What is OSINT? A practical guide to open-source intelligence

Open-source intelligence (OSINT) is intelligence produced from information that anyone can lawfully obtain: websites, social media, news, public records, company filings, patents, academic papers, satellite imagery and data sets. The word that matters is intelligence, not open source. A pile of screenshots is not OSINT. OSINT starts when someone collects public information to answer a specific question, checks how reliable it is, and analyses it into a judgement that a decision-maker can use.

OSINT used to be treated as the junior partner of secret collection. That has changed. In 2024 the US intelligence community published its first community-wide OSINT strategy and called open source “INT of first resort”. Human-rights investigators now use the Berkeley Protocol, published by the UN Office of the High Commissioner for Human Rights, to make open-source evidence admissible in court. Journalists, compliance teams, academic researchers and market analysts use the same methods.

The intelligence cycle, applied to open sources

Most intelligence work, open or secret, follows some version of the intelligence cycle:

1. Direction. Define the question and who needs the answer. “Who supplies components to company X?” is workable. “Find everything about company X” is not.
2. Collection. Gather information from sources likely to answer the question. Keep a log of what you searched, where and when.
3. Processing. Translate, de-duplicate, geolocate, extract metadata and archive. Raw material becomes usable evidence.
4. Analysis. Weigh the evidence, consider alternative explanations and reach a judgement with an explicit level of confidence.
5. Dissemination. Deliver the answer in a form the decision-maker can use: a short brief, not a data dump.
6. Feedback. Find out whether the product helped, and refine the next question.

The cycle is a simplification. Real investigations loop back constantly: analysis exposes a gap, which sends you back to collection. Even so, the cycle is a useful discipline because it forces you to start from a question rather than from a tool.

What counts as an open source

Source type	Examples	Typical use
Web and social media	Posts, profiles, forums, archived pages	Events, networks, sentiment, claims to verify
Public records	Company registries, court records, procurement notices, land registries	Ownership, contracts, legal exposure
Scholarly and technical	Papers, patents, standards, conference talks	Capabilities, research directions, who works with whom
Imagery and geodata	Commercial satellite imagery, street-level imagery, maps	Verifying locations, monitoring change over time
Trade and transport	Ship and aircraft tracking, customs data	Supply chains, sanctions evasion
Grey literature	Think-tank reports, government white papers, theses	Context and expert interpretation

“Open” does not mean “free” or “harmless”. Commercial databases are open sources if anyone can buy access. Personal data that is public may still be protected by data-protection law, and collecting it can raise ethical issues. Responsible OSINT practice sets limits before collection starts. It decides what is necessary for the question, how personal data will be minimised and stored, and when to stop.

Analysis: where OSINT succeeds or fails

Collecting public information is easier than ever. Analysing it well is not. The classic reference is Richards Heuer’s Psychology of Intelligence Analysis. Heuer argued that analysts’ biggest problems are cognitive, not informational. People see what they expect to see, anchor on their first hypothesis and under-weight evidence that does not fit.

Structured analytic techniques (SATs) are designed to counter these habits:

• Key assumptions check. List the assumptions your conclusion depends on and ask what happens if each is wrong.
• Analysis of Competing Hypotheses (ACH). Write down every plausible explanation and score each piece of evidence against all of them. The best-supported hypothesis is the one with the least evidence against it, not the one with the most evidence for it.
• Source evaluation. Rate each source for reliability and each item for credibility, separately. A reliable outlet can carry a false claim, and an unknown account can post genuine footage.
• Red teaming and devil’s advocacy. Ask a colleague to argue the opposite case before you publish.
• Indicators. Decide in advance what you would expect to see if a hypothesis were true, then look for it.

US analytic standards (Intelligence Community Directive 203) add a reporting discipline that works for any OSINT product. Describe the quality of your sources. Express uncertainty with consistent terms such as “likely” or “almost certainly”. Separate what the evidence shows from your own judgement. Say what alternative explanations you considered.

Verification: the core OSINT skill

For many OSINT tasks, the hardest question is “is this real, and is it what it claims to be?” A standard verification workflow checks:

1. Provenance. Is this the original? Use reverse image and video search to find earlier copies.
2. Source. Who posted it first? What is their history, location and track record?
3. Date. When was it created? Metadata is often stripped, so compare shadows, weather records and other events in the frame.
4. Location. Where was it taken? Match landmarks, signs, terrain and road layouts against maps and satellite imagery. This is geolocation.
5. Motivation. Why was it shared, and who benefits if it spreads?

Open communities such as Bellingcat publish step-by-step guides to these methods. The Berkeley Protocol sets out how to document each step so that someone else can reproduce your findings and evidence holds up later. That means recording URLs, timestamps and hashes, and archiving pages before they disappear.

Tools come second

New OSINT tools appear every month, and lists of them go out of date quickly. Tools fall into a few stable categories: search and archiving, social media monitoring, image and video verification, geolocation, company and records lookup, network analysis and data cleaning. More and more of them now include AI features, such as automatic translation, transcription, entity extraction and summarisation.

AI speeds up processing, but it adds a new verification burden. Language models can invent sources, mistranslate names and present guesses as facts. Treat AI output as a lead to check, never as evidence.

Ethics and law

Good OSINT practice is defined as much by what you do not do:

• Do not access accounts, systems or data without authorisation, even if the password is “public”.
• Do not deceive people to obtain information unless your organisation has a lawful basis and an ethical review for doing so.
• Minimise personal data, protect it and delete it when the purpose is served.
• Think about harm before publishing. Naming the wrong person in a viral investigation can ruin a life.
• Respect the law where you are, where your subjects are and where your data is stored.

Where OSINT is heading

Three trends are reshaping the field. First, the volume of open data, especially commercial satellite imagery and machine-readable public records, keeps growing, so prioritisation matters more than access. Second, generative AI makes fake images, documents and personas cheaper to produce, so verification becomes more central, not less. Third, the line between government, commercial and civil-society OSINT is blurring. The same techniques serve sanctions enforcement, investigative journalism, due diligence and academic research.

The fundamentals have not changed. Start from a question, document your collection, evaluate every source, test competing explanations and say how confident you are. That discipline turns public information into intelligence.