Open evidence · Edition 1.1.14

Edikka’s library of instruments

Instruments made to be examined, reused and cited. Protocols, audit grids, registers and datasets are published with their source, version and limits.

Browse the 26 instruments Download the complete edition Explore the demonstrations

Instruments: 26; Documented instruments to observe, audit, decide and automate under human control.
Explore the catalogue
Catalogued files: 172; Grids, protocols, registers and datasets collected in a versioned archive.
Download the archive
DOI deposits: 6; 6 deposits associated with 4 of the 26 instruments, including the MIA-FR protocol and its J0 and J+30 results.
Explore source studies
Open licence: CC BY 4.0; Reusable resources with attribution to Edikka, the source and the version used.
Read the licence

Demo and GitHub links open in a new tab.

Keyboard interaction

HTML & ARIA

What we observeAn ARIA button role alone does not add keyboard access.

Compare three identical-looking controls. In the archived tests, only the native button is reached with Tab and activated with Enter or Space.

English guide; original variants and observations in French. Safari/macOS: enable keyboard navigation. VoiceOver/Safari has not been tested.

Explore the demonstration Code on GitHub Test protocol

Content extraction

Semantic HTML

What we observeThe same caveat is retained in a paragraph and lost in an aside by Readability.

Explore this French case with Readability 0.6.0, then compare the archived outputs from seven tools.

One documented case. Attachment to the claim is not evaluated; this is not a general verdict on aside.

Explore the demonstration Code on GitHub Test protocol

Visual stability

Images & stability

What we observeA visible shift can still fall below the “good” CLS threshold.

Without reserved space, the marker moves by 233 px in the mobile viewport, with a CLS of 0.051149 — below 0.1. With reserved space: 0 px in these trials.

12 archived Chrome laboratory trials, observed over 0–6 s. No field data or overall Core Web Vitals assessment.

Explore the demonstration Code on GitHub Test protocol

Archived link paths

Internal linking

What we observeRemoving one relation leaves 34 shortest paths of three links.

Explore the archived graph, examine each relation and compare a reversible hypothesis. The original example has 37 shortest paths.

Archive from 9 September 2026. A simulation of structure, without a ranking or traffic prediction.

Explore the demonstration Code on GitHub Test protocol

Editorial review

Can we publish this title?

What we observeA technically acceptable title can still lack evidence.

Compare sources and proposals in five fictional scenarios. Correct a claim, test a lock and see why a changed version requires another review.

A local simulation, separate from the six read-only Edikka observations. No automatic approval or actual publication.

Explore the demonstration Code on GitHub Test protocol

UX writing & recovery

Forms & receipt

What we observeA lost response can hide an already recorded request.

Compare wording, correct errors and check receipt. Four scenarios show what the interface knows and how a retry can preserve a single record.

Fictional data. Public simulation and a reproducible local HTTP lab. No message sent to Edikka.

Explore the demonstration Code on GitHub Test protocol

Decisions & coverage

SEO migration review

What we observeFictional case: 14 missing URLs account for 31% of the recorded clicks.

A plan without loops or chains can still omit important URLs. Compare the fictional plan with its inventory, inspect the evidence and complete the missing decisions.

French interface and documentation. CSV files stay in your browser. No crawl or traffic-loss forecast.

Explore the demonstration Code on GitHub Test protocol

Content & markup

Page & JSON-LD

What we observeValid JSON can still describe an older version of the page.

Compare two templates across eight fictional scenarios, then review your own HTML files locally. Dates, authors, FAQ and breadcrumbs, with evidence for each field and JSON/CSV exports.

French interface and documentation. A shared record can also be wrong. Agreement does not certify truth or Google eligibility.

Explore the demonstration Code on GitHub Test protocol

Rules & declared intent

robots.txt review

What we observeAdding a GPTBot restriction can reopen a path excluded by the general group.

Compare the proposal with its correction across six synthetic scenarios. Review two files against your expectations, inspect decisive lines and export the evidence as JSON or CSV.

French interface and documentation. Local analysis. Rules do not prove crawler compliance and do not enforce access control.

Explore the demonstration Code on GitHub Test protocol

Complete catalogue

Choose the evidence before the format.

Start from the decision to be supported. Every source page explains the context; downloadable files come second.

Observe

Studies, baselines and protocols that document a state without turning a snapshot into a universal claim.

Public-evidence study Reference instrument

French GEO public-evidence observatory · wave 1

Thirty GEO-provider websites observed through 18 checks and 300 editorial decisions, published only as anonymised aggregates.

Distinguish public evidence from a commercial claim.

Version: 1.0.2
Status: Published · anonymised aggregates
Published: 2026-07-16
Languages: FR · EN

Read the HTML source

Preview & files 3 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
Check	Observed	Inconclusive
Pages lisibles	24	6
Offre indexable	24	6
Canonical cohérente	23	6

Read the original file — French GEO public-evidence observatory · wave 1 · v1.0.2 · original excerpt in French

Dated observations from this edition; these are not a new assessment of the current sites.

Cite this version

Edikka (2026). French GEO public-evidence observatory · wave 1 (v1.0.2). https://doi.org/10.5281/zenodo.21943813. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

A deterministic, non-random, single-wave panel; no agency ranking or individual result.

Public-observation dataset

French web-accessibility observatory · pilot wave 01

A dated record of public statements and signals across twelve major French-city websites, including observed sources.

Read an accessibility statement within its dated scope.

Version: 1.1.0
Status: Published · pilot wave
Published: 2026-06-23
Languages: FR · EN

Read the HTML source

Publication on data.gouv.fr

Documented edition 1.1.0 of the 23 June 2026 observations: CSV, JSON, method, dictionary, schema and provenance. The deposit adds no new collection.

Preview & files 6 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
Site	Published state	Declared rate
Paris	Partiellement conforme	53,52 % critères RGAA
Marseille	Non détecté	Not recorded
Lyon	Partiellement conforme	81,03 %

Read the original file — French web-accessibility observatory · pilot wave 01 · v1.1.0 · original excerpt in French

Dated observations from this edition; these are not a new assessment of the current sites.

Cite this version

Edikka (2026). French web-accessibility observatory · pilot wave 01 (v1.1.0). https://www.edikka.com/en/insights/accessibility-observatory#library-source-accessibility-observatory-wave-01. Accessed 2026-10-11. CC BY 4.0.

Version history from the instrument

v1.1.0 · 2026-09-10 Documented edition of the 23 June 2026 observations: numeric rates, dictionary, schema and provenance. No new collection. Version 1.0 remains available at its original URLs.

Report an issue with this version by email

Interpretation limit

Observation of public declarative signals: this dataset is neither an RGAA audit nor an exhaustive conformance assessment.

Multi-engine baseline

France SEO/GEO baseline · 27 August 2026

Twenty queries replayed in Google France, Google AI Mode and Perplexity to freeze an initial state of rankings and citations.

Choose a stable query set before comparing SEO/GEO observations.

Version: 1.0.0
Status: Frozen baseline · J+30 follow-up available
Published: 2026-08-27
Languages: FR

Read the HTML source

Manifest JSON — France SEO/GEO baseline · 27 August 2026

France SEO/GEO follow-up · J+30 · dataset dated 26 September 2026

The same 20 queries, one pass across three surfaces. Complete: Google France 20/20, Google AI Mode 13/20, Perplexity 6/20. The 21 incomplete observations do not mean absence. The baseline is unchanged.

Read the J+30 findings and limitations

J+30 files · v1.0.0

Preview & files 5 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Corpus query	Intent
Q01	qu'est-ce que le GEO	definition
Q02	GEO définition marketing	definition
Q03	Generative Engine Optimization	definition

Read the original file — France SEO/GEO baseline · 27 August 2026 · v1.0.0 · original excerpt in French

Cite this version

Edikka (2026). France SEO/GEO baseline · 27 August 2026 (v1.0.0). https://www.edikka.com/en/insights/seo/what-is-geo#library-source-seo-geo-france-baseline-2026-08-27. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

One pass per environment: neither probabilistic frequency nor future stability can be inferred.

Bilingual claim register

Edikka GEO Claim Register

Thirty-nine GEO claims linked to sources, evidence levels, limitations and maintenance status in French and English.

Check which sources support a GEO claim.

Version: 1.1.1
Status: Published · maintained
Published: 2026-08-27
Languages: FR · EN

Read the HTML source

Preview & files 5 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Claim	Status
GEO-001	The framework named Generative Engine Optimization was introduced by Aggarwal et al. in a November 2023 preprint and later published at KDD 2024.	Supported
GEO-002	The founding study reports visibility gains of up to 40% in its experimental environment.	Supported
GEO-003	The founding study’s gains concern content already present in a fixed context; they do not demonstrate its organic discovery.	Supported

Read the original file — Edikka GEO Claim Register · v1.1.1

Cite this version

Edikka (2026). Edikka GEO Claim Register (v1.1.1). https://doi.org/10.5281/zenodo.22128746. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

The register qualifies documentary claims; it guarantees neither rankings, citations nor commercial outcomes.

AI-visibility measurement protocol

MIA-FR protocol

A workbook, an initial 25-prompt corpus and an observation schema separating citations, sources, mentions, impressions and post-exposure value.

Define prompts and conditions before measuring AI visibility.

Version: 1.0
Status: Unchanged protocol · J0 and J+30 available
Published: 2026-08-26
Languages: FR · EN

Read the HTML source

External record · no endorsement

This record covers the protocol only, not the J0 results. Listing derived from Zenodo metadata; neither an evaluation nor an endorsement by NLM or NIH.

MIA-FR follow-up · J+30 · dataset dated 26 September 2026

25 French prompts, four platforms, three repetitions: 295 valid runs out of 300. J0 comparison, invalid runs and documentary checks retained. Protocol v1.0 is unchanged.

Read the J+30 findings and limitations

J+30 · DOI 10.5281/zenodo.23052222

Executable code and calculation inputs · supplement 1.1.1 · Collection conditions and corrections. Separate supplement dated 30 September; the original files and archive remain unchanged.

J+30 files · v1.0.0

AI Overviews and AI Mode · exploratory case dated 8 October 2026

Five French questions, three rounds across two surfaces, initial annotations and a French/English audit worksheet. Separate from MIA-FR J0/J+30; independent review has not been completed.

Read the J+30 findings and limitations

J+30 files · v1.0.0

Preview & files 3 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Corpus prompt	Intent
P001	Comment mesurer sa visibilité dans les réponses IA en France ?	Informationnelle
P002	Qu’est-ce que la visibilité IA d’une marque et comment la définir ?	Informationnelle
P003	Quelle différence entre visibilité IA, visibilité SEO et trafic organique ?	Comparative

Read the original file — MIA-FR protocol · v1.0 · original excerpt in French

Cite this version

Edikka (2026). MIA-FR protocol (v1.0). https://doi.org/10.5281/zenodo.22106520. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

Visibility varies by service, session, location and date; J0 results are not a forecast.

Audit

Grids and registers connecting every check to a method, evidence, status and limitation.

44-check audit grid Reference instrument

Technical SEO audit grid

Forty-four checks across thirteen domains, with method, expected evidence, required access, severity, owner and decision.

Connect an SEO check to the evidence needed for a conclusion.

Version: 1.1
Status: Published · bilingual
Published: 2026-08-22
Languages: FR · EN

Read the HTML source

Preview & files 6 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Check	Expected evidence
TS01	The final page returns an actionable HTTP status code.	Tested URL, timestamp and final 200 status for a public canonical page.
TS02	Redirects are intentional, direct and loop-free.	One final destination, documented chain and no loop.
TS03	Server errors and soft 404s do not replace an explicit response.	Consistent 4xx/5xx responses and an error template that does not return 200 by default.

Read the original file — Technical SEO audit grid · v1.1

Cite this version

Edikka (2026). Technical SEO audit grid (v1.1). https://www.edikka.com/en/insights/web-development/technical-seo-foundations#library-source-technical-seo-audit-grid. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

A public observation proves neither Search Console, logs nor configuration; the grid makes no ranking promise.

Measurement and counter-test protocol

Core Web Vitals protocol

An LCP, INP and CLS protocol separating field data, first-party RUM, lab runs and traces, with a local collector.

Qualify the source and population before interpreting Web Vitals.

Version: 1.2
Status: Published · quarterly review
Published: 2026-09-04
Languages: FR · EN

Read the HTML source

Manifest JSON — Core Web Vitals protocol

Preview & files 4 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
Metric	Unit	Good threshold
LCP	ms	2500
INP	ms	200
CLS	score	0.1

Read the original file — Core Web Vitals protocol · v1.2

Cite this version

Edikka (2026). Core Web Vitals protocol (v1.2). https://www.edikka.com/en/insights/web-development/core-web-vitals-guide#library-source-core-web-vitals-protocol. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

A lab run does not establish population experience; public field data remains distinct from local evidence.

Audit grid and evidence collector

Web image audit grid

Thirty-six checks across weight, responsive selection, quality, LCP, CLS, accessibility, image SEO, caching and CI.

Document which images need correction and the supporting evidence.

Version: 1.1
Status: Published · replayable collector
Published: 2026-08-23
Languages: FR · EN

Read the HTML source

Observe image loading

Images and visual stability: observe before correcting

What happens when an image arrives after the text? Compare the same image with and without reserved space, then inspect twelve archived laboratory trials. Linked to controls IMG22 and IMG23 of the existing grid.

These links open in a new tab.

Explore the demonstration Browse the code Reproduce the measurements

Demo 1.0.0 · grid 1.1. The visual replay and archived laboratory CLS are distinct. No field data is collected.

Preview & files 6 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Check	Expected evidence
IMG01	Is every visible image inventoried with its final URL, template and role?	Dated inventory with page URL, currentSrc, role and owner.
IMG02	Is each image classified as LCP, content, decoration, interface, social or data?	Reviewed role and short rationale for each priority image.
IMG03	Do critical images have an owner and a review date?	Owner, last review and next review date.

Read the original file — Web image audit grid · v1.1

Cite this version

Edikka (2026). Web image audit grid (v1.1). https://www.edikka.com/en/insights/web-development/web-image-optimisation#library-source-web-image-audit-grid. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

The grid separates browser facts from decisions; the historical v1.0 replay does not validate the corrected IMG07 check in v1.1.

Bilingual 36-check foundation

Professional website accessibility foundation

Thirty-six checks in eight families, with statuses, severities, decisions, expected evidence and a dated public self-assessment.

Prepare a journey assessment and identify checks needing further review.

Version: 1.1
Status: Public candidate
Published: 2026-08-26
Languages: FR · EN

Read the HTML source

Practical companion · English guide, original trials in French

HTML and ARIA: compare three controls with a keyboard

Try a clickable div, an ARIA role and a native button. Compare keyboard access with the targeted Playwright snapshots.

These links open in a new tab.

Try the demonstration Browse the code Reproduce the tests

Observation dates, configurations and status · Screen readers and independent AI agents: not tested.

Demo version and history: README. Code under MIT; separate from the instrument version and its CC BY 4.0 licence.

Preview & files 5 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Check	Expected evidence
A11Y01	Identifiable page title	Browser-tab screenshot and head extract.
A11Y02	Declared primary language	DOM extract and test log.
A11Y03	Coherent heading hierarchy	Exported heading outline.

Read the original file — Professional website accessibility foundation · v1.1

Cite this version

Edikka (2026). Professional website accessibility foundation (v1.1). https://www.edikka.com/en/insights/web-development/web-accessibility-professional-website-basics#library-source-professional-website-accessibility-foundation. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

A public candidate for journey diagnosis; the distributed workbook remains edition 1.0 and the package is neither a certification nor an RGAA conformance audit.

Eight-axis grid and validation protocol

Landing Page audit grid

Eight weighted axes, 0/1/2 criteria, expected evidence, validation schemas and a public self-audit of three journeys.

Prepare a landing page using explicit criteria and evidence.

Version: 1.0
Status: Published · external validation pending
Published: 2026-08-21
Languages: FR · EN

Read the HTML source

Preview & files 7 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Area	Expected evidence
LP01	Intent and message match	Campaign brief, query, ad, email, or documented segment.
LP02	Promise and value	Cold comprehension test and hero copy.
LP03	Evidence and trust	URL, dated capture, case, method, author, or scoped result.

Read the original file — Landing Page audit grid · v1.0

Cite this version

Edikka (2026). Landing Page audit grid (v1.0). https://www.edikka.com/en/insights/ux-ui-design/effective-landing-page#library-source-landing-page-audit-grid. Accessed 2026-10-11. CC BY 4.0.

Version history from the instrument

v1.0 · 2026-08-21 Publication des huit axes, poids, niveaux, preuves, formule et seuils.
v1.0.1 · 2026-08-21 Précision de la résolution et du protocole inter-évaluateurs ; poids et seuils inchangés.
v1.0.2 · 2026-08-21 Parité documentaire FR/EN, rattachement du résultat primaire à son DOI et métadonnées descriptives.
v1.0.3 · 2026-09-07 Noms canoniques, libellé 01 harmonisé et règle de fraîcheur des scores ; poids et seuils inchangés.

Report an issue with this version by email

Interpretation limit

A preparation heuristic with no predictive power over conversion; external validation is preregistered but has no results yet.

Documented decision register

Digital audit decision register

Findings applied to edikka.com: proposition, evidence, treatment, priority, owner and counter-check.

Turn a documented finding into a decision, owner and follow-up check.

Version: 1.2.5
Status: Published · verifiable register
Published: 2026-10-06
Languages: FR · EN

Read the HTML source

Preview & files 6 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Dated observation	Recorded decision
ADA05	No commercial page was said to carry the generative-search visibility offer.	Finding withdrawn. No additional page is created.
ADA19	The typography filter inserts non-breaking spaces into technical syntax and applies French punctuation to English content.	Fix the source filter: French prose only, inherited language, literal elements, URLs, attributes and JSON-LD preserved. Fix published and production acceptance passed on 6 October 2026.
ADA07	Does this article meet the field Core Web Vitals criteria for LCP, INP and CLS?	Close the documented data search for this edition. Keep field performance undetermined; transfer neither a home-page result to the article nor a laboratory score to field INP.

Read the original file — Digital audit decision register · v1.2.5

Dated observations from this edition; these are not a new assessment of the current sites.

Cite this version

Edikka (2026). Digital audit decision register (v1.2.5). https://www.edikka.com/en/insights/digital-strategy/digital-audit-complete-method#library-source-digital-audit-decision-register. Accessed 2026-10-11. CC BY 4.0.

Version history from the instrument

v1.2.5 · 2026-10-06 Production acceptance and closure of ADA03/ADA19. ADA21 reveals an empty confirmation despite a received enquiry: FR/EN fix published, four submission/receipt scenarios verified and check closed. Plan now focuses on ADA20 then ADA11, with clearer next-check dates and examples. Generic audio capsules retained.
v1.2.4 · 2026-10-06 First publication of this edition on 6 October 2026: shared Insight presentation and reading calculation, with new French and English audio capsules. Production acceptance remains to be verified for ADA03, ADA19 and ADA21.
v1.2.3 · 2026-10-06 Shared reveal sequence restored for the heading, baseline, image and introduction, with the shared hidden-tab fallback. Rich table cells made readable through the shared component. ADA03 decision clarified; observations and counts preserved. Local edition, not deployed.
v1.2.2 · 2026-10-06 ADA21 remains open and To test for preparing, submitting and receiving an enquiry; navigation evidence preserved. Introductory figure now counts defects to verify; redundant sentence removed. Local edition, with no deployment or submission test.
v1.2.1 · 2026-10-06 Priorities reordered, figures added to the introduction and shared Insight order restored, older audio moved to history, ADA02 lesson added and Expertise → redesign offer → contact journey observed in production (ADA21). Local edition, not deployed.
v1.2.0 · 2026-10-06 Standard FR/EN layout, qualified-enquiry objective, hidden-startup fix, explicit ADA02 result, archived page-filtered FR/EN Search Console exports and new ADA20 finding on English visibility. Not deployed.
v1.1.0 · 2026-10-05 Local edition: reader introduction and decisions in both languages, treatment states separate from evidence, reviewed dependencies, 5 October PageSpeed and Bing checks, integration of 1 October Search Console exports and new ADA19 typography finding. ADA03 remains historically unresolved; ADA09 now explicitly qualifies the disproved original claim. This edition has not been deployed.
v1.0.1 · 2026-09-07 Editorial-integrity correction: targets made bilingual, ADA03 evidence strength downgraded because no dated artifact is attached, CRAWL linked to ADA09 and crawl scope clarified for ADA13. States, severities, due dates and counts are unchanged.
v1.0 · 2026-09-05 Register frozen after counter-verification. Eighteen findings on edikka.com: nine established, five contradicted, two to test and two out of scope. The public crawl and pre-existing accessibility evidence invalidated two additional statements. No average score.

Report an issue with this version by email

Interpretation limit

An Edikka-specific case with no aggregate score or generalisation to other organisations; unknowns remain “to test”.

Bilingual 16-check protocol

Edikka UX writing protocol

Sixteen before/after checks across three Edikka interfaces, covering labels, accessible names, errors and fidelity to evidence.

Check whether interface copy helps people understand and act.

Version: 1.0.1
Status: Published · quarterly review
Published: 2026-09-03
Languages: FR · EN

Read the HTML source

UX writing · errors · receipt

“Message sent”: was your request actually received?

A lost response can hide an already recorded request. Compare wording, correct an error, recover after a refusal and check receipt without creating a duplicate. Four fictional scenarios extend the UX writing protocol.

These links open in a new tab.

Try the form states Browse the code Reproduce the checks

Public browser simulation and reproducible local HTTP lab. No message is sent to Edikka. Historical UXW observations are unchanged; no global accessibility or conversion claim.

Manifest JSON — Edikka UX writing protocol

Preview & files 2 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Check	Acceptance
UXW01	The overall verdict only depends on scans that pass the published reliability threshold.	An inconclusive result is excluded from the confirmed rank and cannot create or erase a lead or lag.
UXW02	Every inconclusive result is named and counted.	Readers know how many results are excluded and why without opening every scan detail.
UXW03	The wording reproduces the actually confirmed gap.	The point difference and its direction match the reliable subset.

Read the original file — Edikka UX writing protocol · v1.0.1

Cite this version

Edikka (2026). Edikka UX writing protocol (v1.0.1). https://www.edikka.com/en/insights/ux-ui-design/ux-writing-clearer-interfaces#library-source-ux-writing-protocol. Accessed 2026-10-11. CC BY 4.0.

Version history from the instrument

v1.0.1 · 2026-09-03 Corrected the per-interface result distribution in the public summary. The sixteen observations, nine corrected gaps and UXW06 to-test result remain unchanged.
v1.0 · 2026-09-03 Protocol frozen; three interfaces audited; nine gaps corrected; UXW06 kept to test.

Report an issue with this version by email

Interpretation limit

An internal three-surface case; outstanding screen-reader tests are not reported as passed.

Matrix, documented example and data package

SEO Architecture Decision Matrix

A reusable matrix, the complete normalised 188-page graph and a standalone script for replaying the Edikka paths without a private archive.

Version: 1.1.0
Status: Version 1.1.0 · standalone replay
Published: 2026-09-21
Languages: FR · EN

Read the HTML source

Explore · verify · simulate

Internal linking: which paths lead to your key pages?

Choose a starting URL and a destination, inspect each archived relation and compare a reversible simulation. In the documented example, removing the FAQ link leaves 34 shortest paths of three links to the French guide.

These links open in a new tab.

Explore paths Browse the code Reproduce the calculation

Archive: 9 September 2026 · data package 1.1.0 · demo 1.0.0. A calculation on a dated graph, not an observation of the current site or a prediction of SEO performance.

Manifest JSON — SEO Architecture Decision Matrix

Preview & files 11 files

Interpretation limit

The graph uses a 9 September archived crawl that post-dates the 15 August intervention; it proves the calculation within that scope, without Search Console data or an SEO effect.

Protocol, cases and collected answers

Citation Fidelity Lab

Three bilingual cases for inspecting scope, exclusions and commitments in an answer, with a GitHub demo and condition-level findings.

Compare information with its restatement, condition by condition, without an overall score.

Version: 0.5.0
Status: 107 complete responses in the new series · archives preserved
Published: 2026-10-11
Languages: FR · EN

Read the HTML source

Manifest JSON — Citation Fidelity Lab

Preview & files 6 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
Case	Examined condition
price	Price, tax and exclusions
time	Indicative duration and signed proposal
hosting	Documented capability and boundary

Read the original file — Citation Fidelity Lab · v0.5.0

Cite this version

Edikka (2026). Citation Fidelity Lab (v0.5.0). https://www.edikka.com/en/insights/seo/ai-citable-website#library-source-citation-fidelity-lab. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

Separate series and biased pilot archived. Explicitly requested diagnostic; C changes length and repetition. One model, simulated retrieval. No citation uplift or certified overall fidelity. MIT code; CC BY 4.0 text.

Bilingual protocol, raw outputs and reproduction

Semantic HTML extraction protocol

Twelve controlled test documents, seven extractors and converters, eight real-page snapshots and one DOM/accessibility-tree scenario, with results and pinned versions.

Version: 1.2.1
Status: Version 1.2.1 · dated review
Published: 2026-09-21
Languages: FR · EN

Read the HTML source

Explore the archived results

Semantic HTML: compare what tools preserve

Compare two variants of a heading, a table or an essential caveat across seven tools. Inspect the outputs, corrected observations and their source files.

These links open in a new tab.

Explore the results Browse the code Reproduce the protocol

Archived extractions, no live test. Package 1.2.1 · protocol 1.2.0 · revised evaluation 2.0.0. The demo has its own version and code licence.

Separate experiment · JSON-LD 1.0.0

Which author does an extractor return?

14 fictional FR/EN inputs, 7 tools, 98 observations per run. Three identical local runs compare identifier references, inline names, microdata and JavaScript injection. No search engine or AI-citation test.

Read the results and limits Method · TXT Source and results · ZIP Cite the dataset · DOI

Original data: CC BY 4.0. Code: MIT. Protocol separate from the HTML study and the Page & JSON-LD agreement tool.

Preview & files 6 files

Interpretation limit

These observations measure neither accessibility conformance nor indexing, ranking or citation. Observed instability in one newspaper4k output is retained.

Decide

Matrices and templates for assigning ownership, framing a project and making GO/NO-GO decisions explicit.

Bilingual matrix of 24 decisions Reference instrument

UX/UI Responsibility Matrix

Twenty-four interface decisions classified as UX-led, UI-led, shared or evidence-required, with acceptance, owner and deliverable.

Assign an interface decision and identify the expected evidence.

Version: 1.0
Status: Published · source matrix
Published: 2026-09-03
Languages: FR · EN

Read the HTML source

Manifest JSON — UX/UI Responsibility Matrix

Preview & files 2 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Situation	Lead
UXUI01	Ordering steps in a complex form	UX / interaction design
UXUI02	Structuring navigation and information architecture	UX / information architecture
UXUI03	Designing recovery after interruption	UX / interaction design

Read the original file — UX/UI Responsibility Matrix · v1.0

Cite this version

Edikka (2026). UX/UI Responsibility Matrix (v1.0). https://www.edikka.com/en/insights/ux-ui-design/ux-vs-ui-differences#library-source-ux-ui-responsibility-matrix. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

Assigning responsibility proves neither usability, preference nor business impact.

Bilingual decision application

Conversion application of the UX/UI matrix

Application of the 24 UX/UI decisions to abandonment, comprehension, trust and conversion hypotheses.

Separate a conversion hypothesis from an effect actually measured.

Version: 1.0
Status: Published · based on UXUI01–24
Published: 2026-09-04
Languages: FR · EN

Read the HTML source

Manifest JSON — Conversion application of the UX/UI matrix

Preview & files 2 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Context	Required evidence
UXUI01	Order the steps of an enquiry, sign-up or payment flow.	Task testing for progression; funnel measurement or a controlled experiment for business effect.
UXUI02	Connect intent, offer, evidence and action in the information architecture.	Tree or findability testing for discovery; journey data for impact on enquiries.
UXUI03	Let people resume an interrupted action without losing valid data.	An interruption scenario for recovery; measured comparison for abandonment.

Read the original file — Conversion application of the UX/UI matrix · v1.0

Cite this version

Edikka (2026). Conversion application of the UX/UI matrix (v1.0). https://www.edikka.com/en/insights/ux-ui-design/design-that-converts#library-source-ux-ui-conversion-application. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

An interface may make a decision easier; this dataset demonstrates no conversion uplift.

Website redesign risk register

Twelve risks, RF01 to RF12, linked to signals, evidence, owners and four GO/NO-GO decision gates.

Decide whether a redesign stage can pass its next GO/NO-GO gate.

Version: 1.0
Status: Published · no aggregate score
Published: 2026-08-23
Languages: FR · EN

Read the HTML source

Preview & files 2 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Risk	Expected evidence
RF01	Refondre sans problème démontré	Diagnostic daté reliant chaque problème à un utilisateur, un objectif métier, un signal mesurable ou une contrainte vérifiable.
RF02	Commencer sans baseline exploitable	Dossier d’état initial horodaté : crawl, exports Search Console et analytics, parcours, formulaires, mesures terrain et captures de configuration utiles.
RF03	Ne pas nommer le propriétaire de chaque décision	Matrice de responsabilités indiquant décideur, contributeurs, délai de réponse et règle d’escalade pour chaque famille de décision.

Read the original file — Website redesign risk register · v1.0 · original excerpt in French

Cite this version

Edikka (2026). Website redesign risk register (v1.0). https://www.edikka.com/en/insights/digital-strategy/website-redesign-costly-mistakes#library-source-website-redesign-risk-register. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

A decision-documentation instrument with no budget, schedule, compliance, ranking or conversion guarantee.

Transparent calculation workbook

Website redesign budget estimator

A visible-formula workbook estimating initial budget, recurring costs and three-year total cost from editable assumptions.

Make budget assumptions explicit before requesting a quote.

Version: 2026-08-20
Status: Published · open formulas
Published: 2026-08-20
Languages: FR · EN

Read the HTML source

Preview & files 2 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
Item	Unit	Initial quantity
Diagnosis and framing	fixed scope	1
Information architecture	fixed scope	1
Editorial support	page	5

Read the original file — Website redesign budget estimator · v2026-08-20

Cite this version

Edikka (2026). Website redesign budget estimator (v2026-08-20). https://www.edikka.com/en/insights/digital-strategy/website-redesign-cost#library-source-website-redesign-budget-estimator. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

An indicative estimator based on entered assumptions and Edikka pricing policy; it is not a quotation.

Editable framing template

B2B website redesign brief

A template connecting requirements, acceptance criteria, evidence, owners and blocking status before procurement or launch.

Write verifiable requirements before selecting a supplier.

Version: 1.3
Status: Published · editable
Published: 2026-10-06
Languages: FR · EN

Read the HTML source

Preview & files 2 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
To specify	Template prompt
Organisation and offers	[Activities, markets, languages, regions, sales cycle, sector constraints]
Role of the current website	[Lead generation, information, recruitment, support, portal, other]
Reasons for the redesign	[Observed obstacles, debt, brand change, new needs, deadline]

Read the original file — B2B website redesign brief · v1.3

Cite this version

Edikka (2026). B2B website redesign brief (v1.3). https://www.edikka.com/en/insights/digital-strategy/b2b-website-redesign-brief#library-source-b2b-website-redesign-brief. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

The template must be adapted to context, contract and risk level; it does not replace legal review.

GO/NO-GO acceptance workbook

Pre-launch website acceptance matrix

Fifty-four pre-launch checks, fifteen post-launch checks and three decision gates with evidence and owners.

Document acceptance testing and reservations before a launch decision.

Version: 1.0
Status: Published · no global score
Published: 2026-08-21
Languages: FR · EN

Read the HTML source

Preview & files 2 files

Inside the instrument

Three excerpts from the published file · abridged where necessary
ID	Check	Expected evidence
PM-01	Critical pages available	Dated crawl, URL, HTTP status and final destination.
PM-02	Primary navigation	Manual journey and link export.
PM-03	Footer and cross-site links	Link crawl and human review.

Read the original file — Pre-launch website acceptance matrix · v1.0

Cite this version

Edikka (2026). Pre-launch website acceptance matrix (v1.0). https://www.edikka.com/en/insights/web-development/website-acceptance-testing-before-launch#library-source-website-acceptance-testing-matrix. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

The matrix organises agreed acceptance; it does not guarantee defect-free delivery or future website outcomes.

Review files, corpus and extractor

Metadata Review Kit

A bilingual review profile, SERP observation sheet, dated Edikka corpus and dependency-free script extracting title, description, H1, canonical, fingerprints and similarities.

Version: 1.0.2
Status: Published · human review pending
Published: 2026-09-22
Languages: FR · EN

Read the HTML source

Sources · checks · editorial decision

Can we publish this title?

A technically acceptable title can still lack evidence. Explore five fictional scenarios: unsupported claims, similar titles, a lock, changed evidence and an unchanged replay. The six archived Edikka observations remain separate and read-only.

These links open in a new tab.

Try the scenarios Browse the code Reproduce the checks

Kit 1.0.2 · observations from 22 September 2026 · demo 1.0.0. A local simulation: no publication, automatic approval or predicted SEO outcome.

Manifest JSON — Metadata Review Kit

Preview & files 8 files

Interpretation limit

The corpus is limited to Edikka, human decisions remain pending and similarity alerts are neither an SEO score nor an editorial verdict.

Automate with controls

Contracts and evaluation sets for assisting production without delegating human validation.

Output JSON Schema Reference instrument

AI-assisted SEO recommendation contract

A machine contract requiring recommendation types, evidence, confidence, risks, application traceability and human validation.

Validate a recommendation’s structure before human review.

Version: 2026-08-19
Status: Published · mandatory human validation
Published: 2026-08-19
Languages: FR · EN

Read the HTML source

Preview & files 1 file

Inside the instrument

Three excerpts from the published file · abridged where necessary
Field	Type	Constraint
type	string	["brief", "title", "meta_description", "internal_link", "content_gap", "cannibalization_risk"]
confidence	string	["low", "medium", "high"]
requires_human_validation	boolean	{"const": true}

Read the original file — AI-assisted SEO recommendation contract · v2026-08-19

Cite this version

Edikka (2026). AI-assisted SEO recommendation contract (v2026-08-19). https://www.edikka.com/en/insights/ai-web-automation/ai-seo-automation#library-source-ai-seo-recommendation-contract. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

JSON SCHEMA · 2026-08-19 · FR + EN · 2.9 KB Open — automatisation-seo-ia-recommandations.schema.json — AI-assisted SEO recommendation contract

Interpretation limit

The schema validates structure, not recommendation truth; no “published” decision is allowed.

JSONL scenarios and control kit

Reliable AI evaluation set

Nominal, missing-data, ambiguity, contradictory-source, injection, secret, tool-failure and regression cases.

Test missing-data and ambiguous cases before delegating a task to AI.

Version: 3.0.0
Status: 12 scenarios · reproducible controls
Published: 2026-10-07
Languages: FR · EN

Read the HTML source

Preview & files 5 files

Inside the instrument

Three excerpts from the published file · abridged where necessary · synthetic examples
ID	Family	Expected decision
EVAL-001	nominal	ready_for_review
EVAL-002	missing_required_data	clarify
EVAL-003	ambiguity	clarify

Read the original file — Reliable AI evaluation set · v3.0.0

Cite this version

Edikka (2026). Reliable AI evaluation set (v3.0.0). https://www.edikka.com/en/insights/ai-web-automation/reliable-ai-prompts-business-rules#library-source-reliable-ai-evaluation-set. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

Interpretation limit

Synthetic scenarios to adapt; the kit checks controls, without measuring a model or verifying response meaning. MIT code, CC BY 4.0 data.

Matrix, journal and server controls

AI back office: from approval to a saved change

Separate review from execution; exercise permissions, concurrent conflicts, replay and rollback.

Version: 1.0.0
Status: Published · synthetic bench
Published: 2026-10-11
Languages: FR · EN

Read the HTML source

Preview & files 8 files

Interpretation limit

Local synthetic bench: 18 scenarios, no model calls or human measurements. Simulated authentication; no security certification. MIT code, CC BY 4.0 data.

Bilingual governance workbook

AI-assisted FAQ matrix

Twelve synthetic B2B questions with sources, priority, answer contract, owners, statuses and acceptance tests.

Connect every FAQ answer to a source and human approval.

Version: 2026-08-19
Status: Published · synthetic data
Published: 2026-08-19
Languages: FR · EN

Read the HTML source

Preview & files 1 file

Inside the instrument

Three excerpts from the published file · abridged where necessary · synthetic examples
ID	Anonymised question	Example status
FAQ-001	Quel est le délai de déploiement multi-sites ?	To validate / À valider
FAQ-002	La solution est-elle compatible avec Microsoft Entra ID ?	Approved / Approuvé
FAQ-003	Quel est le budget minimal ?	To validate / À valider

Read the original file — AI-assisted FAQ matrix · v2026-08-19 · original excerpt in French

Cite this version

Edikka (2026). AI-assisted FAQ matrix (v2026-08-19). https://www.edikka.com/en/insights/ai-web-automation/ai-assisted-faq-customer-questions#library-source-ai-assisted-faq-matrix. Accessed 2026-10-11. CC BY 4.0.

Version history: this catalogue documents the version shown above. No earlier change log is provided here.

Report an issue with this version by email

XLSX · FR + EN · 16 KB Open — matrice-faq-assistee-ia-questions-clients.xlsx — AI-assisted FAQ matrix

Interpretation limit

A synthetic case with no demonstrated traffic, conversion, support-deflection or AI-citation gain.

Reuse

Cite the instrument, not just the file.

Keep the title, version, publication date, publisher and source URL. When a DOI exists, use it as the permanent identifier.

Suggested citation for this catalogue

Edikka (2026). Edikka Open Instrument Library, edition 1.1.14. https://www.edikka.com/en/library

Edikka — instrument title, version, source URL and access date. CC BY 4.0.

Catalogue governance

A curated edition, maintained and versioned.

Edition 1.1.14 adds the back-office suggestion control kit, its 18 server scenarios and FR/EN matrix. Previous editions and their limitations are retained. No human measurements or model calls are claimed.

Maintenance remains versioned; no recurring publication obligation is created. Every new addition must support a distinct decision, expose its sources and retain unfavourable results.

Editorial direction and maintenance : Bertrand Morel · ORCID · Contact

Help improve this edition.

An error, a missing source or a reuse to share? Your email draft includes the edition. Add the instrument, your evidence and the context; do not include confidential client data.

Report an error by email Share a reuse by email

No email app configured? Use the contact form.

External review is invited, not claimed as completed. A reported issue must be checked before a correction is published.

Catalogue version history

1.1.0 · 2026-09-20
Two instruments added: the SEO Architecture decision matrix and semantic HTML extraction protocol, with evidence, complete packages and reproduction instructions. No DOI, accessibility certification or SEO outcome is claimed.
1.0.2 · 2026-09-11
Accessibility observatory updated to documented edition 1.1.0: six files identical to the data.gouv.fr deposit, method, provenance, checksums and aligned citations. The original observations and the 21-instrument scope are preserved.
1.0.1 · 2026-09-11
Resource identities aligned, source membership added, previews and citations made accessible, vocabulary documented and feedback channels added. The collection remains at 21 instruments; the instrument files are unchanged.
Download archive 1.0.1
1.0.0 · 2026-09-08
First edition: 21 instruments, bilingual catalogue, archive and checksums. The NLM catalogue reference was clarified on 9 September.
Download the previous archive

For search engines and agents

The entire library. One file.

One versioned bilingual archive contains all 172 catalogued files and their verification dependencies, both JSON and Markdown catalogues, citation metadata, the licence, a manifest and verifiable SHA-256 checksums.

ZIP Download the complete archive 172 catalogued files · proof dependencies · FR/EN catalogues · SHA-256 manifest · 7.3 MB

Archive edition 1.1.14 · 2026-10-11

Page updated on 2026-10-11. The MIA-FR supplement dated 30 September is linked separately; the original archive is unchanged.

Verify the downloaded files

Extract the ZIP, open a terminal in its root folder and run the command for your system. Every listed file must return OK. This checks file integrity, not the validity of its conclusions.

Linux

sha256sum -c SHA256SUMS

macOS

shasum -a 256 -c SHA256SUMS

From evidence to decision

Need to apply one of these instruments to a real project?

Edikka can frame the perimeter, conduct the review and turn the findings into a prioritised decision.

Discuss the project