not submitted |
| Signal term | An invented string with zero results at the start |
| Checks | Weekly, Search Console URL inspection plus a site search |
| Duration | 12 weeks, or until the destination is indexed |
| Control | A second destination page with no inbound links at all |
The control matters more than it sounds. Without it, a destination that gets discovered proves nothing, because Google finds URLs through routes nobody planned.
What I expect to happen
My prediction, written before the result: the destination page gets discovered within three weeks and indexed within six, and the control page does not. I expect the link to work for discovery, which is the part most people actually care about.
I am less confident about ranking. A single link from a noindexed page is a very thin signal, so the invented term may return the page without that telling us much about value passing.
Results
The result is pending, and this section stays empty until the twelve weeks are up. When it fills, it will carry the discovery and indexing dates for both the destination and the control, whether the invented term returns the page, and the raw Search Console screenshots behind each of those claims.
[PAUL TO ADD: the dates and the outcome, once the test has actually run.]
I would rather publish a page that says pending than one that says what I assumed.
What I concluded
The conclusion is pending too, and it gets written from the data rather than from the prediction above. If the destination never gets discovered, that is the finding and it goes up unchanged. A test that contradicts what I expected is more useful published than quietly abandoned, so that is what happens here.
How to repeat this
Running this yourself needs two sites you control, about ten minutes of setup and the patience to leave it alone for three months. The whole method is five steps, and the only part people reliably get wrong is the robots.txt handling in step one, which quietly invalidates everything that follows.
- Publish a page on site A with a
noindexmeta robots tag and confirm it is crawlable, not blocked in robots.txt. Blocking it means Google never sees the link at all, which tests something different. - Publish a page on site B, linked only from that page. Keep it out of your sitemap and do not submit it.
- Put an invented, unique string in the destination page’s body copy. Check it returns zero results first.
- Publish a second page on site B with no inbound links anywhere. That is your control.
- Check weekly with URL inspection and a search for the invented string.
The most common way this goes wrong is blocking the source page in robots.txt as well as noindexing it, which makes the whole test meaningless.
Limitations
Two sites is a sample of two, so whatever this produces is a signal rather than proof. It shows what happened for these specific sites at this specific time. Google’s handling of long-term noindexed pages has changed before and can change again, which makes any result here a snapshot rather than a rule.
FAQ
Does noindex stop Google crawling a page?
No. Noindex tells Google not to include the page in its index, but the page still has to be crawled for that directive to be read. Blocking the page in robots.txt is what stops crawling, and the two are frequently confused.
Is noindex, follow still a thing?
Google has said it treats long-term noindexed pages as though their links are nofollowed, because it eventually stops crawling them. That makes the interesting question one about timing rather than a simple yes or no, which is part of what this test is checking.
Should I noindex thin pages?
Sometimes, though merging or improving them is usually better. Noindex is a display decision. If a page is genuinely useful to visitors but not worth indexing, noindex fits. If nobody benefits from it, removing it is cleaner than hiding it.
Can I run this on a client site?
I would not, and I do not. The test involves publishing pages that exist only for the experiment and leaving them untouched for months, which is not something to do on a site somebody pays you to improve. Run it on a property you own instead.
Sources
- Google Search Central, robots meta tag documentation, on
noindexand the requirement that the page be crawlable for the directive to be seen. Accessed 16 September 2026. - Google Search Central, robots.txt documentation, on the difference between blocking crawling and preventing indexing. Accessed 16 September 2026.
Got a site with a large noindexed section?
Large noindexed sections are common on ecommerce sites with filters and on older sites with thin archives, and whether they are costing you anything depends on what is linked from inside them. That is a specific thing to check rather than a general worry, and it takes a look at your internal linking to answer properly.
Get in touch, or start with a free assessment. More tests are on the tests hub.