The Core Paradigm: Information vs. Intelligence

Information becomes useful intelligence when it is evaluated, interpreted in context, and connected to a decision. A collection of links, screenshots, or company records can be valuable information. It becomes an intelligence assessment when the reader can understand what those sources support, what remains uncertain, why it matters, and what should happen next.

This distinction is central to open-source intelligence, or OSINT. It also helps anyone preparing a research brief, checking a supplier, evaluating a public claim, or monitoring changes that affect an organization. The goal is decision-useful reasoning with a traceable evidence base.

Expanded November 5, 2026 with AI assistance and linked primary-source review. The earlier article disclosed Gemini assistance. Examples in this guide are hypothetical, and the templates are editorial teaching aids.

Data, information, and intelligence: a working distinction

In a practical research workflow, data are recorded elements such as dates, names, measurements, and document identifiers. Information organizes those elements into a meaningful statement. Intelligence adds evaluation and interpretation for a particular user’s question or decision.

For example, a registry number is a data element. A dated registry record connecting that number to a legal entity is information. An assessment explaining whether the entity matches a proposed contracting party—and what remains unresolved—can support a purchasing decision.

Terminology varies across professions. NIST’s intelligence glossary includes definitions involving collection, processing, integration, analysis, evaluation, and interpretation. The important distinction for this guide is the work done between retrieving material and giving a reader a defensible answer.

Do not assume that every useful source must be transformed into an elaborate report. A reader may only need a specific document. The ODNI guide for first responders describes an intelligence cycle while recognizing that some users need factual reporting or raw imagery. Fit the output to the actual need.

Begin with the reader’s decision

A good intelligence requirement is a specific question linked to an action. “Research this company” leaves the scope open indefinitely. “Does the available public evidence identify the legal entity that would sign our contract?” establishes a clearer task.

Ask the reader what they must decide, by when, and what level of uncertainty is acceptable. Identify the consequence of a mistaken conclusion. A low-stakes background note and a public allegation should not use the same threshold for review.

Write down the boundaries: jurisdictions, date range, named entities, and excluded personal information. This prevents a research assignment from expanding into unrelated profiling. It also makes it possible to explain what the resulting report does not cover.

A useful question can have an inconclusive answer. If the available sources cannot establish authority to sign, the report can explain that gap and recommend direct clarification. Uncertainty becomes actionable when the reader knows what it blocks and what evidence would resolve it.

Keep observations, assertions, and inferences separate

An observation describes what you directly inspected: a page displayed a particular sentence at retrieval. An assertion is something a source claims: the publisher says its product has a certification. An inference is your interpretation: the publication may indicate a planned market launch.

These categories can coexist in one source. A press release is evidence that an organization publicly made a claim. It does not independently verify every claim inside it. A court filing is evidence of what was filed, with a procedural role that must be understood before its contents are treated as findings.

Use wording that preserves the distinction. “The company announced a launch” is different from “The product launched.” “The complaint alleges a breach” is different from “The company breached the agreement.” A small change in the verb can change the factual meaning of a report.

Material	Supported observation	Further question
A supplier’s certification page	The supplier displays the claimed certification.	Does the issuing body confirm it for the relevant product and period?
A dated corporate filing	The filing reports a specified relationship or event.	Is it current, amended, and relevant to this exact entity?
An image with a location caption	The publisher associates the image with that place.	Do landmarks and source history support the caption?
Several articles repeating one release	The claim circulated through several publications.	Is there independent reporting or only one source lineage?

Evaluate the source and the claim independently

A source can be well positioned to know one fact and poorly positioned to know another. A company knows what it intends to announce but has an interest in presenting itself favorably. An official registry has a defined administrative role, but its contents may include information supplied by the registered entity.

Ask how the source knows, whether the statement is first-hand, what incentives could affect it, and whether the record is complete. Then evaluate the particular claim against other evidence. Do not give every sentence the same reliability rating because it appears on an official-looking page.

Record dates carefully. A source can be accurate for its publication period while being unsuitable for a current-status question. A historical ownership statement may explain a past decision without establishing who controls the entity today.

Where a source has a correction or amendment, examine the change. Preserve the earlier version if it matters to the question, but use the updated information when describing current facts. An audit trail should help explain the evolution of the record rather than freeze an error into the assessment.

Corroboration depends on independence and relevance

Count evidence lineages, not webpages. If five articles quote the same company statement, they provide broad dissemination of one assertion. A regulator’s separate record or a directly inspectable artifact may add evidence of a different kind.

Independence alone is also insufficient. Two independent sources may describe different entities or different periods. Match the evidence to the exact claim before combining it. A shared name is a starting point for identity resolution, not a reason to merge records.

Ask which observation would distinguish the explanations you are considering. A generic website logo may fit both a legitimate supplier and an impersonator. A confirmation through an established contact route has more diagnostic value for an unexpected payment request.

The CIA’s Tradecraft Primer describes structured techniques for examining assumptions and alternative explanations. Use such techniques to expose the reasoning. A table of hypotheses is useful because another person can challenge it, not because the table creates certainty.

Express uncertainty in terms the reader can use

Separate the likelihood of an event from confidence in your assessment. A statement about what may happen is different from a statement about the strength and limitations of the evidence used to judge it.

ODNI’s Analytic Standards distinguish underlying information, assumptions, judgments, uncertainty, and confidence. They govern a particular intelligence-community context; for this public-source guide, they are a reference for making reasoning and limitations explicit rather than a claim of official certification.

For a general audience, plain language can be more helpful than unexplained probability labels. Say which part is well supported and which part is weak. For example: “The entity identifier matches the contract, but the public records do not establish the representative’s current signing authority.”

Avoid a universal confidence percentage assembled from arbitrary points. Adding points for three source types and subtracting points for an old website can create the appearance of measurement without a validated model. Describe the actual evidence and the specific uncertainty instead.

Worked example: from supplier information to a decision brief

A fictional organization is considering a supplier that says it will deliver a specialized research instrument within a month. The reader asks whether the available public evidence supports including that supplier in the final purchasing round.

The initial collection contains a product webpage, a corporate record, a trade article, and a certification statement. A link list would simply present those four items. An assessment asks what each one establishes about the purchasing question.

The corporate record connects the legal name to the proposed contracting entity. The trade article repeats the supplier’s announcement, so it is not independent delivery evidence. The product page describes planned availability. The certification body’s public search does not resolve whether the named model is covered.

The resulting brief might say: “The legal-entity match is supported. Public materials describe a planned delivery date but do not demonstrate completed deliveries. Certification coverage for this model remains unresolved. Request the relevant certificate and delivery references before final selection.”

This output helps the buyer act without claiming that the supplier is dishonest or that the product cannot be delivered. It identifies the evidence boundary and the next useful information request.

A reusable one-page intelligence brief

Use the following structure when the reader needs a concise assessment. Keep a separate evidence log for detailed source handling. The brief should make the decision clear without forcing the reader to reconstruct the research from scratch.

1. Question and date: state the exact question and the time at which the assessment is current.
2. Key assessment: give the direct answer in one or two sentences.
3. Evidence: cite the strongest relevant records beside the claims they support.
4. Alternatives: explain another plausible interpretation and why it is weaker or unresolved.
5. Uncertainty: identify the missing or conflicting information that matters.
6. Implication: explain what the finding changes for this reader.
7. Next action and review trigger: identify the information or event that should prompt a new assessment.

Keep recommendations proportionate to your role. A researcher can recommend obtaining a missing document or escalating an unresolved issue. A legal, medical, or regulated compliance decision may require the responsible professional to interpret the evidence.

Put the most consequential caveat near the conclusion. A reader should not have to discover in a footnote that all reported relationships come from one unconfirmed source.

Related tools and what they contribute

Zotero can organize references and snapshots for selected sources. Check the imported author, date, and title, then connect each source to a specific claim. A well-formatted bibliography does not establish the truth of the linked material.

OpenRefine reconciliation can help compare messy records with an external dataset. Review candidate matches and keep stable identifiers. A name-similarity suggestion should not silently become a confirmed identity.

A spreadsheet is often enough for a small evidence matrix. Use columns for claim, source, date, observation, inference, limitation, and reviewer. An additional source-lineage column helps reveal when several records depend on the same underlying statement.

Use continuous monitoring when the question requires updates over time. Collection automation can identify changes, but it should not automatically publish a revised judgment without the relevant review.

Where AI can assist—and what must remain traceable

AI can help turn notes into a draft outline, suggest alternative questions, or identify inconsistencies for review. Keep the underlying records available and require every substantive factual claim to point to an inspected source.

Do not use an AI-generated source summary as a replacement for reading the document that matters to the conclusion. A summary can omit a qualification, confuse a party’s assertion with a finding, or join two similarly named entities.

Record where AI assistance affected the product. A useful disclosure states the actual contribution, such as drafting or organization. It should not claim that a model independently verified sources or conducted an investigation if that work did not happen.

For consequential reports, ask a reviewer to reconstruct one or two key claims directly from the sources. This spot-check tests the evidence connection and often reveals problems that a prose-only review misses.

Common failures that reduce decision value

Collecting without a question. Large archives can hide the absence of a decision. Revisit the requirement before adding another source.

Using stronger verbs than the evidence permits. Announced, alleged, reported, confirmed, and found describe different things. Choose the one supported by the source’s role and procedural status.

Treating missing results as a clean bill. Explain where you searched, the relevant coverage limits, and what an absent record cannot establish.

Failing to update a dated assessment. Put an as-of date on the report and define a review trigger. A useful conclusion can become stale when a filing, decision, or source correction changes the evidence.

Adding recommendations unrelated to the finding. If the evidence only establishes an identity mismatch, recommend resolving it. Do not jump to a sweeping conclusion about intent or future behavior.

Frequently asked questions

Is a verified fact already intelligence?

It can be enough for a particular user’s need. In a broader assessment, it becomes part of intelligence when its relevance and implications are explained. The distinction is about purpose and analytical work, not the prestige of the label.

Do more sources always produce a better answer?

No. More independent, relevant evidence can help. More copies of an assertion, or more records about a different entity, can create noise. Track source lineage and relevance before counting.

What if the evidence does not support a clear conclusion?

Report the uncertainty, explain its practical consequence, and identify the next test. “Not established by the available public records” can be a useful answer when it prevents an unsupported decision.

Further practice and sources

The linked primary references explain the formal concepts and tool features. The supplier scenario, distinctions, and brief template are editorial applications. For an end-to-end procedure, use the Landry Scientific Investigative Loop.

Visit the OSINT Research Hub, Free Research Library, and Start Here guide. For your next brief, write the decision question at the top, separate source claims from your conclusions, and make the missing evidence visible.