Disclaimer: This article is for education and for lawful, authorised professional research. Use these methods only where you have a legitimate purpose and a lawful basis, and follow the laws and platform terms that apply to you, including data protection law such as the UK GDPR and EU GDPR. See our Terms of Service.
TL;DR
- Google dorks (search operators) are still one of the fastest OSINT routes to company documents, filings, press releases, public PDFs and job adverts, but several operators that older guides rely on no longer work.
- Google documents only a handful today: quotes, minus, site:, filetype:, before: and after:. Others such as intitle: and inurl: work without official documentation, and cache: was retired in 2024.
- Below is a quick-reference table with 2026 status, the Bing, DuckDuckGo and Brave equivalents, and worked recipes for due diligence, brand misuse and verification.
- Google's Custom Search JSON API is closed to new customers and ends for existing ones on 1 January 2027, so scripted dorking needs a rethink.
- UserSearch covers many of the same questions through structured Search types, with a search history you can hand to a reviewer.
Why Search Operator Guides Go Stale
Google dorks for OSINT are search queries built from operators such as site:, filetype: and exact-match quotes, used to find documents organisations deliberately publish: filings, reports, press releases, public PDFs and job adverts. Search a phrase like "Google dorks for OSINT" and most results are lists copied from lists written years ago. They still tell you to use cache: to see a deleted page, related: to find similar companies and + to force a word. None of those is a method you can rely on now. Worse, many of those lists stray well beyond the documents organisations choose to publish, which is where professional research belongs.
This quick reference does two things differently. It records the 2026 status of each operator, based on what Google and the other engines document today. And every recipe is framed around documents organisations deliberately publish: annual reports, filings, press releases, procurement notices, public PDFs and job adverts. That is where most due diligence, brand and verification answers come from anyway.
What Does a Search Operator Actually Do?
A search operator is a word or symbol that narrows what a search engine returns: a specific site, a file type, a date range, or a phrase in the page title. "Google dork" is simply the community's name for a query built from operators. Google's own help page on how to refine web searches lists the operators it currently documents for ordinary users: exact-match quotes, site:, the minus sign, before:, after: and filetype:.
Google's developer documentation for site owners adds a few more, such as src: and imagesize: for image searches, and notes two limits worth remembering: a site: query does not necessarily return every indexed URL, and site:https://www.example.com does not return the same results as site:https://example.com/. Operators narrow the index. They never show you everything.
If you prefer forms to syntax, Google's Advanced Search page builds many of the same queries with fields for exact phrases, sites, file types and dates.
Why Do Operator Changes Matter to Research Teams?
Two changes should shape how your team uses search engines in 2026.
First, the cache is gone. In February 2024 Google confirmed that its cache links were being retired, and The Verge reported that the cache: operator would go too, with Google's search liaison saying it was no longer needed now that pages load reliably. For researchers, the cached copy was often the quickest proof of what a page said before it was edited. That job now belongs to web archives and to your own captures.
Second, automation has narrowed. Google's documentation states that the Custom Search JSON API is closed to new customers, and that existing customers have until 1 January 2027 to move to an alternative. Teams that ran saved dork lists through that API for scheduled brand reviews or adverse media sweeps need a new plan before then, and automating queries against the normal results page runs into Google's terms and its bot defences quickly.
Newsrooms feel both changes directly: verification depends on proving what was published and when. Our journalism page covers how structured, attributed searches support that work.
Which Google Dorks Still Work for OSINT? The 2026 Quick Reference
Status key: Documented means listed in Google's current help or developer pages. Works, undocumented means widely used and still honoured in practice, but not promised by Google, so results can be partial. Retired means do not build a method on it.
| Operator | What it does | 2026 status | Example |
|---|---|---|---|
| "exact phrase" | Matches the words in that order | Documented | "Harrowdene Freight Ltd" |
| - (minus) | Excludes a word or site | Documented | -site:example.com |
| site: | Limits results to a domain, subdomain or URL prefix | Documented | site:gov.uk |
| filetype: | Limits results to a file type (pdf, docx, xlsx, pptx, csv) | Documented | filetype:pdf |
| before: / after: | Limits results by date (year or YYYY-MM-DD) | Documented | after:2025-01-01 |
| OR | Either term; must be upper case | Works, undocumented | "annual report" OR "strategic report" |
| intitle: / allintitle: | Word(s) in the page title | Works, undocumented | intitle:"modern slavery statement" |
| inurl: / allinurl: | Word(s) in the URL | Works, undocumented | inurl:investors |
| intext: | Word in the body text | Works, undocumented | intext:"company number" |
| * (inside quotes) | Wildcard for a missing word | Works, undocumented | "appointed * as chief financial officer" |
| AROUND(n) | Two terms within n words of each other | Unofficial, inconsistent | "Harrowdene" AROUND(5) "acquisition" |
| cache: | Showed Google's stored copy of a page | Retired (2024) | Use a web archive instead |
| + (plus) | Forced a word to appear | Retired | Use quotes instead |
| related:, info:, link: | Similar sites, page info, backlinks | Not in current documentation | Do not rely on them |
Two habits make these operators more dependable. Put no space after the colon (site:gov.uk, not site: gov.uk). And never treat Google's "about N results" figure as a count; it is an estimate that changes between pages of the same query.
Equivalents on Other Engines
Different engines index different pages, so a second engine is the cheapest way to widen coverage. Bing documents its advanced search keywords, including intitle:, inbody:, filetype:, contains: and ip:, which returns sites hosted on a given IP address. DuckDuckGo's search syntax page supports site:, filetype:, intitle: and inurl:. Brave Search publishes its search operators, including inpage: and lang:.
| Task | Bing | DuckDuckGo | Brave | |
|---|---|---|---|---|
| Word in title | intitle: | intitle: | intitle: | intitle: |
| Word in body | intext: | inbody: | not documented | inbody: |
| File type | filetype: | filetype: or ext: | filetype: | filetype: or ext: |
| Date range | before: / after: | date filter in the interface | date filter in the interface | date filter in the interface |
| Sites on one IP | none | ip: | none | none |
Worked Query Recipes for Professional Research
Every recipe below uses fictional names and example.com placeholders. Replace them with the organisation in your brief. Each is meant to be run by hand, read, and captured if it answers the question.
Due Diligence: Documents the Company Published
site:harrowdene-freight.example.com filetype:pdf after:2025-01-01Finds PDFs the company itself has published since the start of 2025: brochures, policies, certificates, reports. site: restricts to the company's domain, filetype:pdf to PDFs and after: to recent material. Compare what the documents claim with what the company told you.
"Harrowdene Freight" (intitle:"annual report" OR intitle:"modern slavery statement") filetype:pdfPulls annual reports and modern slavery statements that mention the company, whether published by it or by customers listing it as a supplier. UK organisations above the turnover threshold also file statements on the modern slavery statement registry, which is worth checking directly.
Due Diligence: Filings and Public Notices
site:thegazette.co.uk "Harrowdene Freight"
site:find-and-update.company-information.service.gov.uk "Harrowdene Freight"The first finds notices in The Gazette, the UK's official public record, including insolvency and strike-off notices. The second surfaces indexed Companies House pages, which occasionally catches former names that the register's own search ranks lower. Always confirm on the register itself: Companies House search is the source, Google is only the finder.
site:gov.uk "Harrowdene Freight" filetype:pdf
"Harrowdene Freight" (site:contractsfinder.service.gov.uk OR site:find-tender.service.gov.uk)Public sector documents and procurement notices. Since 24 February 2025, notices under the Procurement Act 2023 are published on Find a Tender, so older contract awards and newer ones may sit on different services. Search both, then use each service's own search, which is more complete than Google's view of it.
Brand Misuse: Copycat Pages and Adverts
intitle:"Marlowby Home Goods" -site:marlowbyhomegoods.example.com -site:linkedin.comPages that put the brand name in their title but are not on the official site or LinkedIn. Expect resellers, review sites and news; you are looking for pages presenting themselves as the brand.
inurl:marlowby-homegoods -site:marlowbyhomegoods.example.comHyphenated brand names in URLs of other domains, which catches many copycat shops and support pages. Pair this with lookalike domain generation and registration checks, covered in our guide to domain OSINT and WHOIS infrastructure, for domains Google has not indexed yet.
"Marlowby Home Goods" ("work from home" OR "remote role" OR "no experience needed") -site:marlowbyhomegoods.example.comJob adverts using the brand name away from the official careers site, a common sign of recruitment scams that borrow a real employer's reputation.
"Marlowby Home Goods" (site:boards.greenhouse.io OR site:jobs.lever.co OR site:apply.workable.com)The company's genuine postings on the applicant platforms it actually uses. You need this baseline to tell a real advert from a copy, and genuine adverts are also useful in due diligence: they reveal office cities, team growth and the software a company runs.
Verification: Where and When Did This Text First Appear?
"we have taken the difficult decision to close our Leeds distribution centre" before:2026-09-01Searches for an exact sentence from a circulating statement, restricted to pages dated before the statement supposedly appeared. A hit from last year suggests recycled text. No hit is weak evidence; Google's dates come from the page and are sometimes wrong.
"Marlowby Home Goods" "distribution centre" (site:marlowbyhomegoods.example.com OR site:londonstockexchange.com OR site:prnewswire.com) after:2026-09-01Checks whether the same announcement appears on the company's own site, a regulatory news feed or a wire service. Genuine corporate announcements usually appear in at least one of those places.
When a page has been edited or removed, move from search to archives. For UK public bodies, the UK Government Web Archive holds captured copies of government websites; for everything else, use the Internet Archive and your own captures.
A Small Script for Logging Manual Queries
Because automated querying is off the table for most teams, we keep manual work disciplined instead. This script turns a recipe file into clickable search URLs and writes a timestamped log, so your method section records exactly which queries you ran. It needs only Python 3, no packages and no API key.
python3 - <<'EOF'
import urllib.parse, datetime
engines = {"google": "https://www.google.com/search?q=",
"bing": "https://www.bing.com/search?q="}
now = datetime.datetime.utcnow().strftime("%Y-%m-%dT%H:%MZ")
with open("recipes.txt") as f, open("query_log.csv", "a") as log:
for q in (line.strip() for line in f if line.strip()):
for name, base in engines.items():
url = base + urllib.parse.quote(q)
print(url)
log.write(f'{now},{name},"{q}",{url}\n')
EOFurllib.parse.quote encodes the operators and quotes safely for a URL; the log line records the UTC time, engine, raw query and URL. Open each link by hand, read the results, and capture what matters.
Where Structured Search Types Beat Dorking
Operators are excellent for documents. They are weaker for questions that depend on structured data: who owns a domain and when it changed hands, which companies share officers, whether an email address appears across services, or where else an image has been published. Those answers exist in registers and specialist data sources that Google indexes poorly or not at all.
UserSearch puts those sources behind one account. Corporate Intelligence answers company questions from registry-grade data. Domain Intelligence covers ownership, history and favicons. Product Intelligence covers trends, patents, scholarly material and app listings, which is often what a brand team is really searching for with intitle:. Picture runs reverse image search, geolocation and synthetic-image checks. OneScan runs one input across several sources at once, with attribution on every result and the Credit cost shown first.
Every search in a Forensic Mode Case is stored with its bookmarks, so the query log the script above builds by hand already exists. When a search engine result is the evidence, the free Forensic Capture extension captures the page with SHA-256 fingerprints and two independent timestamps. Our searching guide explains how the Search types fit together.
Worked Scenario: A Supplier That Looked Established
Context. A fictional housing association is about to award a maintenance contract to Pellingford Facilities Ltd, which claims ten years of public sector work. The procurement lead asks for evidence of that history before signing.
Actions. The analyst runs the procurement recipe and finds no contract notices naming the company on either Contracts Finder or Find a Tender. site:gov.uk "Pellingford Facilities" filetype:pdf returns two council committee minutes from 2023 listing it as a subcontractor, which partly supports the claim. site:pellingford-facilities.example.com filetype:pdf finds an accreditation certificate; searching the certificate number in quotes returns the certifying body's public register, where the certificate shows as expired. The Gazette search returns a proposal-to-strike-off notice from 2024 that was later withdrawn.
Outcome. The report says the company has a genuine but modest public sector history as a subcontractor, not ten years as a prime contractor, and that its accreditation has lapsed. The association asks for a current certificate and two references before award. Every query is in the analyst's log with a timestamp, and each result that mattered is captured.
Worked Scenario: Job Adverts Borrowing a Brand
Context. A fictional retailer, Marlowby Home Goods, hears from three people who were asked to pay for "training materials" after applying for a remote job in its name. The brand team wants to know how widespread the adverts are.
Actions. The analyst first runs the genuine-postings recipe to confirm the retailer only advertises through its careers site and one applicant platform. The brand-misuse job recipe then returns 14 adverts on free job boards and classified sites, all using the same wording and a Gmail contact address. inurl:marlowby-homegoods finds a copycat careers site on marlowby-homegoods-careers.example.net. The same search on Bing adds four adverts Google had not indexed.
Outcome. The brand team reports the adverts to each job board with captures, asks the registrar to act on the copycat careers domain, and publishes a short notice on its real careers page listing where genuine vacancies appear. The recipes are saved in the team's recipe file and rerun at each quarterly brand review.
Advanced Patterns for Operator-Based Research
Pair Date Operators with Archives
before: and after: rely on the date Google assigns to a page, which can be the publication date, the last edit or a guess. Use them to find candidates, then confirm dates with the web archive or the document's own metadata.
Build Recipe Files per Case Type
Keep a text file of tested recipes for each recurring job: supplier checks, brand reviews, statement verification. New analysts start from the file, and your method stays consistent. For handle searches, our guide to mapping accounts from one username covers operator-based methods.
Run Every Important Query on Two Engines
Google and Bing index different corners of the web. For anything that will go in a report, run it on both and note which engine produced each result.
Search the Source, Not Just the Index
Operators tell you a document exists. The register, the procurement service or the company's own newsroom is where you confirm it. Treat Google as a finder, never as the source you cite.
Staying Proportionate with Search Operators
Search operators find what organisations chose to publish, but results often contain personal data too: names in minutes, signatures on certificates, contact details in adverts. Collect only what answers your question, and do not let a broad query become a reason to keep everything it returned. The ICO's guidance on purpose limitation is a good test: you should be able to say why you are collecting each item and how that fits the purpose you recorded at the start. This is general information, not legal advice.
- Keep recipes focused on company documents, filings, press releases, public PDFs and job adverts.
- If a result turns up material that was plainly not meant to be public, do not open, download or use it; note that it exists if relevant and tell the owner through a proper channel.
- Respect search engines' terms: run queries by hand or through an official API, at a sensible pace.
- Treat every result as a lead to corroborate at the original source.
Frequently Asked Questions About Google Dorks for OSINT
What are Google dorks in OSINT?
Google dorks are search queries built from operators, words or symbols that narrow what Google returns to a specific site, file type, date range or phrase in a page title. In OSINT, Google dorks are one of the fastest ways to find company documents, filings, press releases, public PDFs and job adverts that organisations have deliberately published.
Which Google search operators still work in 2026?
Google currently documents exact-match quotes, the minus sign, site:, filetype:, before: and after:. Operators such as OR, intitle:, inurl:, intext: and the wildcard inside quotes still work in practice but are not promised by Google, so results can be partial. The cache: and plus operators are retired, and related:, info: and link: are no longer documented.
Does the Google cache operator still work?
The Google cache: operator no longer works. In February 2024 Google confirmed that its cache links were being retired, and the cache: operator went with them. For OSINT researchers, the cached copy was often the quickest proof of what a page said before an edit, so that job now belongs to web archives such as the Internet Archive and to your own captures.
Can you automate Google dorks for OSINT?
Automating Google dorks has become much harder. Google's Custom Search JSON API is closed to new customers, and existing customers have until 1 January 2027 to move to an alternative. Automating queries against the normal results page runs into Google's terms and bot defences, so many teams run recipes by hand and keep a timestamped query log instead.
What is the difference between Google and Bing search operators?
Google and Bing share operators such as site:, intitle: and filetype:, but differ elsewhere. Bing uses inbody: for body text, adds contains:, and offers ip: to return sites hosted on a given IP address, which Google has no equivalent for. DuckDuckGo supports site:, filetype:, intitle: and inurl:, and Brave adds inpage: and lang:.
Is it legal to use Google dorks for research in the UK?
Google dorks are standard search features, so the legal questions in the UK come from what a researcher collects and why. Results often contain personal data, such as names in minutes or contact details in adverts. The ICO's purpose limitation principle means collecting only what answers the recorded question, and material plainly not meant to be public should be left alone.
From Saved Queries to Checkable Research
Search operators remain one of the most useful skills in OSINT, as long as you know which ones still work, which are retired, and which only find documents organisations deliberately publish. Keep a recipe file, run important queries on two engines, confirm at the source and log what you ran.
Stop guessing. Start researching with UserSearch. One account gives your team 18 Search types and 100+ third-party data sources for the questions dorks cannot answer, OneScan with attribution on every result, Cases that record every search you run, Forensic Capture for hashed and timestamped page evidence, and SargeBot to help build entity searches and draft a report you verify. Legal teams can see how this fits dispute and pre-action work on our law professionals page.