Complete Centropic guide

Introduction

Centropic measures and improves your site's visibility in generative models. AIO (AI-Driven Visibility) and GEO (Generative Engine Optimization) are not GIS nor “all-in-one”: they are AI citation metrics. The workspace has five pages: Overview, Benchmark, Prompt, Trend, Guide.

Centropic dashboard overview — Workspace: CVI on the mark, AIO/GEO charts, and packs across five pages.

The five pages

The workspace has a single top bar. Five destinations, same active domain. Account and settings stay in the avatar menu.

Overview

Domain and CVI at the top. Below: seven KPIs, Composition and Engines at the same height, then Suite and Mix, then AIO Crawl with AIO distribution and Criticality index in the same column, pages ledger on the right. At the bottom: run history and findings.

Benchmark

Compare the active site with rivals on AIO, GEO, SoV, and engines. Uses a measured snapshot, not an invented chart.

Prompt

The queries used to measure SoV and the open findings. See what was asked of the engines and the pack to apply.

Trend

Run history on the active domain. Two side-by-side charts: AIO/GEO over time and CVI per run, with dates under each point. Needs at least two analyses.

Guide

This manual, inside the workspace. Same content as the public /guida page.

Services

Each product block, with illustration and what it does.

Workspace

Five pages, one active domain, charts from the same run.

Overview: CVI on the mark (DD–AA grade inside the C, score on the right), KPIs and charts
Benchmark, Prompt, and Trend on the domain selected from the chips
Account and settings in the avatar menu — not a sixth pill

AIO / GEO analysis

HTTP probe + HTML parsing on the domain: measured what is observable.

AIO = AI-Driven Visibility (how much models “see” the brand)
GEO = Generative Engine Optimization (citability in generated responses)
CVI (Centropic Visibility Index): DD→AA letter grade on the AIO+GEO composite

Findings and critical pages

Prioritized gaps: critical, warn, ok — with concrete actions.

Close criticals first (e.g., AI bots blocked in robots)
Criticality index: (critical × 100 + warning × 40) / total, under AIO distribution in the Crawl stack
Critical pages highlight URLs with severe gaps

Optimization pack

HTML pack with the original logo, score, findings, and snippets to publish.

Download centropic-fix.html: opens offline, original logo at the top
On Plus you can send the pack via email
Plus/Business: pack.ready webhook (checksums + Edge URLs) so your CMS can apply the pack without copy-paste

Edge Signals

Dynamic hosting of artifacts on centropic.ai/e/<token>/…

Free: llms.txt + signals.json
Plus: live robots, organization.jsonld, Worker/Vercel/embed
Updates on every re-scan and when the AI crawler list changes

Universal CMS Connector

A ZIP for WordPress, Drupal, Shopify, PHP, Netlify, Cloudflare, Vercel.

Activate Edge, then download the connector from the dashboard
One adapter on your host: proxy to Edge
APPLY.md in the ZIP: pack.ready event, idempotent on signals_version

Share of Voice

Proxy (estimated) on every plan; measured on Plus/Business with the citation monitor.

Estimated: heuristics on site signals
Measured: prompt bank → brand mentions in LLM responses
Gemini and Azure AI remain honest proxies (not native Overview/Copilot)

GEO Suite

Modules: entity graph, citability, schema, publish verify, llms lint, locales.

Entity: Organization, sameAs, contacts
Publish verify: check if the artifacts are live
Locales: hreflang and consistency

Competitor snapshot

Quick comparison AIO/GEO/rating on competitors (Plus).

Same metrics as your site
Useful for relative priorities, not absolute ranking

Trend and re-scan

Time series on the active domain: AIO/GEO and CVI, with dates under each point.

Two side-by-side charts: AIO/GEO and CVI per run
Set frequency and UTC time in Settings
before-after.md confronta due analisi successive

Operational allowance and coverage

Every plan includes a monthly operational allowance. Extra packs increase coverage without changing plans.

Choose the plan based on domains, re-scan frequency, and API/white-label needs
Extra packs in euros for re-analysis spikes or clients
The monthly renewal allowance is included in the Plus/Business subscription

API and white-label

Agency automation: API key, analyze, sites, edge; white-label report.

Bearer ct_… on /api/v1/* (legacy gp_… accepted)
GET /sites/<id> and /runs: AIO, GEO, CVI, SoV, criticality, checksums — OpenAPI at /api/v1/openapi.json
analysis.completed and pack.ready webhooks (plus regression alerts)

Analysis and modules

What a run produces: measured probes, GEO Suite modules, and breakdown by engine.

Crawl and probe

Centropic requires title, meta, JSON-LD, FAQ, robots, llms.txt, sitemap, ai.txt, humans.txt, and site pages. Free: page cap; Plus: full crawl (with platform operational cap).

Engine breakdown

View for ChatGPT, Gemini (proxy AI Overview), Claude, Perplexity, Grok, and Azure AI. By default, it is Estimated; with SoV measured, the available engines switch to Measured.

Entity graph

Evaluate consistency of Organization / brand, sameAs, contacts, and reusable entity signals from crawlers and models.

Citability

How citable the copy is: clear claims, definitions, verifiable facts, less ambiguity about 'who you are / what you do'.

Schema quality

Quality and completeness of JSON-LD (Organization, WebSite, FAQPage, SoftwareApplication, Article where relevant).

Publish verify

Check if llms.txt, robots, and schema published on the domain match the drafts of the pack / Edge.

llms lint

Check the presence and quality of /llms.txt: brand, topic, absolute URLs, preferred citation.

Locales

hreflang signals / markets: linguistic consistency and international coverage for multi-language answer engines.

Citation monitor / measured SoV

Only Plus: citation monitor with prompt bank. Counts brand/domain mentions in responses — sample probe, not a guarantee of ranking in consumer UIs.

After the analysis

Recommended order to close the gaps and publish.

CVI (Centropic Visibility Index) Read the letter and score: DD is critical, AA is top. The grade sits in the mark, the number on the right. Aim to climb at least one CVI grade between re-scans.
Score AIO / GEO CVI components. Below 60 = priority intervention. Then sample citation share — do not isolate the numbers.
Findings Close critical and warn items; use the criticality index and the pack. Then republish and verify.
Publish llms.txt Site root: https://yourdomain.com/llms.txt — or Edge + CMS connector.
JSON-LD + FAQ + meta Paste in the <head> of key pages (or let Edge serve organization.jsonld).
robots.txt Merge the draft with the existing file; keep useful AI bots open (GPTBot, ClaudeBot, …).
Re-scan On Plus, set frequency/time UTC and compare before/after.md.

Pack and files

Where the files go

centropic-fix.html → single deliverable: original logo at the top; open and copy head / llms.txt / robots onto the site
signals.json → Edge /geopulse/signals.json via CMS (optional)

Edge Signals and CMS

In dashboard: Activate dynamic hosting → Download CMS connector (.zip) → install a single adapter. Deep dive in FAQ Edge Signals.

Plans: domains, re-scans, API

Free

1 monitored domain, limited crawl, initial analyses included
Score, findings, single HTML pack, basic Edge (llms + signals)
Estimated SoV (proxy)

Plus · €19.99 excl. tax

Up to 5 sites, crawl up to 120 pages (Deep 500), competitors, extended history
Measured SoV, prompt bank, re-scan, alerts, full Edge
Are API and white-label included in Plus?

Business · €89.99 excl. tax

Everything in Plus + up to 50 sites and 400 tokens/month
API /api/v1 and white-label MD/HTML with agency brand
Recommended profile for client portfolios

See updated prices · Methodology

What we measure — and what we don't

Three questions that come up often in sales conversations aren't in the product. That's not a hole in the roadmap: we can't measure them honestly, or they'd burn your monthly quota without making the site more citable.

Attribution and ROI from LLM citations

ChatGPT, Claude, Perplexity, and Copilot do not expose impressions, clicks, or CTR. A “revenue from AI visibility” number would be a made-up correlation. Centropic measures brand mentions against your prompt bank (Measured SoV) and, on Plus, can connect Search Console as a Search baseline — not as AI traffic. When we have sessions with AI referrers reported by your analytics, we'll put them on Trend and label them Estimated.

Agent simulations (buyer personas and purchase funnels)

A multi-turn journey (CTO, procurement) multiplies LLM probes: a full simulation can cost as much as 4–10 Measured analyses and is still stochastic. The Plus prompt bank already accepts buyer-style questions. We do not launch agents on POST /analyze — that would empty the monthly quota without publishing an artifact. If you need a journey report, that is a separate SKU, not the default.

Pre-training data audits (Common Crawl, Reddit, GPT “weights”)

No closed model will certify that your brand is in its weights. Putting that on a dashboard would be a marketing claim, not a measurement. The GEO Suite already reads your site (entity graph, sameAs, schema). A Wikipedia/Wikidata presence is a useful finding; “you are in ChatGPT’s training data” is not. We stay with what we can observe.

Read more about Measured vs Estimated in the Methodology.

Glossary

Voices used in dashboard, reports, and Edge.

CVI (Centropic Visibility Index): Centropic’s proprietary metric: a 0–100 index with a DD→AA letter that synthesizes AIO+GEO (with findings penalties). On Overview the grade sits in the mark; the numeric score sits to the right. It is the standard to compare brands — not third-party Domain Authority.
AIO: AI-Driven Visibility: how understandable/visible the brand and site are to AI models and crawlers. Does not mean All-in-One.
GEO: Generative Engine Optimization: optimization to be cited in responses generated by answer engines. Does not mean GIS.
Citation share (SoV): Sample of brand mentions in generative responses (stochastic), not advertising Share of Voice. Proxy = estimated from structure; Measured = citation monitor (Plus).
Measured: Signal observed directly (HTTP/HTML probe) or mention counted via configured LLM API.
Estimated (proxy): Heuristic value derived from site signals, not a live citation in the LLM UI.
Scale from DD to AA: Centropic Visibility Index letter grade, from critical (DD) to excellent (AA). Historical alias: “Index.”
Criticality index: 0–100 score from the run’s finding mix: (critical × 100 + warning × 40) / total. On Overview it sits under AIO distribution, same Crawl column.
Workspace: The product’s five pages: Overview, Benchmark, Prompt, Trend, Guide. Account and settings stay in the avatar menu.
Finding: Single gap or confirmation (critical / warn / ok) with context and priority.
Pack: Set of artifacts from the analysis. centropic-fix.html includes the original logo at the top, score, findings, and snippets (llms, schema, meta, robots, checklist).
llms.txt: Machine-readable root file that explains brand, topics, and preferred URLs to AI models/crawlers.
JSON-LD: Schema.org structured data embedded in the page (Organization, FAQPage, …).
Edge Signals: Centropic endpoint /e/<token>/… that serve dynamically updated artifacts.
CMS Connector: Universal package (plugin/rewrite) that points your domain's AIO paths to Edge.
signals.json: Machine payload with versioning and Edge metadata for discovery and integrations.
Citation monitor: Plus job that sends prompt bank to LLMs and measures brand mentions.
Prompt bank: Prompt list used for measured SoV; customizable in Settings.
Vertical pack: Set of prompts/checklists for industry (e.g., SaaS, local, ecommerce).
Publish verify: Check that the artifacts published on the domain are reachable and consistent.
Entity graph: Entity signal network (name, sameAs, contacts) that disambiguates the brand.
Citability: How easily the content can be cited by a model (clarity, facts, structure).
Schema quality: Completeness and correctness of structured data on the site.
AI Crawler: Bots like GPTBot, ClaudeBot, PerplexityBot, Google-Extended that index for models.
robots.txt: Crawler access policy; Centropic draft opens useful AI bots without breaking existing rules.
Operating quota: Credit included in the plan (and in extra packs) that covers analyses and measured jobs. In the dashboard it appears as remaining coverage in euros — there is no need to calculate API “tokens.”
Hold: Temporary credit reservation before a job; released or converted to debt at the end of the run.
Re-scan: New analysis of the same site, manual or scheduled (Plus).
White-label: Exportable report with agency brand (logo/colors) for end clients.
API key (ct_…): Bearer ct_ key for /api/v1 (analyze, sites, edge). Legacy gp_ keys are still accepted. Business only.
Engine breakdown: View for engine (ChatGPT, Gemini*, Claude, Perplexity, Grok, Azure*) of Share of Voice.
Critical pages: Crawl URL with severe gaps that drag down score or citability.
before-after.md: Pack document that compares two successive runs after fixes.