Blog
Google search operators
Which operators still function, which have been retired, why site: counts are not an index count, and the specific diagnostic jobs operators do well.
Short answer
Which Google search operators still work?
The reliable ones are site:, intitle:, allintitle:, inurl:, allinurl:, filetype:, exact-phrase quotes, the minus sign for exclusion, OR, and wildcards. The cache: operator was retired by Google in 2024. The link: operator has not returned usable results for many years, and info: and related: have been unreliable or removed. Operators are diagnostic tools for specific jobs, not measurement instruments: a site: result count in particular is an estimate and should never be reported as an index count.
Updated
The operators worth knowing
A small set does almost all the useful work. Everything else is either a variant of these or no longer functions.
| Operator | Example | What it is good for |
|---|---|---|
| site: | site:example.com/blog/ | Seeing roughly what Google holds for a domain or a section |
| intitle: / allintitle: | allintitle: seo agency dubai | Gauging how many competitors target a phrase in the title |
| inurl: / allinurl: | site:example.com inurl:?s= | Finding parameter, search and pagination URLs that should not be indexed |
| filetype: | site:example.com filetype:pdf | Finding indexed PDFs nobody remembers publishing |
| "exact phrase" | "request a free audit" | Finding duplicated boilerplate across a site or scraped copies elsewhere |
| - (minus) | site:example.com -inurl:https | Excluding a pattern; useful for spotting leftover HTTP URLs |
| OR | site:example.com (dubai OR abu dhabi) | Widening a check across variants in one query |
| * (wildcard) | "best * agency in dubai" | Finding phrasing variants used in titles and headings |
What no longer works
The cache: operator was retired by Google in 2024, and the cached link was removed from search results around the same time. For checking what a crawler actually sees, the replacement is the URL Inspection tool in Search Console for your own site, and fetching the raw HTML for any site.
The link: operator was deprecated long ago and has not returned a usable backlink set for many years. Anyone offering a backlink audit built on link: is not doing a backlink audit. The info: and related: operators have been unreliable or removed; where related: returns anything, it is not a dependable competitor set and should not be used as one.
Why the site: count is not an index count
This is the most consequential misunderstanding in practical SEO. The number Google shows above a site: query is an estimate produced for a results page, not a count of indexed URLs. It fluctuates between runs, between data centres and between the first and last page of results, and Google has said for years that it should not be relied on.
The authoritative figure for your own site is the Pages report in Search Console, which separates indexed from not-indexed and gives a reason for each exclusion. Reporting a site: count to a client as an indexation metric is a habit worth breaking: it is wrong, and when it moves for no reason it produces alarm or false comfort.
What the site: query is good for is the shape of the index rather than its size — whether a section exists at all, whether staging or parameter URLs have leaked in, whether titles are duplicated across a template, whether an old subdomain is still showing. Those are visible at a glance and are often the first sign of a real problem.
The diagnostic jobs operators do well
Finding what should not be indexed. A site: query narrowed with inurl: against parameter patterns, search URLs, tag archives or paginated series surfaces the low-value pages consuming crawl budget — usually faster than a crawl, and on live Google data rather than your own.
Finding leaked environments. A site: query against a staging subdomain, or an exact-phrase search for a distinctive string from a development build, regularly turns up environments nobody meant to publish. This is worth doing before a migration and again after.
Finding duplicated boilerplate. An exact-phrase search for a sentence from a template, restricted to the site, shows how many pages repeat it — which is how doorway-page patterns and thin city-swap pages reveal themselves. The same phrase searched without the site restriction shows who has copied the content.
Gauging competition for a phrase. allintitle: against a commercial term gives a rough sense of how many pages are explicitly targeting it in their title. It is crude, it is not a difficulty score, and it is directionally useful in a way that a vendor's difficulty number sometimes is not.
Finding link and citation opportunities. Exact-phrase searches for the patterns that directories, roundups and resource pages use, combined with a location term, produce a list of pages that already link out in your category. For the UAE this is more productive than most outreach lists, because the pool of genuinely local sources is small and finite. That work sits alongside the method on SEO across the UAE.
What operators cannot tell you
They cannot tell you your rankings. A query run from your own browser is personalised and localised, and the position you see is not the position your market sees. Rank tracking needs a tool that controls location and personalisation, or Search Console's average position, which is measured from real impressions.
They cannot tell you your backlinks, your traffic or your competitors' traffic. They cannot confirm a penalty — a page absent from a site: query may be unindexed for any number of mundane reasons, and reading a manual action into it is a leap. And they are increasingly unreliable at scale: heavy automated operator queries trigger blocking, which is a sign the method is being used as a measurement tool rather than a diagnostic one.
What is confirmed and what is observed
Confirmed: Google retired the cache: operator and removed cached links from results in 2024, and Google has repeatedly stated that site: result counts are estimates. Search Console's Pages report is the documented source for indexation status of a verified property.
Observed rather than documented: the current reliability of link:, info: and related:, which degraded gradually rather than through announcements, and the practical thresholds at which automated operator queries get blocked. Operator behaviour changes without notice, so test rather than assume.