Skip to main content

OSINT Techniques in 2026: 15 Methods Research Teams Use

Fifteen OSINT techniques research teams use in 2026, grouped by starting point: accounts, organisations, infrastructure and content. Each has a working command, what the flags do, API key needs and the pitfall that most often leads to a wrong answer.

· By UserSearch Team · 15 min read

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

  • Fifteen OSINT techniques research teams actually use in 2026, grouped by what you start with: an account or contact detail, an organisation, a piece of infrastructure, or a piece of content.
  • Each technique comes with one working command or query, what the flags do, whether you need an API key, and the pitfall that most often produces a wrong answer.
  • The techniques are strongest in chains: an email domain leads to a certificate, the certificate to sibling domains, the domains to a shared page resource.
  • Two worked scenarios show chains in practice: a security team mapping lookalike infrastructure and an AML analyst checking a crypto counterparty.
  • UserSearch runs most of these checks from one account across 100+ data sources, with source attribution and a stored search history.

Why Most Technique Lists Fail in Real Casework

OSINT techniques are repeatable methods for turning one known identifier, such as a handle, company number, domain or photo, into verified information or new identifiers using publicly or commercially available sources. Most "OSINT techniques" articles are tool lists in disguise. They name a website, paste a screenshot and move on. Then an analyst tries the method on a live case and hits the things the list never mentioned: the API that now needs a key, the service that returns a 502 error half the time, the result that looks like a match and is not.

This guide is built the other way round. For each technique we give you the method first, then a command you can paste, then the pitfall. We tested the commands against public example resources, and we tell you where the friction is. Some of it we hit while writing: the Internet Archive was showing a "Temporarily Offline" page on the day we checked the CDX query, and crt.sh returned a 502 before it answered. That is normal. Plan for it.

If you are new to the field, start with our OSINT fundamentals knowledge base for the vocabulary, then come back here.

What Counts as an OSINT Technique?

A technique is a repeatable way of turning one known identifier into verified information or new identifiers, using publicly or commercially available sources. It is not a tool. Maigret is a tool; checking whether a handle exists across platforms and then confirming which accounts are really linked is the technique. Tools change every year. Techniques last longer, which is why we organise this guide around them.

For a broader catalogue of methods, grouped by task, the Bellingcat Online Toolkit is the best free reference we know. It is maintained by practitioners and flags when tools stop working.

Why Does Technique Discipline Matter in 2026?

Sources move under your feet. On 28 January 2026 the OFSI Consolidated List closed, and the UK Sanctions List became the only source for UK sanctions designations. Any team whose screening procedure still pointed at the old list was, from that day, checking a frozen snapshot. Nothing broke loudly. The results just quietly aged.

That is the real risk in OSINT technique: not that a method is hard, but that it silently degrades. An API starts requiring a key. A registry changes its data model. A platform removes a search feature. Teams that write down which source and which version they used catch these changes. Teams that rely on habit do not. Security teams feel this most, because infrastructure data changes daily; our page for cyber security teams covers how attributed, dated results help there.

How Are the 15 OSINT Techniques Organised?

We group the techniques by starting point, because that is how casework arrives: someone hands you a handle, a company name, a domain or a photo. Each technique lists what it is, a command or query, and its main pitfall. Replace placeholders such as example.com and example_handle with your own inputs, within your agreed scope.

Techniques 1 to 3: Accounts and Contact Details

1. Handle Checks Across Platforms

What it is. Checking whether a handle exists on hundreds of sites at once, to find accounts that may belong to the same organisation or operator. Maigret is the most capable free option; no API key is needed.

pip3 install maigret
maigret example_handle --html
maigret example_handle -a --html

By default Maigret checks the 500 highest-traffic sites from its list; -a checks all of them (several thousand), and --html writes a report you can attach to a case.

Pitfall. Existence is not ownership. Short or common handles return dozens of unrelated accounts, and site checks break when platforms change their pages, producing false positives. Confirm each hit with a second signal, such as a matching link back to the organisation's website.

2. Email Address and Mail Domain Checks

What it is. Establishing whether an email address and its domain are plausible for the organisation that uses them. Start with DNS, which needs no tools beyond dig:

dig +short MX example.com
dig +short TXT example.com | grep -i spf

MX returns the mail servers; the TXT query with the grep filter shows the SPF record, which lists who may send mail for the domain. A company that claims fifty staff but whose domain has no MX record at all deserves a question.

Pitfall. Catch-all mail domains accept every address, so "the domain accepts mail" says nothing about a specific mailbox.

3. Phone Number Parsing and Range Checks

What it is. Checking a number's format, type (mobile, landline, VoIP) and original range holder before you search it anywhere else. Google's libphonenumber is available in Python as phonenumbers; no key needed.

pip install phonenumbers
python3 -c "
import phonenumbers as p
from phonenumbers import carrier, number_type
n = p.parse('+44 7700 900123', None)
print(p.is_valid_number(n), number_type(n), carrier.name_for_number(n, 'en'))"

parse reads the number (the None region works because it starts with +44), is_valid_number checks it against the numbering plan, number_type returns a type code, and name_for_number gives the operator the range was originally assigned to.

Pitfall. Number portability. The operator returned is the original range holder, not necessarily today's provider. Never state "this is a Vodafone number" from range data alone.

Techniques 4 to 6: Organisations and Official Records

4. Company Register Officer and Filing Checks

What it is. Pulling officers, appointment dates and filing history from an official register, then comparing them with what the company told you. For UK companies, the Companies House API needs a free key; it allows 600 requests per five minutes, as its rate limiting guide explains.

curl -s -u "$CH_API_KEY:" \
  "https://api.company-information.service.gov.uk/company/00000000/officers" \
  | jq '.items[] | {name, officer_role, appointed_on, resigned_on}'

-u "$CH_API_KEY:" sends your key as the basic-auth username; jq (install with your package manager) trims each officer record to the four fields you need. Swap /officers for /filing-history to see when accounts were filed and whether they were dormant.

Pitfall. Register data is largely self-declared. An officer appointed yesterday, a registered office shared by hundreds of companies, or dormant accounts from a firm claiming years of trading are all findings, but none proves anything alone.

5. Sanctions List Screening

What it is. Checking organisations and their controllers against official sanctions lists. In the UK that means the UK Sanctions List search service; for US-designated parties, OFAC's Sanctions List Search. For bulk or offline checks, download the list and search it locally:

curl -sL -o sdn.csv \
  "https://sanctionslistservice.ofac.treas.gov/api/PublicationPreview/exports/SDN.CSV"
grep -i "dunmarsh" sdn.csv

-o saves the file; -L follows redirects. grep -i is a crude first pass only.

Pitfall. Transliteration and name order. Exact matching misses variants; fuzzy matching floods you with near-misses. Use the official search tools' fuzzy scoring, record the threshold you used, and check registration numbers, addresses or other listed identifiers before calling anything a match.

6. News and Adverse Media Sweeps

What it is. Finding press coverage of an organisation across languages. The GDELT DOC 2.0 API, described in its launch post, searches a rolling window of recent worldwide coverage without a key.

curl -s "https://api.gdeltproject.org/api/v2/doc/doc?query=%22dunmarsh%20logistics%22&mode=artlist&format=json&maxrecords=50&timespan=3m"

The quoted, URL-encoded phrase keeps the name together; mode=artlist returns a list of articles; timespan= sets the window within GDELT's rolling coverage period.

Pitfall. GDELT asks for no more than one request every five seconds and will tell you so if you loop too fast. Coverage is recent only, so older history needs news archives. And name collisions are the norm: always tie an article to a company number, domain or named director before it goes in a report.

Techniques 7 to 12: Infrastructure and Transactions

7. Domain Registration Data and DNS

What it is. Establishing when a domain was created, through which registrar, and where it currently points. RDAP is the structured replacement for WHOIS; about.rdap.org explains the protocol and the bootstrap service.

curl -sL "https://rdap.org/domain/example.com" | jq '.events'
dig +short NS example.com
dig +short A example.com

.events holds registration, last-changed and expiry dates. The NS and A queries show name servers and the current IP address.

Pitfall. DNS is a snapshot of today. Where a domain pointed last year needs passive DNS data from a commercial provider, and registrant details are usually redacted under GDPR. Our guide to domain OSINT and WHOIS infrastructure goes deeper.

8. Certificate Transparency Searches

What it is. Every publicly trusted TLS certificate is written to public logs, as explained at certificate.transparency.dev. Searching them reveals subdomains and sibling domains issued on the same certificate.

curl -s "https://crt.sh/?q=%25.example.com&output=json" \
  | jq -r '.[].name_value' | sort -u

%25 is an encoded % wildcard, so the query matches every subdomain; output=json returns machine-readable rows; sort -u removes duplicates.

Pitfall. crt.sh is a free community service and often times out or returns 502 on large domains; retry later or narrow the query. Wildcard certificates leave individual hostnames unnamed, and a certificate proves a domain was configured, not that a site is live.

9. Lookalike Domain Generation

What it is. Generating typo, homoglyph and TLD variants of a brand domain and checking which are registered. dnstwist is the standard tool; no key needed.

pip install dnstwist
dnstwist --registered --format csv example.com > lookalikes.csv

--registered shows only variants that resolve; --format csv gives you a file to sort and share.

Pitfall. Registered is not abusive. Many lookalikes are parked, held defensively by the brand itself, or belong to an unrelated business with a similar name. Check content and registration dates before escalating.

10. Internet-Facing Host Data

What it is. Reviewing what services a server offers to the internet, as recorded by a search engine for devices. The Shodan CLI needs an API key; install steps are in the Shodan CLI guide.

pip install -U --user shodan
shodan init YOUR_API_KEY
shodan host 203.0.113.10

init stores your key; host returns open ports, service banners, hostnames and the organisation the IP range is registered to.

Pitfall. Banners can be months old, and shared hosting means one IP serves thousands of unrelated sites. Use the data you are given; do not connect to or probe systems you have no authorisation to test.

11. Shared Page Resource Pivots

What it is. Finding websites that load the same file, such as a logo, script or stylesheet, or share an identical page title. urlscan.io indexes scans and lets you search by the SHA-256 hash of any response, as its search documentation shows. An API key is needed for scripted searches.

curl -s -H "API-Key: $URLSCAN_KEY" \
  "https://urlscan.io/api/v1/search/?q=hash:<sha256-of-logo-file>" \
  | jq -r '.results[].page.domain' | sort -u

hash: matches any HTTP response with that fingerprint; page.domain lists the sites where it appeared. A page.title: query works the same way for cloned page titles.

Pitfall. Visibility. Scans you submit with public visibility are published for other users to see. Submit sensitive material as private or unlisted, and search before you scan. Also, popular frameworks share files across millions of sites; pivot on rare resources only.

12. Blockchain Address Review

What it is. Reading the public transaction history of a cryptocurrency address to establish activity dates, volumes and counterparties. Blockstream's Esplora API serves Bitcoin data without a key.

curl -s "https://blockstream.info/api/address/1A1zP1eP5QGefi2DMPTfTL5SLmv7DivfNa" \
  | jq '.chain_stats'

This example uses the well-known genesis block address. chain_stats returns transaction count and total funded and spent amounts.

Pitfall. Addresses are not owners. Exchanges pool funds, and clustering heuristics that group addresses are estimates. Treat any wallet-linking result as a lead to corroborate, not an attribution.

Techniques 13 to 15: Content, Media and Time

13. Web Archive History

What it is. Listing every archived copy of a site to see what it said and when. The Wayback Machine's CDX server, documented in the wayback-cdx-server README, answers without a key.

curl -s "https://web.archive.org/cdx/search/cdx?url=example.com/*&output=json&fl=timestamp,original,statuscode&collapse=digest&from=2020"

url=example.com/* covers every page on the domain; fl= limits the fields; collapse=digest drops captures whose content did not change; from=2020 sets the start year.

Pitfall. The archive is not complete and not always up. Sites can ask to be excluded, and outages happen. Record "no capture found" as a negative with the date you checked.

14. Image Provenance and Metadata

What it is. Finding earlier copies of an image with reverse image search (Google Lens, Bing Visual Search, TinEye, Yandex) and reading any embedded metadata. ExifTool reads both images and documents.

exiftool -a -G1 -s photo.jpg
exiftool -Author -Creator -Producer -CreateDate statement.pdf

-a shows duplicate tags, -G1 shows which metadata group each tag came from, and -s prints short tag names. The second line pulls authorship and creation fields from a PDF, which often reveals the software and account that produced a document.

Pitfall. Most social platforms strip metadata on upload, so absence means nothing. Metadata can also be edited. The earliest copy found by reverse search is only the earliest one indexed.

15. Geolocation and Chronolocation

What it is. Verifying where and when a photo or video was taken by matching visible features to map data and shadows to sun position. The Overpass API queries OpenStreetMap for combinations of features; SunCalc then checks shadow angles against a date and time.

[out:json][timeout:60];
area["name"="Cardiff"]["boundary"="administrative"]->.a;
node["amenity"="fuel"](area.a);
out center;

Paste this into Overpass Turbo. It defines the city boundary as area .a, then returns every fuel station inside it; add more tags to narrow to the feature combination in your image.

Pitfall. Map data is incomplete, and once you have a candidate spot it is easy to see only the features that fit. Write down the features that do not match as well as those that do.

Running the Same Techniques from One Account

Look back at the list. Fifteen techniques, at least five API keys, several rate limits, a dozen output formats and no shared record of what was searched. The techniques are sound. The overhead is what slows teams down and causes mistakes.

UserSearch brings most of these checks into one platform. Username Intelligence, Email Intelligence, Phone Intelligence, Corporate Intelligence, Domain Intelligence (ownership, history and favicon), IP Intelligence, Cyber Intelligence through Shodan Modules, Cryptocurrency, Picture (reverse image search, geolocation and synthetic-image checks), Threat Intelligence and more sit behind one account with access to 100+ third-party data sources, so you do not hold separate accounts with each provider.

  • OneScan runs one input across the sources you select and merges results with source attribution; the Credit cost is shown before you run.
  • Bulk search runs up to five inputs through one Module in a batch.
  • Cases in Forensic Mode store your search history and bookmarks, so the chain of techniques you used is recorded as you go.

For teams that want to call these sources from their own systems, API access through OneQuery currently has a waiting list.

Chaining Techniques: Two Worked Scenarios

Scenario: Mapping Lookalike Infrastructure for a Retailer

Context. The security team at a fictional retailer, Copperlea Homewares, receives customer reports of a phishing page on copperlea-returns.example.net. The question: is this a one-off page, or part of a wider set of infrastructure copying the brand?

Actions. dnstwist on the brand domain (technique 9) returns 41 registered variants. RDAP (7) narrows these to nine registered in the last 60 days, six through the same registrar. crt.sh (8) shows four of the six sharing a single multi-domain certificate. The analyst hashes the logo file on the reported page and searches urlscan (11), privately, which returns two further domains that do not look like the brand name at all but load the same logo. The CDX query (13) shows none of the domains had content before last month.

Outcome. The team reports eight domains to registrars and hosting providers with evidence of shared certificates and resources, rather than one. The two domains that did not resemble the brand would never have come out of lookalike generation alone; they came from the chain. The analyst records the three variants that turned out to be the retailer's own defensive registrations, so nobody reports them next time.

Scenario: Checking a Crypto Counterparty for an AML Team

Context. A fictional payments firm's AML analyst reviews a business customer, Brightwater Trading Ltd, which asks to settle invoices through a Bitcoin address. The question: is the address consistent with the business's stated activity, and does anything link it to sanctioned parties?

Actions. The Esplora API (12) shows the address first received funds eleven days before the customer applied, with a handful of large incoming transfers. The Companies House API (4) shows Brightwater incorporated fourteen months ago with one director appointed last month. Screening (5) of the company and director names against the UK Sanctions List returns no match at the recorded threshold. GDELT (6) finds no coverage. The website's archive history (13) begins three weeks ago.

Outcome. No sanctions link, but a pattern of newness across every source: new director, new site, new address. The analyst escalates for enhanced due diligence, asking for proof of trading and the source of the incoming funds, and records each negative check with its date. The finding is "consistent with a recently repurposed company", not an accusation, and the report says so.

Advanced Habits That Make Techniques Reliable

Anchor Every Chain on a Confirmed Identifier

A chain is only as strong as its first link. Confirm the starting domain, company number or handle before pivoting, or every later result inherits the error.

Log Negative Results with Coverage

"No sanctions match" should read "no match on the UK Sanctions List search service at fuzzy threshold X, searched 2026-09-24". Without the coverage, the negative is worthless.

Version Your Methods

When a source changes, as the UK sanctions list did in January 2026, update your procedure and note the date. A method document with a change log is one of the cheapest controls a team can adopt.

Prefer Rare Pivots

Shared hosting IPs, popular JavaScript libraries and common handles connect everything to everything. Pivot on rare values: a unique logo hash, a custom analytics ID, a distinctive page title.

Every technique here touches data that is publicly or commercially available, but several can reach personal data quickly: usernames, email addresses and phone numbers belong to people. Under the UK GDPR, as amended by the Data (Use and Access) Act 2025, you still need a lawful basis, a defined purpose and proportionate collection. Write the purpose down before you run the first command. This is general information, not legal advice.

  • Use only the access you are authorised to have. Reading published data is research; connecting to systems to test them without permission is not.
  • Respect platform terms and rate limits. Throttling is the service telling you how it wants to be used.
  • Keep research on organisations and their public activity unless your purpose clearly requires an individual, and record why when it does.
  • Treat every output as a lead to corroborate. Coverage and freshness depend on the source.
  • Set retention periods for exported results and delete what you no longer need.

Frequently Asked Questions About OSINT Techniques

What are OSINT techniques?

OSINT techniques are repeatable ways of turning one known identifier into verified information or new identifiers using publicly or commercially available sources. Checking whether a handle exists across platforms and then confirming which accounts are really linked is a technique. Techniques outlast the tools that support them, which change every year, so teams should document methods rather than habits.

What are the most common OSINT techniques in 2026?

Common OSINT techniques in 2026 include handle checks across platforms, email and mail domain checks, phone number parsing, company register and sanctions list checks, news sweeps, RDAP and DNS lookups, certificate transparency searches, lookalike domain generation, host data review, shared page resource pivots, blockchain address review, web archive history, image metadata checks, and geolocation.

What is the difference between an OSINT technique and an OSINT tool?

An OSINT tool is software such as Maigret, dnstwist or ExifTool. An OSINT technique is the method the tool supports, including how results are confirmed. Maigret is a tool; checking a handle across platforms and confirming which accounts are really linked with a second signal is the technique. Tools change often, while techniques last much longer.

Do OSINT techniques need API keys?

Many OSINT techniques need no API key: Maigret, dig, the phonenumbers library, RDAP, crt.sh, dnstwist, GDELT, Blockstream's Esplora API and the Wayback Machine CDX server all answer without one. The Companies House API needs a free key, the Shodan CLI needs a key, and urlscan.io needs a key for scripted searches. Most free services also enforce rate limits.

Why do OSINT techniques produce wrong answers?

OSINT techniques produce wrong answers when analysts overread the data. A handle that exists is not proof of ownership, a phone range holder may not be today's provider after porting, news results often involve name collisions, host banners can be months old, and a blockchain address is not an owner. Confirming each result with a second signal prevents most errors.

OSINT techniques are lawful in the UK when used with a lawful basis, a defined purpose and proportionate collection under the UK GDPR, as amended by the Data (Use and Access) Act 2025. Teams should use only authorised access, respect platform terms and rate limits, keep research on organisations unless the purpose requires an individual, and set retention periods.

Fifteen Methods, One Audit Trail

The techniques that matter in 2026 are not exotic. They are handle checks across platforms, contact detail checks, register and sanctions lookups, domain and certificate research, resource pivots, archive history and careful image work, chained together and written down. The difference between a good team and an average one is discipline: confirmed anchors, rare pivots, logged negatives and methods that are updated when sources change.

Stop guessing. Start researching with UserSearch. One account gives your team 18 Search types across 100+ third-party data sources, OneScan to run one input across the sources you choose with attribution on every result, bulk search for batches, Cases that record every step of the chain, and SargeBot to help build entity searches and draft a report you then verify.

About the author

UserSearch Team
Updated on Nov 13, 2026