I Noindexed a Third of My Own Pages

01Asking about every page

A small tool site I run has struggled with Google since the June spam update. On 29 October I stopped guessing and asked Search Console about every address in its sitemap, one by one, through the inspection API. Fifty six pages were indexed. Forty two had been crawled and not indexed. Twenty were unknown to Google, and eleven had been discovered but never crawled. Of the 60 templated tool pages, 34 were refused. So were 26 of its 37 blog posts. In fifteen days Googlebot had fetched 13 pages. A page Google refuses to index is not neutral. It is a vote about the site, cast every time it is crawled.

How it got there

Back in May the site published 61 templated pages in a single day, each a variant of the main tool with a different preset, and its sitemap went from 42 addresses to 107 overnight. A month later the spam update arrived and the site's Google traffic collapsed. Some of those template pages were genuinely useful. Most were near copies of each other, and a search engine looking at a young site that doubled its size in a day with near copies has every reason to be cautious about the rest.

02Take the refused pages out of the queue

Noindex, keep the links

The same day, the 34 template pages Google had refused got a noindex, follow tag, and the sitemap dropped from 135 addresses to 102. The pages still exist and still work for anyone who lands on them; their links still pass through. They simply stop asking Google to reconsider them. A sitemap is a list of pages you are asking to be judged on, so do not put pages there you would not defend. With 13 Googlebot fetches in two weeks, every fetch spent on a page Google had already refused was a fetch not spent on one it might accept.

Why not delete them

Deleting would have removed pages that a few people use and that assistants occasionally send visitors to. The site gets more human visits from ChatGPT tagged links than from Google right now, 21 to the homepage alone in fifteen days. Assistants do not care whether Google indexed a page; they care whether it answers. Noindex separates the two audiences: the page stays useful to people and to anything that links to it, and stops counting against the site in Google's eyes. It is the opposite conclusion from the thin pages that earn, where a templated set earned real clicks and stayed. The difference is the data. Templates are not the problem; templates nobody searches for, published all at once, are.

03A bet with a date on it

What would count as working

I do not know yet whether this helps, and I am not going to pretend otherwise. What would count as working: more of the remaining 102 pages moving to indexed over the next six to eight weeks, Googlebot fetching more pages per day, and the core pages climbing out of the positions in the sixties. What would count as failing: nothing moving at all, which would point at the site as a whole rather than the templates. A cleanup without a stated test is a ritual, so write down what success looks like before you start. The earlier recovery story in I got deindexed overnight took a support ticket and two days on Bing; Google does not offer that door.

Run the same audit on your site

The inspection API answers one address at a time with a daily quota, so it suits sites with hundreds of pages, not millions. Ask it about every sitemap address, group the answers by page type, and look for a type where most pages are refused. That type is either thin, duplicated or unwanted. The coverage idea from seventy five impressions applies here too: the share of your pages Google is willing to show is a number worth watching weekly, and it is invisible until you ask for it page by page.

Tomorrow: forty one pages that existed at the edge and nowhere else.