What could cause Google to not honor canonical URLs?

[IrvCo_Interactive]

I have a strange situation on a website, when I do a Google query of site:example.com all the top indexed results appear to be queries that users can perform on the website. So any random term the user searches for on the website for some reason is causing the search result page to get indexed - like example.com/search/query/random-keywords

However, the search results page has a canonical tag on it that points to example.com/search, but that doesn't seem to be doing anything. Any thoughts or ideas why this could be happening?

[GastonRiera]

Hi there,

First of all, its a mistake to think that when searching with _site:  _operator, the first results are the most important nor the more relevant. Google has said a few times that we shouldn't rely that much on what that search in terms of what's being shown.

Blocking search results with robots.txt wont be of help, as it will not remove already indexed pages and cant prevent for new pages to be indexed (if there's an external link to a robots.txt blocked page, google can still index it) it'll only prevent Googlebot from discovering new ones FROM YOUR SITE.

Again, i'd try to dig deeper to understand where are the links to internal searches that google is finding. Googlebot will not do any search in your site.

The thing with GSC, might be related to quite a few reasons. I cant say much because I don't know any more specifics, but from what you are telling me it looks like you are getting impressions in searches that you don't relate to your site and that land on pages that google is noindexing. Yeah im repeating the obvious, hehe. 
In my experience, Google can have these strange behaviours. You know, there are cases when a page is canonicalized, but it can still be shown in SERPS. Dont ask me why, but it happens. It takes a little time to google fully replace it with the correct one.
I'd wait a little longer to see how Google is handling them.

I don't know if im helping you. 
it kinda took me a few minutes to understand/process what you wrote and come up with an answer.

Please, feel free ask again or comment on my reply if I misunderstood something.

Best luck,
Gaston

[IrvCo_Interactive]

Hi here's some more background info on this situation that makes it even stranger. I can perform some pretty specific searches on Google where these indexed search result pages show up. And I can look in Google Search Console under the performance section and see that those pages receive impressions and clicks. However, if I inspect the URL, Search Console says it is not included in Google's index, and the reason it gives under indexing is because it says it is honoring the canonical URL. So search console is saying it isn't indexed because of the canonical, but I can do searches and find that exact URL in the index. Any ideas what this could be from?

[IrvCo_Interactive]

Hi Gaston,

Thanks for the response. I can confirm that the example, /search and /search?q=foo are pretty much identical. However that may not always be the case, only when a user searches for something that would return no results. So, a website that sells widgets, /search and /search?q=widgets would not be identical, and in that case it would make sense that Google would not honor the canonical link. What's really strange is if I search google for the site: operator of the domain, the top pages are not user queries for things that make sense. The top indexed pages are random, non-relevant user searches.

I do not have a way with this system to control noindex tags on these search result pages. The only thing I could do is take the nuclear option and just block it all with robots.txt using wildcards. But that means no search result pages would get indexed, relevant or not.

[GastonRiera]

Hi there,

in my experience, when google doesn't honor Canonicals, is because pages arent similar.
In its definition, canonical are there for two or more pages that have the same content.

If you are finding it problematic, i'd suggest to use noindex tags for that search pages.
I'd investigate If there are links pointing to those internal search pages, as its not common for google to discover search pages.

Hope it helps,
Best luck.
Gaston