Google search operators that still work in 2026, tested

You've typed site: into Google before, and it did what you expected. Then you tried related: or cache: from an old cheat sheet, got a page of results, and couldn't tell whether the operator did anything.

Most operator lists don't say. So I sent 32 of them through Google on October 2, 2026, three searches each, and compared every search with the same words minus the operator. Each verdict comes from the results that came back, checked against what the operator promises.

At a glance

What a Google search operator is, and the six Google documents

A search operator, sometimes called a Google search command, is a word or symbol that tells Google how to read the rest of your query. Type jaguar speed and Google decides what you meant. Type jaguar speed -car and you've told it: no cars.

Google's Refine Google searches help page lists six for web search: quotes for an exact match, site:, the minus sign, before:, after:, and filetype:. It also warns that site:nytimes.com works and site: nytimes.com, with a space, doesn't.

Google's Search Central overview for site owners adds two more, imagesize: and src:, which only work on Google Images. Everything else on this page, from intitle: to AROUND(n), is an operator people pass around that Google doesn't document today.

If you've heard the term "Google dorking", it's this same set of operators, used to find pages a site didn't mean to make easy to find. The results below apply to it too.

All 32 operators, and what each one did

Each operator ran on three searches. An operator "works" when, on all three, most results showed what it promises and more of them did than on the same search without it. It works "sometimes" when that held on one or two of three, and has "no effect" when it held on none. "Can't tell" means what it promises isn't something a list of results can show.

The documented ones work, and so does intitle:

Five of Google's six documented operators did their job on every search. site: returned 30 of 30 results from the site named. The minus sign took a word out of every result, and quotes turned 1 phrase match in 26 results into 20 in 23.

before: and after: are the date filter to use. Every dated result fell on the right side of the date, on all six searches. Without the operator, most results showed no date at all.

intitle: isn't in Google's docs, and it held up anyway: 22 of 24 titles had the word. So did the * wildcard inside quotes and define:.

filetype: and the URL operators work some of the time

filetype:pdf worked, and the other file types barely did. Six of nine results were PDFs. filetype:docx returned a single result, and filetype:xlsx returned none for "monthly budget template". ext:, which most lists treat as the same operator, returned 3 PDFs in 26 results.

inurl: and allinurl: worked on two of three searches. On the third, the share of URLs with the word was the same with and without the operator. allintitle: worked once. OR and the vertical bar passed the check only on the kayak search, where most results mentioned just one of the two boats.

-site: left out nasa.gov and reddit.com completely. On photosynthesis -site:wikipedia.org, all 10 results were from Wikipedia, on both runs. If you rely on -site:, look at what comes back.

related:, link:, info:, and cache: are gone, even when they return results

Google has retired five of the operators in this test, and said so.

Most of the retired ones don't fail visibly. related:nytimes.com returned nine results, all on New York Times sites, where the operator used to list similar sites. link:github.com returned eight GitHub pages out of nine, where it used to list pages that link to GitHub. info: isn't in Google's docs, and 2 of its 25 results were the homepage it's meant to summarize. Only cache: fails outright, with no results at all.

daterange: never had Google's documentation, and on all three searches no dated result fell inside its range. before: and after: do the same job.

Some operators can't be judged from a results page

Eight operators promise something a list of titles, links, and snippets can't show. intext: and inanchor: are about a page's body text and the links pointing at it. AND is already Google's default. For $ and AROUND(n), the plain search already did what the operator would add: every result for "camera 400" showed "$400" without the dollar sign in the query.

~ and + are retired, so there's no reason to use them. For the other six, the results moved when you added them, but results also move when you run the same search twice, so that alone doesn't show the operator did anything.

How this compares to what is already published

• Oxylabs (August 2026) explains the common operator types and says search engines "periodically deprecate or modify legacy search features". It describes related:, which Google removed from its docs in 2023.
• Apify (October 2025) lists the Google dorking operators it considers supported, partly from its own tests, and notes that "some operators that worked in the past are now deprecated or return inconsistent results", without naming which.
• SerpApi covers Google's URL parameters, such as hl, gl, and tbs, rather than the operators you type in the query.

All three explain what operators are meant to do. None of them puts each operator next to the same search without it, which is what shows that related: and link: now return a full page of results without doing their job.

How I measured it

1. I picked 32 operators from Google's help page and the operator lists that rank for this search, and wrote three searches for each.
2. I sent each operator search within about a minute of its plain twin, the same words without the operator.
3. I checked every result against what the operator promises: the domain for site:, the URL ending for filetype:, the title for intitle:, the date at the start of the snippet for before: and after:, and so on.
4. I replaced two searches whose plain version already passed the check on every result, such as "history coffee", where every title already said "history". A search like that can't show an operator doing anything.

The 209 calls, including repeat runs of some plain searches, ran on October 2, 2026, between 22:17 and 22:52 UTC, and cost $0.11 in total. They all went through google.search, which returns Google's organic results as JSON.

The limits of this data

• Three searches per operator is enough to see whether an operator does anything, and not enough to say how often it fails. One odd result moves a verdict.
• Results came from Google in English, for the US, on one day. A different country, language, or week could move the "sometimes" verdicts.
• The checks read titles, links, snippets, and the date at the start of a snippet. A page that matches in its body text but not in its snippet counts as a miss.
• Google chooses the title it shows for each result from several sources, so a page can match intitle: without the word in the title you see. That can make intitle: and allintitle: look weaker than they are.

The big results hold either way: site:, quotes, minus, and the date operators changed every search, and related:, link:, and cache: didn't do their job once.

What to do with it

1. Use site:, quotes, the minus sign, before:, and after: freely. They worked on every search.
2. Use filetype:pdf for documents, and expect little from other file types or from ext:.
3. Treat inurl:, allinurl:, allintitle:, and OR as hints. Check the results.
4. Drop related:, link:, info:, cache:, daterange:, ~, and + from your notes.
5. For backlinks, use a backlink tool instead of link:.

If you run searches from code, the same operators go in the query string. Here's one call:

curl -sS -X POST 'https://api.getanyapi.com/v1/run/google.search' \
  -H "Authorization: Bearer $ANYAPI_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"query":"site:nasa.gov mars rover"}'

And the first two of the 10 results it returned on October 2, 2026:

{
  "output": {
    "found": true,
    "data": {
      "query": "site:nasa.gov mars rover",
      "results": [
        {
          "position": 1,
          "title": "Mars Exploration Rovers: Spirit and Opportunity",
          "link": "https://science.nasa.gov/mission/mars-exploration-rovers-spirit-and-opportunity",
          "snippet": "NASA's Spirit and Opportunity rovers were identical twin robots who helped rewrite our understanding of the early history of Mars."
        },
        {
          "position": 2,
          "title": "Mars Rover Opportunity - Mars Missions",
          "link": "https://www.jpl.nasa.gov/missions/mars-exploration-rover-opportunity-mer",
          "snippet": "Opportunity was the second of the two rovers launched in 2003 to land on Mars and begin traversing the Red Planet in search of signs of ancient water."
        }
      ]
    }
  },
  "costUsd": 0.0005,
  "items": 10
}

google.search costs $0.40 per 1,000 searches, billed per call from a prepaid wallet, so rerunning this whole test would cost about a dime.

To compare Google search APIs on price, see the cheapest SERP API comparison.

Use data responsibly and follow AnyAPI's Acceptable Use Policy.