Skip to content
Repliq[1]

The method

Six dimensions to measure how citable a site is by AI.

Every Repliq audit gives a score out of 100, built from six weighted dimensions. The weighting reflects how much each factor actually matters in a generative engine's decision to cite a brand or not. The sections below detail what Repliq checks in each one.

Citabilité IA et visibilité25%
Autorité de marque20%
Qualité de contenu et E‑E‑A‑T20%
Fondations techniques15%
Données structurées10%
Optimisation par plateforme10%
Chaque axe est mis à l'échelle du poids le plus élevé (25%) : plus une dimension pèse dans le score, plus son point s'approche du bord.

The six weights add up to 100% of the score.

01 AI citability and visibility

25% of the score

How likely a generative engine is to cite or rephrase a page from the site.

This dimension measures whether generative engines (Google AI Overviews, ChatGPT, Perplexity, Gemini, Bing Copilot) cite the site on queries related to the brand and its industry. Repliq tests real queries, records which sources are cited instead of the client, and identifies the page formats AI systems pick up most readily: a direct answer at the top of a section, sourced figures, a question-and-answer structure.

02 Brand authority

20% of the score

The strength of the entity graph: Wikipedia, Wikidata, press coverage, verified social profiles.

Generative engines cross-check several sources before describing a brand. This dimension checks the consistency of the name, figures and description across the site, Wikipedia, Wikidata, social profiles and press coverage. An outdated Wikipedia entry or contradictory figures between two pages weaken the trust an AI places in the brand, even when the site itself is well built.

03 Content quality and E-E-A-T

20% of the score

Experience, expertise, authoritativeness and trustworthiness, measured page by page.

Repliq evaluates the actual depth of each page, the presence of an identifiable author, a visible publication or update date, and whether statistics are attributed to a source. A 250-word page with no author or date gives an AI less to work with when justifying its answer than an 800-word page structured as questions and answers, signed and dated.

04 Technical foundations

15% of the score

AI crawler access, performance, and the absence of technical blockers.

This dimension checks that generative engine crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) can actually read the site: robots.txt, meta robots tags, server-side rendering of the main content, load time and visual stability. A site that relies entirely on client-side JavaScript to display its main text is often unreadable for these crawlers.

05 Structured data

10% of the score

JSON-LD Organization, Article, FAQPage, Product: what an AI can extract without guessing.

Schema.org markup gives an AI facts that are already qualified: who the author is, what the entity is, what the questions and answers on a page are. Repliq checks the presence, validity and coverage of this markup on strategic pages, and flags missing schemas that would force the AI to infer information instead of reading it.

06 Platform-by-platform optimization

10% of the score

The behavior gaps between Google AI Overviews, ChatGPT, Perplexity, Gemini and Bing Copilot.

Each generative engine favors different signals: Perplexity favors pages with sourced citations, Google AI Overviews relies on the classic index, Bing Copilot cross-references its own entity graph. Repliq tests the site on each platform separately and gives a readiness score per engine rather than a single average.

See the method applied to a real site.

The example report applies these six dimensions to a real site and shows how each score turns into an action plan.