LEXICON / 218 TERMS

Digital intelligence glossary

Operational definitions for clear discussion of AI, OSINT, search, visibility, evidence and sovereignty.

Search the 218 terms218 terms shown

Answer Engine Optimisation

AEO

Structure answers so answer engines can understand and reuse them directly.

AI agent

Agent IA

A system pursuing a bounded mission, using tools and following an action policy.

Authority

Autorité

Trust granted to a source based on evidence, consistency and citation.

Digital due diligence

Due diligence numérique

Structured verification of visible footprint, assets, relationships and risks.

Entity

Entité

A person, organisation, place, product or concept identified through attributes and relationships.

Generative Engine Optimisation

GEO

Optimisation of presence and citation inside generative answers.

Knowledge graph

Graphe de connaissances

A structured network of entities, properties and verifiable relationships.

Human in the loop

HITL

Human validation inside an automated chain before sensitive action.

Large Language Model

LLM

A model trained on large corpora to process and generate language.

LLM Optimisation

LLMO

Work on sources, formats and signals enabling LLM understanding, retrieval and citation.

Open Source Intelligence

OSINT

Intelligence produced from legally accessible and methodically qualified sources.

Provenance

Provenance

The chain linking a claim to its source, date and transformations.

Reversibility

Réversibilité

Ability to leave a tool or supplier without losing data, continuity or freedom of choice.

Search Engine Optimisation

SEO

Optimisation of content discovery and understanding by search engines.

Weak signal

Signal faible

An incomplete indicator whose trajectory may announce significant change.

Digital sovereignty

Souveraineté numérique

Ability to understand dependencies and preserve technical and decision options.

Search Experience Optimisation

SXO

Joint optimisation of visibility, understanding, journey and conversion.

Retrieval-Augmented Generation

RAG

Generation enriched by retrieving passages from a bounded corpus.

Structured data

Donnée structurée

Information expressed through an explicit vocabulary for machine interpretation.

Semantic footprint

Empreinte sémantique

The themes, attributes and associations publicly connected to an entity.

Answer share

Part de réponse

Frequency and quality of an entity’s presence across a panel of search or AI answers.

Prompt injection

Prompt injection

A hostile or untrusted instruction attempting to divert an AI system.

Decision log

Journal de décision

Record of options, arguments, decisions, owners and review conditions.

Degraded mode

Mode dégradé

Planned minimum operation when the normal system is unavailable.

AI Overview

AI Overview

A generative summary shown in some search results using information drawn from multiple web sources.

Answer engine

Moteur de réponse

An interface that directly formulates an answer and, where available, presents the sources used.

Canonical URL

URL canonique

The reference address indicated to search engines when the same content has several accessible variants.

Cited source

Source citée

A page or document explicitly presented as support for an answer, synthesis or claim.

Grounding

Ancrage factuel

Supporting a generated answer with information retrieved from identifiable sources.

IndexNow

IndexNow

A protocol for promptly notifying participating engines when a URL is added, changed or removed.

Knowledge panel

Panneau de connaissance

A panel that gathers structured information about an entity from sources considered relevant.

Multimodal search

Recherche multimodale

Search combining several input or result formats such as text, image, voice and video.

Retrieval

Récupération

The step of selecting from an index or corpus the items relevant to a query.

Semantic HTML

HTML sémantique

Use of HTML elements that clearly express the role and hierarchy of content.

Vector embedding

Plongement vectoriel

A numerical representation of content used to compare semantic similarity.

Crawl budget

Budget d’exploration

The volume and pace of URLs a search engine may choose to crawl on a site over a given period.

Crawlability

Explorabilité

The ability of an authorised crawler to reach a URL and follow required resources.

Indexability

Indexabilité

A page’s technical eligibility for an index, without a guarantee that it will be included.

Hreflang

Hreflang

An annotation connecting equivalent language versions of a page.

Canonical cluster

Cluster canonique

A set of URL variants a search engine may consolidate around one reference URL.

Primary source

Source primaire

The original document or actor directly responsible for information, considered in context.

Evidence register

Registre de preuves

A record linking a claim to its source, date, status and next verification.

Confidence level

Niveau de confiance

A documented assessment of how strongly available sources support a conclusion.

Correction log

Journal de correction

A dated record of corrected errors, reasons and replaced versions.

Question panel

Panel de questions

A stable set of questions used to compare AI answers over time.

Model variation

Variabilité des modèles

Differences in answers across models, sessions, contexts or observation times.

Digital provenance

Provenance numérique

The chain used to recover a content item’s origin, transformations and rights.

Content lifecycle

Cycle de vie éditorial

Rules for creating, validating, updating, merging and retiring content.

OAI-SearchBot

OAI-SearchBot

OpenAI crawler used to surface sites in ChatGPT search results; its access is controlled separately from GPTBot.

GPTBot

GPTBot

OpenAI crawler for content that may be used to train its foundation models.

Robots.txt

Robots.txt

A file telling compliant crawlers which URLs they may request; it does not make content private.

XML sitemap

Sitemap XML

A file signalling important URLs to search engines, with optional dates, images and language versions.

Core Web Vitals

Core Web Vitals

Three field experience metrics: main-content loading (LCP), responsiveness (INP) and visual stability (CLS).

LCP

Largest Contentful Paint

The time until the largest content element in the initial viewport is displayed.

INP

Interaction to Next Paint

A measure of how responsive a page is to interactions throughout a visit.

CLS

Cumulative Layout Shift

A measure of unexpected movement of visible elements during a page’s lifetime.

Entity disambiguation

Désambiguïsation d’entité

Distinguishing people, organisations or products with similar names using verifiable attributes.

Passage language

Langue du passage

The language of a text segment that differs from the page language and should be marked locally in the markup.

First-party data

Donnée de première main

Information collected directly by an organisation through its relationship with its audiences, with context and usage rights to establish.

Risk classification

Classification du risque

Grouping cases by impact, likelihood or context to choose proportionate review and controls.

Evaluation set

Jeu d’évaluation

A documented collection of representative and difficult cases used to test a system against stable criteria.

Acceptance criterion

Critère d’acceptation

An observable condition set before launch to decide whether a result is reliable enough.

Least privilege

Moindre privilège

Limiting a user’s, agent’s or tool’s permissions to the actions required for its task.

Tool boundary

Frontière d’outil

The explicit limit on data and actions an agent can access through a connected tool.

Supplier exit plan

Plan de sortie fournisseur

A documented process for recovering data, replacing integrations and maintaining service when changing suppliers.

Restore test

Test de restauration

A controlled trial confirming that a backup can actually return a service and its data to a usable state.

JSON-LD

JavaScript Object Notation for Linked Data

A format expressing machine-readable entities, properties and relationships in a page without replacing visible content.

Schema.org

Schema.org vocabulary

A shared vocabulary of types and properties used to describe entities and pages in structured data.

Rich result

Résultat enrichi

A search presentation that may use a supported structured data type; valid markup does not guarantee display.

Semantic validation

Validation sémantique

Checking that values declared in markup agree with facts readers can actually see on the page.

Regression test

Test de non-régression

Repeating the same cases after a change to detect a capability or boundary that no longer works.

Indirect prompt injection

Injection indirecte

A hostile instruction placed in a source read by an agent, such as a page or document, to redirect its actions.

Content drift

Dérive éditoriale

The gap that develops over time between a page, its sources, other pages and language versions.

Content consolidation

Fusion éditoriale

Combining similar pages into a more useful answer while explicitly handling old URLs and links.

Accessible name

Nom accessible

Text used by assistive technology to identify a control or field; it should match its visible purpose.

Focus order

Ordre de focus

The sequence in which keyboard focus moves between interactive elements while preserving the journey’s meaning.

Contrast ratio

Contraste

Difference in luminance between text or a component and its background, checked against the relevant criterion.

TTFB

Time to First Byte

Time from navigation start to the first response byte, including connection, redirects and server response.

FCP

First Contentful Paint

Time until the browser first renders visible content such as text or an image; it does not measure the largest element.

TBT

Total Blocking Time

A lab measure adding up the time long tasks block the main thread between FCP and Time to Interactive.

Field data

Données terrain

Measurements from real visits, interpreted by period, device population and sample size.

Server-side validation

Validation côté serveur

Checking received input on the server even when the browser provides immediate validation.

Alternative text

Texte alternatif

Text conveying an image’s purpose or information in context without needless repetition.

Decorative image

Image décorative

A visual with no information needed for the task; its alternative can be empty to avoid noise.

Source independence

Indépendance des sources

Two observations do not depend on the same collection; repeated accounts of one document remain one origin.

Circular sourcing

Reprise circulaire

Publications cite or repeat a claim without adding an independent observation.

Corroboration

Corroboration

Support for a claim from distinct items whose origin, context and independence have been checked.

Signal timeline

Chronologie des signaux

Dated observations separating when an event happened, was published and was discovered.

Timestamp

Horodatage

Time attached to a trace, whose precision, timezone and source need to be recorded.

Name collision

Homonymie

Different entities sharing a name; the label alone cannot establish identity.

Visual geolocation

Géolocalisation visuelle

Comparison of visual clues with spatial references to support a place claim without automatic certainty.

Reverse image search

Recherche inversée d’image

Search for indexed versions of an image; it does not guarantee the original or capture date.

Web archive

Archive web

A dated copy available to an archive service; it does not validate the page’s claims.

Provenance manifest

Manifeste de provenance

Assertions linked to a file and its history; technical validation does not prove a caption is true.

Agent mandate

Mandat d’agent

Explicit scope of tasks, data, tools and decisions an agent may use, including human approvals and exclusions.

Stop condition

Condition d’arrêt

Predefined rule that suspends action when result, cost or risk exceeds the accepted scope.

Reference sample

Échantillon de référence

Fixed set of questions, situations or pages used to compare observations across a stable scope.

Competing hypothesis

Hypothèse concurrente

Alternative explanation retained until the available evidence can distinguish it from others.

Escalation threshold

Seuil d’escalade

Observable criterion that moves a signal from routine monitoring to a designated decision or response.

100

Baseline

Ligne de base

Initial state measured with a known method and period, used to compare change without assuming its cause.

101

Control map

Carte des contrôles

Mapping of risks to safeguards, operating evidence, review frequency and owners.

102

Exit test

Essai de sortie

Bounded exercise checking whether data and functions can be transferred or restored outside a supplier.

103

SIREN

French legal-unit identifier

Nine-digit French identifier for a legal unit. It helps reconcile company records without implying that an address or establishment is the same.

104

SIRET

French establishment identifier

Fourteen-digit French establishment identifier: a SIREN followed by an internal number. An establishment change may alter the SIRET without changing the SIREN.

105

Sirene status notice

Avis de situation Sirene

Snapshot of publishable Sirene information for a legal unit or establishment; Insee states that it has no legal value.

106

RNE

French national business register

French business register accessible through the INPI portal. Compare a record at its consultation date with the available filings and documents.

107

BODACC

French official commercial notices bulletin

French publication of notices on events in business life. A notice documents a published event, not the whole current situation.

108

BRIS

Business Registers Interconnection System

European infrastructure connecting national business registers; useful for tracing a cross-border company back to the issuing register.

109

RDAP

Registration Data Access Protocol

Structured access protocol for domain or Internet resource registration data. Public fields may be limited and do not by themselves identify a site operator.

110

ASN

Autonomous System Number

Number identifying an autonomous system in Internet routing. A network relationship indicates infrastructure, not necessarily ownership of content.

111

Applicant

Déposant

Person or organisation named in a patent or trademark application. Distinguish it from the inventor, current owner and commercial operator.

112

Procurement notice

Avis de marché

Publication of a public purchasing stage. Check whether it announces competition, an award or a change before inferring a contract.

113

Dataset metadata

Métadonnées de jeu de données

Information on a dataset’s source, producer, period, coverage, format, updates and licence that shapes how it can be interpreted.

114

Temporal coverage

Couverture temporelle

Period represented by observations in a dataset, distinct from its publication and last update dates.

115

Reuse licence

Licence de réutilisation

Terms specifying permitted uses of a dataset or document and obligations such as attribution.

116

Capture date

Date de capture

Time when an archive service or investigator saved a version. It does not alone establish creation or first-publication time.

117

Acquisition date

Date d’acquisition

Time an image or measurement was collected by a sensor, distinct from product publication and later processing.

118

Cloud cover

Couverture nuageuse

Estimated share of a satellite image covered by clouds; a scene-wide figure does not guarantee visibility of the area under study.

119

Patent family

Famille de brevets

Group of patent documents linked through one or more priority claims; membership can vary with the database definition.

120

Nice classes

Classes de Nice

Administrative product and service categories used in trademark filings; a class alone does not define the exact scope of a filing.

121

Trademark status

Statut d’une marque

State shown for an application or registration on a given date. Check register, territory, listed goods and later events.

122

Document identifier

Identifiant de document

Registry reference for a specific item. Retain it with the URL, version and date to find the same item again.

123

Correction notice

Procurement amendment

Publication changing an earlier procurement notice, such as deadline, lot or conditions. Keep both versions and dates before comparing tenders.

124

Award notice

Procurement result

Publication stating an award decision or procedure result. It does not, by itself, prove delivery or contract performance.

125

Operative part

Decision outcome

Part of a judgment or authority decision stating what was decided. Read it with the reasoning, procedure and facts.

126

Advocate General opinion

Reasoned proposal

Analysis presented to the Court of Justice of the European Union in some cases. It informs the debate but is not the judgment.

127

Archival call number

Holding reference

Identifier assigned to an archival unit so it can be retrieved. Record it with the institution, fonds and description.

128

Archive fonds

Record group

Documents created or received by a person or organisation during its activity. Their origin helps interpret an individual item.

129

Finding aid

Archive description

Inventory or guide describing holdings and call numbers. It points to documents but its summary does not replace reading an item.

130

Digitised edition

Document reproduction

Digital version of an identifiable edition. Separate the work’s date, the copy’s date and its online publication.

131

Design right

Product appearance

Intellectual property relating to a product’s appearance in a territory and protection state. Distinguish it from a trade mark or patent.

132

Office of origin

Issuing register

Authority where an intellectual property right is filed or registered. Aggregators may show several offices with different coverage.

133

Appeal status

Procedural state

Dated indication of a known challenge to a decision. A published decision may still change through the procedure.

134

Cited passage

Precise location

Page, view, paragraph or exact passage supporting a claim in a document. It enables a check beyond a general link.

135

Statistical unit

Observed object

Person, establishment, transaction or area represented by one observation. Similarly named series may count different units.

136

Series break

Non-comparable change

Discontinuity caused by a changed definition, method, coverage or base year. An apparent trend may reflect that break.

137

Statistical revision

Recalculated value

Change to an already published figure after new information or methodological improvement. Keep each extraction date and version.

138

Data distribution

Dataset resource

File, API or other concrete representation of a dataset. One catalogue entry may point to distributions with different formats or dates.

139

File checksum

Integrity check

Digest calculated from a file to identify the exact version processed. A matching checksum does not establish data accuracy.

140

Metadata quality

Description quality

Extent to which producer, licence, resources and coverage are documented. It does not directly measure record accuracy.

141

Hero image

Image principale

The visual representing a page topic in its HTML content. It should agree with the title and context.

142

Responsive image variant

Variante responsive

An image version suited to screen width or density. It avoids unnecessary transfer while preserving a fallback source.

143

Image caption

Légende d’image

Visible text placing an image in context, such as subject, date or source, without overclaiming.

144

EXIF metadata

Métadonnées EXIF

Information embedded in some image files, such as description, creator or rights. It guarantees no search ranking.

145

Image rights metadata

Métadonnées de licence

Information about image owner, licence or rights-acquisition page; always check the original licence.

146

@id identifier

Identifiant @id

Stable URI joining nodes in a JSON-LD graph. Person, article and page should retain separate identities.

147

Rich result eligibility

Éligibilité aux résultats enrichis

A page meeting supported types, properties and provider rules. Eligibility does not guarantee rich result display.

148

Denominator

Dénominateur

Reference population or quantity used to calculate a rate. A different denominator may make figures incomparable.

149

Modelled estimate

Estimation modélisée

Value produced statistically when direct observations are incomplete. Retain model, assumptions and uncertainty range.

150

Editorial revision

Révision éditoriale

Verified update to content, sources, dates, links and language variants; distinguish a minor correction from a substantive change.

151

ORCID

Researcher identifier

Persistent identifier connecting a person with a record of contributions and affiliations. Check who supplied each entry and when; the profile does not prove current employment.

152

ROR

Organisation identifier

Persistent Research Organization Registry identifier distinguishing research organisations despite name variations. It does not independently validate an author’s affiliation.

153

PMID

PubMed record

Identifier of a bibliographic reference in PubMed. The record and abstract do not replace the full text; a PMID differs from a DOI and PMCID.

154

PMCID

Text archived in PMC

Identifier of a document in PubMed Central. Check its version, updates and reuse rights: free access does not remove copyright.

155

Preprint

Manuscript before review

Version made public before peer review. Identify its repository, date and version, then look for any later publication without treating the two texts as identical.

156

Memento

Prior resource state

In the RFC 7089 framework, a resource representing a prior state of another resource.

157

TimeMap

Archived-state list

List of Memento addresses for a resource. Distinguish listed states from readable ones.

158

WARC

Web preservation format

Format grouping resource records and metadata. Preserving a file does not resolve its interpretation.

159

Archive replay

Display of a preserved state

Consulting an archival resource through an interface. Record useful components and destinations actually read.

160

Idempotence

Stable intended effect

Property whereby repeating an operation has the same intended effect as executing it once. It does not guarantee an identical response or safety of every retry.

161

Control case

Reference result

Case with a known expected result used to compare observed behaviour. Describe the control and conditions; it does not replace real cases.

162

SHA-256

File digest

256-bit cryptographic digest of file bytes. Supports integrity comparison; it proves neither authorship, time nor factual truth.

163

Retraction

Publication status

Notice withdrawing a publication from reliable scientific literature. Read its scope and reasons, which do not automatically establish misconduct.

164

Expression of concern

Editorial alert

Notice of uncertainty about a publication during checks. Distinguish it from correction and retraction; preserve the notice with the text.

165

API pagination

Result traversal

Dividing results into pages or cursors. A successful response may return only part of a collection; check pages, unique identifiers and stopping rules.

166

RNA

Répertoire national des associations

French national association register; its identifier replaces neither SIREN nor SIRET and coverage excludes the Alsace-Moselle local regime.

167

VIES

VAT Information Exchange System

European system querying national VAT databases. Distinguish its dated response from an availability error.

168

Clinical trial identifier

Stable registry reference for a trial. Secondary identifiers connect registrations of one trial without making them independent studies.

169

Data dictionary

Description of dataset fields: meaning, type, format, unit, codes, missing values, keys and schema version.

170

Missing value

Absence of a measurement or information whose cause needs documentation. Zero, empty string and not applicable are not automatically equivalent.

171

Composite key

Fields jointly identifying a row, such as entity and period. Removing a component can create multiple matches.

172

Working translation

Translation prepared for analysis or reading, distinct from a publisher-issued version. It links to the original and explains ambiguous terms.

173

Financial authorisation

Competent authority permission for a defined scope. It validates neither every product nor a contact copying an authorised entity’s name.

174

Data grain

Object and level represented by one row. Grain determines keys and valid aggregation: company, establishment and transaction differ.

175

Join cardinality

Number of matching rows between tables: one-to-one, many-to-one or many-to-many. Row multiplication can change totals.

176

UTC

UTC

Time reference for expressing an instant. UTC conversion needs a known offset; it does not certify the source clock.

177

Calendar date

Calendar date

A calendar day without an implied time or offset. Converting it to midnight UTC adds unsupported precision.

178

UTC offset

UTC offset

Difference between local time and UTC at an instant. A fixed offset does not describe every rule of an IANA zone.

179

M49

M49

UN Statistics Division classification of countries or areas and statistical groupings. Its codes contain no geometry.

180

Constant prices

Constant prices

Amounts expressed at a reference price basis to separate price change. Method and index must suit the variable; currency conversion is insufficient.

181

Statistical suppression

Statistical suppression

Withholding a value to protect statistical information. A suppressed cell is not zero; retain its status and respect the protection.

182

SDMX

Statistical exchange

Standard for statistical data and metadata exchange. Producer structures define dimensions, codes and attributes; a shared standard does not guarantee identical methods.

183

DSD

Series structure

Data Structure Definition: SDMX structure describing dimensions, measures and attributes. Dimension order helps interpret series keys.

184

DCAT

Catalogue vocabulary

W3C vocabulary describing catalogues, datasets, distributions and data services. It makes relationships explicit without certifying quality or availability.

185

Catalogue record

Registration metadata

Description of a resource’s registration in a catalogue. Its modification date may differ from that of the described dataset or file.

186

GeoJSON

Geographic format

JSON format for geographic geometries and features. RFC 7946 uses WGS 84 in degrees, ordered longitude then latitude.

187

Coordinate reference system

Axis meaning

System giving meaning to coordinates, axes and units. Identify reference systems and transformations before comparing numeric pairs.

188

Check digit

Transcription detection

Character or digits calculated from an identifier to detect some copying errors. A coherent check proves neither assignment nor identity.

189

ISBN-13

Edition and format

Thirteen-digit publication identifier starting with 978 or 979, with weighted modulus-10 check. Verify edition and format; arithmetic does not verify allocated ranges.

190

ISSN

Continuing resource

Eight-character identifier of a continuing resource, ending with a check digit or X. ISSN assignment does not guarantee content quality.

191

Soft 404

Soft 404

A page returned with a success status, usually 200, whose content signals absence or an error. Status alone therefore does not prove that the expected resource was served.

192

Sampling frame

Sampling frame

The list or mechanism used to select units from a population. Record what it covers and omits: random selection from an incomplete list does not correct its coverage error.

193

HTTP 202 — request accepted

202 Accepted

A response indicating that a request was accepted for processing, without confirming completion or eventual success. Keep the offered monitoring mechanism rather than concluding from this code alone.

194

CSV dialect

Reading parameters

A table’s delimiter, quoting, encoding and header conventions. Establish them before counting cells or interpreting values.

195

Exact duplicate

Repeated record

A record whose values all match another record. It differs from a repeated key; deletion requires knowing the intended grain.

196

Observed candidate key

Uniqueness in one file

A column or combination with no missing or repeated values in the studied file. This observation does not guarantee uniqueness in future versions.

197

Surrounding whitespace

Leading or trailing character

Whitespace before or after a value. It may explain failed exact matches; remove it only under a documented rule suitable for the data.

198

Unresolved source conflict

Open disagreement

Incompatible claims after aligning object, time and scope. Retain both documents and state the check that could resolve the conflict.

199

Redirect chain

Successive HTTP destinations

Responses and addresses between a requested URL and its final destination. A successful final status does not prove the cited document was preserved.

200

Pagination cursor

Collection continuation

Service-provided value for continuing through results. Keep it with filters and version; it need not be a number or an indefinitely reusable address.

201

Statistical stock

Position at a date

Quantity describing a position at a reference date. Summed monthly stocks are not automatically an annual stock; define the measure before aggregation.

202

Statistical flow

Events during an interval

Quantity covering events over a period. Before adding flows, check compatible definitions and non-overlapping intervals.

203

Cumulative total

Sum from an origin

Value summing flows from a defined starting point. Retain origin, resets and revisions; changes need not represent new events.

204

Percentage point

Difference between rates

Difference unit for percentages: 20% to 25% is +5 points. This differs from the +25% relative change on the initial base.

205

Relative change

Change against a base

Final minus initial value divided by the initial value. Multiply by 100 for a percentage; it is undefined on a zero base.

206

Negative search

No result within scope

Search with no matching result under documented query, filters, service and date. It is not by itself general evidence that the fact is absent.

207

End-to-end latency

Duration of a defined journey

Time between explicitly selected events, such as request sent and accepted output. Name boundaries: first displayed content is not journey completion.

208

Cold cache

No relevant cached reuse

Test condition without relevant data in the cache under study. State which cache; a fresh browser does not guarantee an empty server cache.

209

Warm cache

Possible reuse state

Test condition allowing reuse of relevant data or components. Describe preparation and separate it from first calls in comparisons.

210

Constant geographic scope

Territorial comparison

Method comparing observations for the same territorial units. It separates changes in the phenomenon from changes created by adding or removing a territory.

211

Weighted mean

Denominator weights

Mean accounting for each observation’s weight. To combine comparable rates, use their relevant denominators; an unweighted mean gives equal weight to different population sizes.

212

Statistical vintage

Dataset version

State of a series available at a date or in a particular release, including revisions then known. Keeping that vintage allows results to be reproduced after updates.

213

Conditional rate

Selected denominator population

Proportion calculated within a defined subset, such as accessible answers. Report its selection condition and excluded counts without treating it as a rate for all attempts.

214

HTTP 206 — partial content

Range response

Response delivering part of the requested representation. Read the received range and request context; an accessible portion does not establish that the whole document was examined.

215

Imputation

Estimated missing value

Method assigning an estimated value to a missing observation under explicit assumptions. Keep the imputation marker and original missing state; the result remains an estimate even when published by the producer.

216

URL fragment

Destination within a resource

Part after # in a URL, often identifying a section. Its presence in a link does not establish that the page still has the corresponding target after migration.

217

Private cache

Reuse for one user

Cache dedicated to one user, for example in their browser. Its state can differ from a shared cache; a new-session test does not represent every previously stored resource.

218

Shared cache

Reuse for multiple users

Cache able to reuse responses for multiple users, for example at an HTTP intermediary. Check its version and rules separately from the browser; renewing a session does not ensure its update.

Ten distinctions for choosing the right concept

These ideas overlap, but they answer different questions. Compare them before setting a goal or metric.

Search Engine Optimisation / Generative Engine Optimisation

SEO concerns how search engines discover and understand a page. GEO examines its presence and citation in a generated answer. A strong search position does not guarantee accurate attribution in that answer.

See definition : Search Engine Optimisation See definition : Generative Engine Optimisation

Generative Engine Optimisation / LLM Optimisation

GEO concerns how an entity or content is represented in generated answers. LLMO more broadly develops the sources, formats and signals that help models retrieve, interpret and cite that information.

See definition : Generative Engine Optimisation See definition : LLM Optimisation

Large Language Model / Retrieval-Augmented Generation

An LLM is the model producing text; RAG is an architecture that supplies retrieved passages from a corpus. Adding RAG does not automatically make sources accurate, current or correctly cited.

See definition : Large Language Model See definition : Retrieval-Augmented Generation

Entity / Knowledge graph

An entity is something identifiable through attributes and links. A knowledge graph connects multiple entities and makes their relationships explicit. Verify the type, source and date of a relationship before adding it.

See definition : Entity See definition : Knowledge graph

Provenance / Authority

Provenance traces where a claim came from and how it changed. Authority expresses how much confidence a source deserves for a particular question. An identifiable origin helps assessment but does not itself make a source reliable.

See definition : Provenance See definition : Authority

Digital due diligence / Open Source Intelligence

OSINT is a method for collecting and analysing open sources. Digital due diligence can use it to examine an organisation, asset or risk before a decision. Due diligence needs a defined scope and explicit criteria for a conclusion.

See definition : Digital due diligence See definition : Open Source Intelligence

Reversibility / Digital sovereignty

Reversibility is tested when leaving a supplier or tool. Digital sovereignty is the broader ability to understand dependencies and retain choices. A theoretical export is insufficient if no one can reuse it.

See definition : Reversibility See definition : Digital sovereignty

Human in the loop / Decision log

HITL inserts human approval into an automated flow before selected actions. A decision log retains reasons, owners and review conditions for decisions. Approval without a record makes a system harder to explain and improve.

See definition : Human in the loop See definition : Decision log

Structured data / Semantic HTML

Structured data describes facts explicitly using a vocabulary. Semantic HTML describes the roles of visible page elements. Both should match what readers actually see.

See definition : Structured data See definition : Semantic HTML

Cited source / Grounding

A cited source is a reference shown as support for an answer. Grounding means building the answer from identifiable sources. A displayed citation alone does not prove that every sentence follows from it.

See definition : Cited source See definition : Grounding

Technical and documentary references

Technical rules, registers and procedures change. Check their current scope in official documentation.

NEXT / GUIDES

Move from definition to action

Guides

DEEPER READING / PRACTICE

Read the terms with their useful distinctions

The aim is to make the need, choices and next steps clear to everyone involved. Defining the scope is useful when it helps people make decisions and check the effects of an action.

Editorial supplement added on 8 October 2026.

Starting situation

Close words are used as synonyms while they describe different operations. Glossary helps to stabilize the vocabulary before comparing methods or results.

Priority checks

Relate the term to a concrete example and to neighboring concepts. Check the definition retained in the source document when the word has multiple uses.

Interpretation pitfalls

A general definition may not correspond to the meaning of a specific indicator or register. The context of the document remains necessary.

Working record to retain

A vocabulary sheet retains term, work definition, distinction and example. It helps the speakers to use the same words for the same transactions.

Apply the approach to an initial case

Starting with a delimited case allows to confront intentions with real constraints. In this case, specify the purpose, the available information and the person who can arbitrate. Then extend the scope when the first observations justify that change, rather than fix all the steps on the basis of unverified assumptions.

Present the result and its limitations

When presenting the results, separate the decisions taken from the topics still open. A expected result must be described accurately enough so that the team can recognize its achievement. The next checkpoint focuses on a observable change and provides for what to be revised if the result is not obtained.

PRACTICAL CASE / WORKED EXAMPLE

Two people use the same term differently.

The following case illustrates an approach to defining the need. Situations are pedagogical: they serve to examine the choices and conditions of success, without describing a fulfilled mission or announcing a guaranteed result.

Example context

One team talks about proof to designate a available document, while another reserves this word to support a specific proposal. The conclusions are difficult to compare.

Suggested approach

Defining the terms in the context of the file and give an example. Discern document, observation and conclusion. Revise the formulations where ambiguity changes the scope of the word.

Conclusion and next step

The common vocabulary allows to discuss methods without misunderstanding. The work definition remains related to the context and can be specified when a new case appears.

Checklist for applying this example

Formulate the expected decision and identify the person who can arbitrate.
Describe the known restrictions and the information still needed.
Compare options according to criteria understandable by the participants.
Join the next action to a observable result and a review.

Questions for deeper practice

How to use the glossary in a project?

Choose the terms that really influence exchanges and decisions. A shared short vocabulary, with distinctions and examples, is often more useful than an exhaustive list without any connection with the work.

How to move from a diagnosis to a concrete action?

Choose an action whose role in the problem is explicit, check the requirements and designate a manager. describe what will show improvement. The first step should also produce useful information if the result is not the expected, so that the choice is revised instead of automatically extending the plan.

How to keep a trace of decisions without delaying the project?

A short report can record question, options, choices, reason and next control. The detailed subjects remain linked to the report for a thorough reading. Updating decisions when their conditions change is more useful than accumulating accounts made without indicating the actions and responsibilities that arise from them.

DETAILED WORKSHOP / APPLICATION

Stable the words that change the reading of a result

This workshop offers a five-step sequence to apply the page labels. The entries, observations and conclusions are kept together in order to facilitate the verification and recovery of the work.

01 / Prepare the necessary information
Choose the terms used in the project and the passages where they become ambiguous. Preparing concrete examples and definitions specific to the source documents.
02 / Carry out the approach
Compare usages and then define a work vocabulary. Discern neighboring notions and specify the corresponding operations. Revise the conclusions whose meaning depended on a word used differently.
03 / Examine an alternative situation
If a term has a specialized use in a source, keep that context. The general definition of the glossary does not automatically replace the definition of the indicator or register.
04 / Decide on the next step
Vocabulary is useful when it reduces misunderstandings about methods and results. Differences must be illustrated by cases of real work.
05 / Hand over the verification records
Transmit a term, definition, difference and example. It can be enriched when new situations show ambiguity without multiplying unnecessary entries.

ANNOTATED WORKING DOCUMENT / REVIEW GRID

A vocabulary list with distinctions and examples

The list retains terms, work definition, context and related notions. It connects words to the project’s operations to reduce misunderstandings about methods and results.

On a small screen, scroll the grid horizontally to read all three columns.

Example situations to assess
Observation	Supported interpretation	Next step
The speakers use the same relevant definition.	Exchanges can compare transactions without ambiguity on this term.	Keep the example and review the new uses.
The source document uses a specialized sense.	The general definition does not replace that of that indicator or register.	Add the context and clarify the difference.
A word combines evidence, observation and hypothesis.	The formulation can hide different levels of support.	Separate the concepts and review the ambiguous conclusions.

Reviewing the working document

The final verification verifies that the vocabulary improves real understanding. A useful vocabulary must be able to illustrate its differences with the cases of work, without claiming to impose a unique meaning on all documents.

REPORTING / EXAMPLE WORDING

Presentation of results: Read the terms with their useful distinctions

Point to clarify

The work definition must preserve the specialized uses useful to the file.

Before handing over

Check the examples and review the conclusions that depend on ambiguity.

Example of a qualified conclusion

The term is used according to the distinction and context indicated. The other senses found in the sources are presented when they change the reading.