
Sources to check companies, associations and authorisations
Choose an identity register, notice, VAT validation or authorisation register for the fact to establish.
16 results
DIRECTORY / OPEN SOURCES
76 starting points for researching an entity, legal text, dataset, image or web trace, with each service’s evidentiary limit.
Search sources
01 / Choose a source for the question
A register records what was filed or published. An archive shows a capture. A map represents a dataset. None alone establishes every aspect of a claim. Define the question, jurisdiction and period, then retain the URL and access date.
Topic pages to select a service and understand what its result can establish.

Choose an identity register, notice, VAT validation or authorisation register for the fact to establish.
16 results

Retrieve the text, version and procedure while separating decisions, announcements and commentary.
6 results

Choose a producer, catalogue or series by unit, coverage and research question.
23 results

Document an older page or infrastructure while retaining observation dates and limits.
5 results

Choose a map, collaborative object or satellite product by scale and observation date.
6 results

Locate a publication and then verify territory, procedure and title status in the relevant registry.
6 results

Separate work, author, organisation, repository, journal and trial to cite the right object and version.
14 results
02 / DIRECTORY / OPEN SOURCES
Find a French entity by name, SIREN or SIRET and compare administrative records.
What it does not prove. This is an aggregator: check dates and the original register before drawing conclusions.
Consult France’s national business register data, trademarks, patents and designs.
What it does not prove. Read each registration or filing with its status, date and protection scope.
Search French civil and commercial notices by company, SIREN, category and publication date.
What it does not prove. A notice records a published event; it does not describe the entity’s entire current activity.
Find French statutes, codes and court decisions, including text versions.
What it does not prove. Use the version applicable on the relevant date and distinguish law from commentary.
Discover datasets and APIs published by French authorities and other producers. Distinguish the dataset record, each resource file and its associated API; formats and update dates may differ.
What it does not prove. Check the chosen dataset’s publisher, coverage, licence, method and last update.
Consult the Official Journal and European Union legal acts.
What it does not prove. Check the act, version, entry into force and relationship with national law.
Identify interest representatives registered with EU institutions.
What it does not prove. Registrants provide their own data; absence and accuracy both require caution.
Explore published European public procurement notices, tenders and awards.
What it does not prove. A notice does not prove contract delivery or outcome; read its stage and linked documents.
Search datasets indexed by the European data portal. Distinguish a dataset from its downloadable distributions and locate the producer’s record before choosing a file.
What it does not prove. The portal indexes metadata: open the publisher and check coverage and freshness.
View UK companies, officers, accounts and public filing history.
What it does not prove. Companies House states that it does not check the accuracy of filed information.
Search filings submitted to the US securities regulator.
What it does not prove. Read a filing with its period, form type and any amendments.
Query public domain, IP network and ASN registration data through RDAP.
What it does not prove. Some data may be redacted; registration does not prove who operates a website.
Find an archived web page, compare captures and retain the URL actually opened with its timestamp. Also check images and links needed for the analysis.
What it does not prove. A capture can be incomplete or combine resources from different dates. Archiving does not validate page claims; absence from replay does not automatically mean an element did not exist at the time.
Explore IGN maps, aerial photography and public territorial layers.
What it does not prove. Check each layer’s origin, scale and date before comparing places.
View and compare satellite imagery and metadata over time. Retain collection, scene identifier and display settings to distinguish the acquisition from the exported rendering.
What it does not prove. Cloud cover, resolution, processing and acquisition date constrain interpretation.
Browse Earth observation layers and compare dates or phenomena. Record the layer name, date and displayed area; retain the legend to explain the screenshot’s colours.
What it does not prove. A coloured visual depends on its layer and is not itself a ground measurement.
Search trademarks and related signs across participating international and national collections.
What it does not prove. Coverage varies; also check national registers and each mark’s status.
Search patent publications and trace technical developments across countries.
What it does not prove. A publication does not prove an enforceable patent: check family, territory and status.
Use a SIREN or SIRET to obtain an identity record for an entity in France’s Sirene register.
What it does not prove. Insee states that this status notice has no legal force.
Review French interest-representative declarations, organisations and reported activities.
What it does not prove. Reporting rules define coverage; absence or declared detail does not capture every form of influence.
Search interconnected business registers in the EU and selected EEA countries through BRIS.
What it does not prove. Covered company forms and documents vary by national register; inspect the originating record.
Search UK legislation in original and revised versions.
What it does not prove. The site does not hold every type of legislation; check amendment status and commencement.
Explore published US federal spending, contract and assistance award data.
What it does not prove. Check definitions, reporting thresholds, corrections and lag before comparing amounts.
Compare trademark applications and registrations from participating offices, including across the EU.
What it does not prove. A result alone establishes neither conflict nor availability; territory, classes and prior rights matter.
Compare European statistical series using dataset codes, dimensions, units and metadata. Keep dimension codes, filters and quality flags with the extract so the selection can be reproduced.
What it does not prove. Check method, revisions and coverage; the live database may not retain past versions.
Find EU framework-funded research projects, participants, reports and published deliverables.
What it does not prove. Funding or a project summary does not prove industrial results or market launch; read dated deliverables.
Search UK public procurement notices and their published stages.
What it does not prove. Coverage varies by period, threshold and nation; a notice alone does not confirm contract performance.
Search U.S. federal contract opportunities by agency, category and status.
What it does not prove. An open opportunity is neither an award nor a payment; cross-check award records and dates.
Check procedural stages, grant, opposition and legal information for a published European application.
What it does not prove. After grant, distinguish unitary effect from national parts. The EPO tracks unitary effect; national registers document the relevant national parts.
Search published PCT international applications and documents from participating national or regional offices.
What it does not prove. A PCT publication is not a worldwide patent; check national phases, grants and territories.
Search published French procurement, concession and award notices.
What it does not prove. A notice records one procedural stage; check corrections and documents before inferring delivery.
Find EU court judgments, orders and opinions by case or subject.
What it does not prove. Read the ruling and its reasoning; a summary or press release cannot replace them.
Consult decisions, sanctions and guidance published by France’s data protection authority.
What it does not prove. Distinguish guidance, formal notice and sanction; check date, scope and any appeal.
Explore digitized documents from BnF and its partners. Identify the edition, open the relevant view and retain the record and reproduction identifier.
What it does not prove. A digital copy does not validate every statement in the document. Check provenance, OCR errors and resource-specific reuse terms; a BnF building photograph’s license does not apply to those documents.
Locate fonds, finding aids and digitized documents at the French National Archives. Prepare consultation using the creator, exact reference and described dates.
What it does not prove. A record or finding aid is not consultation of the item. Check fonds context, holding location and access conditions for the reference; a site photograph does not confirm them.
Find identified French statistical time series to track measured change.
What it does not prove. Keep the series code, unit, coverage, method and revisions; different scopes are not directly comparable.
Access the registered-design search signposted by the EUIPO.
What it does not prove. A visual search does not establish complete rights coverage or freedom to operate; check status and territory.
Search datasets and information resources published by Canadian government institutions.
What it does not prove. Check the chosen dataset’s publisher, geographic coverage, licence, method and date.
Explore OECD socioeconomic series and indicators with their dimensions and metadata.
What it does not prove. Compare countries and periods only after checking units, definitions, breaks and revisions.
Find development indicators by country, period and subject in the World Bank portal.
What it does not prove. A value depends on its source, method, unit and coverage; it cannot establish a cause by itself.
Consult European Central Bank statistical series on money, markets and the economy.
What it does not prove. Retain series code, frequency, unit, adjustment and date; a movement alone does not prove a policy effect.
Explore the common United Nations system data and statistics gateway; identify the producer of each indicator.
What it does not prove. Check each table’s producer, coverage, definitions and date; two tables may follow different methods.
Explore open metadata for scholarly works, authors, institutions and citation links.
What it does not prove. Coverage and record matching vary; inclusion or citation counts do not validate scientific quality.
Search metadata deposited with Crossref by DOI, title, author or citation.
What it does not prove. A DOI identifies a record but does not prove peer review or guarantee that all deposited fields are complete.
Explore WHO health indicators by topic, country and period.
What it does not prove. Check definitions, coverage and dates; an aggregate series does not describe every local situation.
Consult UNESCO Institute for Statistics series on education, science and culture.
What it does not prove. Check indicator, country availability, method and year; missing data is not a zero value.
Search ILO statistics on employment, unemployment, wages and working conditions.
What it does not prove. Distinguish national surveys from modelled estimates, and check definitions and series breaks.
Examine IMF series on national accounts, prices, public finance and international trade.
What it does not prove. Retain dataset, frequency, units and revisions before comparing values.
Explore FAO statistics on agriculture, food, land and resources.
What it does not prove. Check domain, method, unit and year; apparent availability does not directly measure consumption.
Consult UNHCR maps, tables and documents for displacement situations tracked by the portal.
What it does not prove. Operational counts have dates, scope and method; they may be revised and do not cover every person.
Examine public routing, IP prefix and autonomous system data in RIPEstat. Identify the API endpoint, resource and time bounds; consult its metadata to understand methodology and coverage.
What it does not prove. A network announcement is a dated technical relationship; it does not prove website ownership or content authorship.
Find USGS satellite and aerial scenes by area, date and collection. Retain the product identifier and exported metadata; a thumbnail helps discovery without replacing the product to analyse.
What it does not prove. Check acquisition, sensor, resolution, clouds and processing level before interpreting an image.
Find a researcher identifier and the contributions or affiliations made public in their record.
What it does not prove. Check each item’s provenance; visibility and completeness vary by record, and an identifier does not certify expertise.
Distinguish research organisations through persistent identifiers, names and documented relationships.
What it does not prove. An organisation record helps disambiguate affiliations; it does not establish a person’s employment on a particular date.
Find published or unpublished research outputs, metadata records and deposited files in a multidisciplinary open archive.
What it does not prove. Distinguish metadata-only records, preprints and published versions; deposit alone is not scientific assessment.
Find biomedical and life-sciences citations and abstracts, then follow full-text links where available.
What it does not prove. PubMed is not a general full-text archive; indexing alone does not validate a finding or medical recommendation.
Search life-sciences literature, preprints and links to full text, data and related resources.
What it does not prove. Identify document type and version; records from different platforms may describe the same work.
Explore DOI-identified resources and metadata links to people, organisations and repositories.
What it does not prove. Check declared relationships against the resource; citations and downloads alone do not measure quality.
Find deposited datasets, software, publications and other research objects with metadata and identifiers.
What it does not prove. Retain the version and licence of the file used; public deposit does not guarantee reproducibility or content quality.
Search open-access journals, their records and articles indexed in the directory.
What it does not prove. Read journal criteria and records; inclusion does not validate every published finding or every site referring to the journal.
Search public clinical-study records and examine identifiers, declared protocols, dates and available results.
What it does not prove. Registration does not establish efficacy or safety; distinguish planned research, recruitment and published results.
Explore a collaborative map and geographic features described by contributors.
What it does not prove. Read feature history and tags; a collaborative map does not replace a field survey or cadastral document.
Explore research outputs and relationships to data, software, projects, people and organisations.
What it does not prove. Graph relationships are leads to check; multiple records may reuse the same original metadata.
Consult OSF projects, preregistrations and preprints, identifying resource type and version. The announced Projects transition starts on 16 November 2026; see the documented schedule in the profile.
What it does not prove. Distinguish a project file from a Registration. Announced preservation of content and identifiers establishes neither protocol adherence nor result quality.
Find an entity with an LEI and examine its legal identity, registration status and reported relationships.
What it does not prove. LEIs do not cover every business. LAPSED concerns renewal of the LEI registration; it does not establish dissolution of the entity.
Find legislative preparation stages, impact studies and public reports indexed by DILA.
What it does not prove. Since 1 April 2026, legislative dossiers previously available on Légifrance are published on Vie-publique. A dossier or summary does not replace the applicable legal text.
Find published notices and annual accounts for associations, foundations and endowment funds.
What it does not prove. Retain RNA, any SIREN, notice number and publication date. A notice and filed accounts do not establish the same operation.
Consult registry data for associations under the 1901 law, including RNA_waldec and RNA_import extracts.
What it does not prove. Coverage excludes Moselle, Bas-Rhin and Haut-Rhin. RNA_import concerns associations with no declared change since 2009; an old record does not establish current activity.
Validate a VAT number for intra-community transactions within the VIES scope.
What it does not prove. Separate valid, invalid and unavailable responses. Validation alone does not determine tax treatment, a contact’s identity or solvency.
Identify entities and authorisation scope in ACPR’s public banking and insurance register.
What it does not prove. The unified portal combines REGAFI and REFASSU since July 2026. Check identity, category, status and scope; an entity’s authorisation does not authenticate a domain or contact.
Find registration and categories of an insurance, banking or financial intermediary.
What it does not prove. Match ORIAS number, SIREN and category. Intermediary registration is not a credit institution licence and does not validate an offer or website using its name.
Find credit institutions and branches listed from competent authority information.
What it does not prove. The European register confers no rights and does not replace national registers. Check category, authority and country; authorisation scope does not cover every activity or product.
Find beneficiaries, purposes and amounts of EU funding covered by the financial transparency system.
What it does not prove. Read accounting year, expenditure type and management scope. The system is not an exhaustive inventory of every organisation’s funding and does not establish project impact.
Discover ongoing and completed trials indexed from multiple registries in ICTRP.
What it does not prove. Return to the source registry and trial ID. Two records from one trial are not two studies; registration does not establish published results or effectiveness.
Find civil-time rules, UTC offsets and daylight-saving changes in an identified tz database release. When comparing times, retain the zone identifier and database version used for conversion.
What it does not prove. A zone name is not a fixed offset. Rules depend on date and release; the database does not establish when an event occurred.
Identify UN statistical countries or areas and groupings using their M49 codes.
What it does not prove. M49 organises statistical units; it provides no boundary geometry and does not replace the territorial classification of a particular dataset.
No source matches these filters.
Dated notes concern the cited points, not validation of every field. A consultation date does not replace a document date.
03 / OSINT
Each route starts with a real question and opens services in a useful order. Its final note keeps conclusions proportional.
Follow workflow
Compare identifier, registration and published notices rather than relying on a trade name.

Separate French and EU law, then record the version in force on the relevant date.

Start with the TED notice, then seek relevant European or French public data for context.

Search relevant trademark collections, then distinguish patents from commercial signs.
OSINT / Keep
Every directory service has a record template: fields to retain and an interpretation check that goes beyond a screenshot.
04 / OSINT
The same discipline applies to a company register, satellite image or archived page.
State the claim, exact entity, period and jurisdiction. Avoid searching only a common name.
Use a stable identifier and open the act, notice, filing or dataset beyond the search snippet.
Record URL, publisher, publication or acquisition date, access date, version and status.
Compare with an independent source and separate established fact, declaration, clue, no result and hypothesis.
05 / FAQ
No. Coverage, publication delays, name variants and jurisdiction may explain it. Record the search and try other identifiers or registers.
It documents a captured version and supports dated comparison. It does not certify the content’s truth or the site’s completeness.
Keep the precise document URL, entity identifier, document date, version or status and access date. A search-results page alone is fragile.
Access varies by service, feature and volume. Some tools, including Copernicus Browser, may require a free account. Check the source’s terms.
The dataset may have been revised, or selected dimensions, units and periods may differ. Keep the code, selections, extraction date and method note before comparing.
No. PATENTSCOPE documents the international publication, among other records. Protection depends on subsequent procedures, decisions and territories; check the relevant registers.
DEEPER READING / PRACTICE
The directory brings together registers, archives, datasets and publications. Choose a source for the fact to establish and the type of document that could support it.
Updated on .
Define territory, entity, period and expected information. Check the source limits and then check the search conditions and the items that are actually available.
An institutional source can only cover a particular scope. The absence of result must be interpreted with the variants of the name and the limits of the service.
The search sheet retains selected source, query, result and known coverage. It specifies the additional traces when the result is insufficient.