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
- What is OSINT? OSINT (open-source intelligence) is not "anything you can find online". It is information that is publicly or commercially available, collected and analysed to answer a specific question someone needs answered.
- The question comes first. Without it you are browsing, not researching, and you will collect personal data you have no reason to hold.
- Compliance, fraud, security, journalism and legal teams all use OSINT, and they all need the same basics: a stated purpose, recorded sources, graded findings and a lawful basis.
- Below is a first workflow you can run this afternoon with a browser and two terminal commands, plus the mistakes we see most often from new teams.
- UserSearch puts 100+ third-party data sources behind one account, with Cases that keep your search history as an audit trail, so the method is repeatable rather than improvised.
Why Teams Struggle to Define OSINT
What is OSINT? OSINT (open-source intelligence) is research on publicly and commercially available information, carried out to answer a stated question and recorded so another analyst could repeat it. Ask five people in the same company what OSINT is and you will get five answers. The security team says it means scanning the internet for the company's own infrastructure. Compliance says it means adverse media checks. The fraud team thinks of social media accounts. Legal thinks of company filings and old web pages. A manager who has just been handed an "OSINT budget" often assumes it means a tool you buy.
All of them are describing a slice of the same discipline. The trouble starts when the slices never meet. Each team runs its own searches, in its own browser, with its own habits, and nothing is written down in a way the next person can check. Then a regulator, an auditor or opposing counsel asks how a conclusion was reached, and nobody can say.
This guide gives professional teams a shared, practical definition. We cover the official wording, the intelligence cycle that turns searching into research, who uses OSINT and for what, where the line between "open" and "not open" sits, a first workflow with real commands, and the legal and ethical ground rules that apply in the UK and EU.
What Is OSINT? The Official Wording and What It Means in Practice
The clearest recent definition comes from the US Intelligence Community. Its IC OSINT Strategy 2024-2026 defines OSINT as "intelligence derived exclusively from publicly or commercially available information that addresses specific intelligence priorities, requirements, or gaps." The document now sits on the ODNI archive site, but the definition is still the one most practitioners quote.
Three parts of that sentence matter for commercial teams:
- "Derived". Raw material is not intelligence. A company register entry, a web page or a domain record is information. It becomes intelligence only after someone has checked it, compared it with other sources and drawn a conclusion.
- "Publicly or commercially available". This includes free public sources and data you pay for from lawful providers. The strategy itself talks about publicly available information (PAI) and commercially available information (CAI) side by side.
- "Addresses specific requirements". OSINT answers a question. "Is this supplier a genuine trading company?" is a requirement. "Find out everything about this supplier" is not.
We use a slightly plainer version with clients: OSINT is research on publicly and commercially available information, carried out for a stated purpose, recorded well enough that another analyst could repeat it. The last clause is ours, and it is the one most teams skip.
OSINT Is a Discipline, Not a Tool
Tools help. They do not define the work. A username search tool that checks 3,000 sites is a collection aid; it does not tell you whether two accounts with the same handle belong to the same organisation. That judgment is the analyst's job, and it is the part that gets challenged later.
How Does the Intelligence Cycle Apply to Everyday OSINT?
Government analysts organise their work around the intelligence cycle. UK defence doctrine, set out in Joint Doctrine Publication 2-00, describes the core functions of intelligence and the cycle that links them: direction, collection, processing and dissemination. Commercial teams do not need the military vocabulary, but the loop maps neatly onto a due diligence or fraud case.
| Stage | What it means for a commercial team | Typical output |
|---|---|---|
| Direction | Agree the question, the scope, the lawful basis and the deadline with whoever asked for the work. | A one-paragraph brief with in-scope and out-of-scope items |
| Collection | Search the sources that can answer the question, and record each search: source, input, date and time. | A source log, captures and exported results |
| Processing and analysis | De-duplicate, check that identifiers really match, grade each source and claim, and compare explanations. | Graded findings and key judgments with confidence levels |
| Dissemination | Write it up for the decision-maker, with the evidence attached for the reviewer. | A report, plus the Case or folder that backs it |
| Feedback | Ask whether the answer was useful and what the next question is. | A new, narrower brief, or case closed |
The loop matters because OSINT rarely finishes in one pass. A first round of collection usually changes the question. The supplier you were checking turns out to share a director with three dormant companies, and now the real question is about that director's other roles. Going back to direction, and writing the new question down, keeps the research proportionate instead of letting it sprawl.
Why a Clear Definition Matters: Lessons from a £21m Fine
In July 2025 the Financial Conduct Authority fined Monzo £21,091,300 for failings in its financial crime controls between 2018 and 2020. One detail in the FCA's statement should interest every OSINT team: the bank onboarded customers on "obviously implausible information", including people giving well-known London landmarks as their address.
That is not an advanced research failure. A single open-source check, comparing the declared address against a map or a public listing, would have flagged it. The lesson is that OSINT is often less about clever techniques and more about doing the obvious checks consistently, on every case, and recording that you did them.
The same pattern appears in supplier onboarding, insurance claims and grant applications. The research needed is usually basic. What fails is consistency and evidence. That is why we pitch OSINT to managers as a process, not a tool. If your team handles checks like these for clients, our page for research professionals shows how the platform supports repeatable, attributed work.
Who Uses OSINT, and for What?
The same discipline serves very different teams. What changes is the question and the sources that answer it.
Compliance and Anti-Money Laundering
Know-your-business checks, sanctions and adverse media screening, and source-of-wealth corroboration. The core sources are company registers, sanctions lists, official gazettes, news archives and domain records. The typical question: is this entity who it says it is, and is anything on the public record that changes our risk rating?
Fraud Prevention and Claims
Linking accounts, emails, phone numbers and payout details across cases to find coordinated activity, and checking claims against what is publicly visible. Fraud teams need speed and volume, and they need their links graded honestly so that a coincidence does not become an accusation.
Security and Threat Intelligence
Mapping an organisation's internet-facing infrastructure, finding lookalike domains that copy a brand, and attributing phishing infrastructure to known campaigns. Here OSINT overlaps heavily with DNS, certificate and hosting data.
Journalism and Research
Verifying images and videos, confirming who is behind a company or a campaign, and building a public-interest story from documents. Newsrooms and NGOs pioneered many of the verification methods commercial teams now use; the free Bellingcat Online Toolkit is still one of the best catalogues of them.
Legal Teams
Pre-action research, asset and corporate structure research, and preserving web evidence for disputes. Lawyers care most about provenance: when was this captured, by whom, and can we show it has not changed since?
What Counts as "Open" and What Does Not
New analysts often draw the line in the wrong place. "Publicly visible" and "fair to use" are not the same thing, and "behind a login" does not always mean "off limits". A working rule set:
- Clearly open: official registers (Companies House, charity registers, the UK Sanctions List), published tribunal and regulatory decisions, news, company websites and filings, public social media posts from organisations, domain registration and DNS data, certificate logs, web archives.
- Commercially available: data you licence from a lawful provider, such as business information databases or aggregated technical data. This is still OSINT under the ODNI wording, but you inherit the provider's terms and you should know where their data comes from.
- Grey areas that need a decision: content visible only after creating an account, closed groups where membership is approved, and data that is public but was never meant to be combined. Your organisation should decide these in advance, in writing, rather than leaving each analyst to improvise.
- Everything else: keep to the categories above, use only the access you are authorised to have, and deal with people openly. Material that did not reach the public through its owner does not become open because someone reposted it.
Personal data inside open sources is still personal data. Public availability affects how you assess fairness and proportionality under data protection law. It does not switch the law off.
Your First OSINT Workflow: One Supplier, One Afternoon
Here is a starter workflow we give new team members. The question is deliberately narrow: is Lathwell Freight Ltd (fictional) a genuine UK company, and is the domain on its invoices plausibly its own?
Step 1: Write the Brief Before You Search
Three lines in your case notes: the question, the lawful basis (for most B2B checks, legitimate interests with an assessment on file), and what is out of scope. For a supplier check, directors' private lives and personal social media are out of scope unless something specific brings them in.
Step 2: Query the Company Register
Start with the official register. You can search Companies House in a browser, but the Companies House API gives you structured, timestamped output you can save as evidence. Register on the developer hub to get a free API key. The key is sent as the username in HTTP basic authentication with the second field left blank, which is why the command below ends the key with a colon.
curl -s -u "$CH_API_KEY:" \
"https://api.company-information.service.gov.uk/search/companies?q=lathwell+freight&items_per_page=5" \
| tee s01_ch_search_$(date -u +%Y%m%dT%H%MZ).json-skeeps curl quiet so only the JSON is printed.-u "$CH_API_KEY:"sends your key, stored in an environment variable so it does not end up in your shell history or screenshots.q=is the search string;items_per_pagelimits the result count.teeprints the result and saves a copy with a UTC timestamp in the file name, which becomes source S01 in your log.
Friction to expect: the API allows 600 requests per five minutes per application, and you get HTTP 429 errors if you loop too fast. Search is fuzzy, so similarly named companies appear; always confirm the company number, not just the name. Then pull the officers for the right number with the /company/{number}/officers endpoint and compare them against the names on the onboarding form.
Step 3: Check the Invoice Domain
Next, look at the domain in the supplier's email signature. RDAP is the structured successor to WHOIS, and ICANN's registration data lookup gives you the same information in a browser. From a terminal:
# replace example.com with the registered domain from the invoice
curl -sL "https://rdap.org/domain/example.com" \
| python3 -m json.tool | grep -A3 '"eventAction"'rdap.orgis a bootstrap service that redirects you to the right registry's RDAP server;-Ltells curl to follow that redirect.python3 -m json.toolpretty-prints the JSON so it is readable.grep -A3 '"eventAction"'pulls out the events block: registration, last changed and expiry dates.
For most domains, personal registrant details are redacted, which is correct under GDPR. What you get reliably is the creation date and the registrar. A domain registered three weeks before the first invoice, for a company incorporated eight years ago, is worth a question.
Step 4: Look at the Past, Then Record Everything
Check a web archive for early copies of the supplier's website, then hash every file you saved so later changes can be detected:
sha256sum s01_ch_search_*.json captures/*.pdf >> source_log_hashes.txtsha256sum prints a fingerprint for each file; >> appends the lines to your hash log rather than overwriting it. That is the whole first workflow: a written question, an official source, a technical source, a historical source and an evidence trail. It is modest. It is also more than many organisations do today.
Mistakes New OSINT Teams Make
- Starting with tools instead of a question. The result is a pile of data and no answer. Worse, it is a pile of personal data you now have to justify holding.
- Matching on names alone. A name like "Lathwell Freight" can belong to more than one company, in the UK and elsewhere. Match on stable identifiers: company numbers, domains, registration dates, officer appointment dates.
- Treating absence as proof. "No results" means no results in the sources you searched, on the day you searched. Record it as a negative check with its coverage, not as a finding that something does not exist.
- Screenshots without dates. An undated screenshot is an anecdote. Capture with a timestamp, hash the file, and note the URL.
- Believing aggregator output. Third-party data is a lead to corroborate, not a fact. Coverage and freshness vary by provider, and some records are years old.
- Letting scope creep. The research drifts from the company to its directors, then to their private lives, with no one asking whether any of it is needed. Go back to the brief and write down any change of scope.
From Browser Tabs to a Repeatable Process in UserSearch
The first workflow above works. It also involves an API key, two terminal windows, a folder of JSON files, a hash log and a separate document for notes. Now multiply it by forty suppliers a month and three analysts. The weak point is not any single check. It is the joins between them.
We built UserSearch 2.0 to remove those joins. One account gives you 18 Search types, including Corporate Intelligence, Domain Intelligence (ownership, history and favicon), Email Intelligence, Phone Intelligence, IP Intelligence and Username Intelligence, and access to 100+ third-party data sources without holding a separate account with each provider.
- OneScan runs one input across several data sources you select and merges the results with source attribution, so you always know which provider said what. The Credit cost is shown before you run it.
- Cases in Forensic Mode store your search history and bookmarks, which gives you the source log and negative checks without typing them up. Private mode is there for exploratory work you choose not to store.
- Teams share Cases and a common Credit pool controlled by the team leader, so a reviewer can open the same history the analyst used.
- SargeBot, our AI research assistant, helps you set a lawful objective, build entity searches and draft a PDF report. You choose the model (Claude, GPT or Grok), and you verify every line before it goes anywhere.
If you are brand new to the platform, our beginner's guide to UserSearch OSINT walks through your first Case, and the OSINT fundamentals knowledge base covers the terms used in this article.
Worked Scenario: A Compliance Team's First Structured Check
Context. Harbour Lane Consulting (fictional) onboards about thirty new business clients a month. Its checks so far have been one analyst "having a look on Google". The head of compliance wants a defensible minimum standard after a client turned out to be a shell company.
Actions. The team writes a one-page procedure based on the cycle above. Direction: every check starts with the same question, "is this a genuine trading entity controlled by the people named?", and a lawful basis note. Collection: Corporate Intelligence on the company number, Domain Intelligence on the website and email domain, and a OneScan on the contact email address, all inside a Case in Forensic Mode named with the client reference. Processing: each finding is graded and anything that conflicts with the onboarding form is flagged for a second analyst. Dissemination: a half-page summary goes to the relationship manager; the Case stays with the compliance file.
Outcome. In the first month, 27 of 31 checks pass without comment. Three are escalated because the website domain is only weeks old while the company claims years of trading; two of those turn out to be genuine rebrands with a clear paper trail, and one is declined. The fourth escalation is a registered office that matches a well-known virtual office address used by hundreds of companies, which is common and lawful but prompts a request for a trading address. When the firm's own auditor reviews the process, every decision has a Case ID, a dated search history and a named reviewer. That is the difference between "we looked" and "we can show what we looked at".
Worked Scenario: Verifying a Statement Before a Story Runs
Context. A business newsroom receives a PDF press statement, apparently from a listed retailer, announcing the closure of a distribution centre. It arrives from [email protected]. The editor wants to know within the hour whether it is genuine.
Actions. The reporter writes the question in the Case: "Did the retailer issue this statement?" Domain Intelligence shows retailer-news.example.org was registered two days earlier, with no history, while the retailer's real press office uses its main domain. The PDF's document properties show it was produced with a consumer word processor under a user name that does not match the retailer's usual press release template. A search of the retailer's own newsroom page and the regulatory announcement feed finds nothing. The reporter calls the press office using the number published on the retailer's own website, not the one in the PDF.
Outcome. The press office confirms the statement is not theirs. The newsroom does not run the story; it runs a short piece on the false statement instead, citing the domain registration date and the retailer's denial. The Case holds the captures and search history in case the retailer's lawyers or a regulator later ask how the newsroom verified the material.
Advanced Patterns Once the Basics Stick
Pivot on Stable Identifiers, Not Names
Once a company number, domain or email address is confirmed, use it as the anchor for the next searches. Identifiers drift less than names, and they cut false positives dramatically. Our guide to domain OSINT and infrastructure shows how far one confirmed domain can take you.
Grade Sources and Claims Separately
An official register is a reliable source, but what a director declares on it is still self-declared. Keep a source grade and an information grade for each finding so readers can see both.
Write Down the Alternative Explanation
For each conclusion, record at least one other explanation and why you rejected it. A new domain might mean an invented company, or a legitimate rebrand. Showing you considered both is what makes the conclusion defensible.
Batch the Routine, Review the Exceptions
Use bulk search to run up to five inputs through one Module at a time for the routine checks, then spend analyst time on the handful of cases that conflict. That is where judgment adds value.
The Legal and Ethical Ground Rules
In the UK and EU, the main framework for OSINT on individuals is data protection law. Most commercial research relies on legitimate interests, which means passing a three-part test: a legitimate purpose, research that is necessary for that purpose, and a balance that does not override the individual's rights. The ICO's legitimate interests guidance was updated in March 2026 to reflect the Data (Use and Access) Act, and it is worth rereading if your assessment templates predate that change. This is general information, not legal advice.
In practice, that means:
- Record your purpose and lawful basis before you start, and keep a legitimate interests assessment for recurring research types.
- Collect only what answers the question, and set a retention date for what you keep.
- Stay on the right side of the Computer Misuse Act 1990: use only access you are authorised to have, and respect platform terms of service.
- Journalists should work within their own codes, such as the Editors' Code of Practice, which sets public-interest tests for intruding on privacy.
- For work that may end up before a tribunal or regulator, follow a documented method; the Berkeley Protocol on Digital Open Source Investigations is the standard many legal and human rights teams use.
Treat every result as a lead to corroborate. Describe people factually and neutrally. And when a finding could affect a decision about an individual, make the uncertainty visible to whoever makes that decision.
Frequently Asked Questions About OSINT
What is OSINT?
OSINT, or open-source intelligence, is information that is publicly or commercially available, collected and analysed to answer a specific question. The US Intelligence Community's OSINT Strategy 2024-2026 defines it as intelligence derived exclusively from publicly or commercially available information that addresses specific intelligence priorities, requirements or gaps. Raw material becomes intelligence only after checking and analysis.
What is the difference between OSINT and searching online?
Searching online collects information, while OSINT turns that information into a checked answer to a stated question. OSINT starts with a written question and lawful basis, records each source with the date and time, grades findings, and compares explanations. Without a question, research becomes browsing and gathers personal data an organisation has no reason to hold.
Is OSINT legal in the UK?
OSINT is lawful in the UK when it follows data protection law and uses only authorised access. Personal data inside open sources is still personal data, and most commercial research relies on legitimate interests, which needs a legitimate purpose, necessity and a fair balance. Teams should also respect platform terms and the Computer Misuse Act 1990.
Who uses OSINT?
OSINT is used by compliance and anti-money laundering teams for business checks and sanctions screening, fraud prevention and claims teams for linking accounts and details across cases, security teams for infrastructure and lookalike domains, journalists for verifying images and statements, and legal teams for corporate research and preserving web evidence for disputes.
What are the stages of the intelligence cycle in OSINT?
The intelligence cycle in OSINT runs through direction, collection, processing and analysis, dissemination and feedback. Direction agrees the question, scope and lawful basis. Collection searches and logs sources. Processing checks and grades findings. Dissemination reports to the decision-maker with evidence attached, and feedback sets the next, narrower question or closes the case.
Does commercially available data count as OSINT?
Commercially available data from a lawful provider counts as OSINT under the US Intelligence Community definition, which covers publicly available information and commercially available information side by side. Teams using licensed business databases or aggregated technical data inherit the provider's terms, so they should know where the provider's data comes from and treat each result as a lead to corroborate.
OSINT Your Team Can Explain and Repeat
OSINT is research on publicly and commercially available information, done for a stated purpose and recorded so it can be checked. The definition is short. The discipline is in the details: a written question, the intelligence cycle, stable identifiers, graded findings, honest negatives and a lawful basis on file.
Stop guessing. Start researching with UserSearch. One account gives your team 18 Search types and 100+ third-party data sources, OneScan to run one input across the sources you choose with attribution on every result, Cases that keep your search history as a ready-made audit trail, shared Cases for peer review, and SargeBot to draft the PDF report you then verify and sign.