Glossary and core concepts

A conclusion directly supported by evidence collected from the analysed scope or a cited external source.

Observed findings point to concrete evidence such as a page, a heading, structured data or a technical file. Observed means CompComp saw it; a company's own marketing claim is still attributed to the company, not treated as independently verified.

A reasoned interpretation of evidence, clearly labelled so it is never mistaken for a confirmed fact.

Inferred findings are drawn from patterns in the evidence, for example the likely audience suggested by tone, offer and examples. They are useful but always shown as interpretation.

There is not enough evidence to make a reliable assessment. Unknown is a valid result, not a weakness.

CompComp prefers Unknown to invented completeness. Missing evidence is not proof of absence, so Unknown results are never presented as competitive gaps. Typical causes are blocked pages, timeouts or content outside the selected scope.

How much of an assessment could actually be checked, weighted by the importance of each check.

Numeric scores are only shown when weighted evidence coverage is sufficient (60% by default). Below that, the score is left empty and the assessment is Unknown.

The exact part of a web presence that is analysed: a host, a subdomain, a path and its child pages, or a single page.

A path scope such as shop.com/product is never broadened to the whole domain, and /product does not match /products. When participants are analysed with materially different scopes, CompComp shows a scope difference warning.

Your own company in a comparison — the participant every competitor is compared against.

Each comparison has exactly one baseline (own) participant plus competitors. All participants are assessed with the same method and criteria.

Site evidence comes from the selected scope; external evidence comes from the wider public web. They are kept separate.

Keeping the two source classes apart makes it clear whether a finding reflects what a company publishes itself or what others publish about it.

A summary of strengths, weaknesses, opportunities and threats drawn from evidence already collected.

CompComp's SWOT is evidence-driven rather than written freehand. Opportunities appear only when the evidence supports them.

The kind of customer a website appears to address. In CompComp it is always inferred from public pages.

A public website never proves a company's internal ICP, so CompComp labels audience and ICP findings as inferred.

How clearly the offer, audience and differentiation are stated within the scope, and how consistently.

Positioning is assessed from what the analysed pages actually say, using the same criteria for every participant.

How well a site's product and purchase information can be found, understood and compared by AI shopping agents.

ACO covers product identity, variants, attributes, price and offers, availability, merchant details, delivery, returns, reviews, machine-readable commerce data and purchase-path clarity. Product schema alone is not enough for a Strong result.

An open protocol for how agents discover commerce capabilities. CompComp reads the published discovery profile only.

CompComp fetches the standard /.well-known/ucp discovery profile and records the declared protocol version, services, transports, capabilities and payment handlers. These are published declarations: CompComp does not call endpoints or test carts, checkout or payments, and never infers UCP support from general product data.

A readiness assessment of on-page signals that help AI systems understand a site — not a measured ranking.

AI Visibility signals are assessed from the analysed pages. They do not claim that a brand is cited or recommended by any AI system.

An optional check of whether answer engines mention a brand for a set of questions, reported per provider.

Results are shown per provider, with the questions and answers used, so they can be judged in context. They reflect the moment the check ran and are not a guarantee of future answers.

A plain-text file at a site's root that summarises the site and links key pages for AI systems.

llms.txt is a supplemental convention. It can help AI systems orient themselves, but it is not a ranking mechanism. CompComp may check it at origin level without widening the content scope.

A machine-readable file at /.well-known/agents.json describing how AI agents may use a site.

It typically points to documentation, sitemaps and any available agent actions or APIs.

Readiness to be understood, verified and attributed by AI-powered search and answer systems.

CompComp's GEO readiness checks structure, factual clarity, attribution, answerable content and crawler accessibility. It is a readiness assessment, not a promise of citations.

Indexability, on-page structure, metadata and structured data within the selected scope.

Search readiness looks at what search engines need to discover and understand pages. It does not use keyword-rank or backlink datasets.

A root file that tells crawlers which parts of a site they may fetch.

CompComp may read robots.txt for technical discovery at origin level without treating unrelated pages as in-scope content.

An XML list of a site's indexable URLs, often with last-modified dates.

Sitemaps help crawlers find pages and judge content freshness.

The preferred address of a page, declared so duplicates point search engines to one version.

Consistent canonical URLs help search engines and AI systems attribute content to the right page.

Machine-readable descriptions of a page, its organisation or its products, usually in Schema.org JSON-LD.

Structured data helps machines interpret page content. CompComp records it as evidence, but its presence alone does not prove the described facts are accurate.

Schema.org Product data such as name, SKU, GTIN, price and availability.

Product schema and identifiers are normal ACO evidence. They are never presented as proof of UCP support.

A category or listing page that shows several products.

When a path scope covers a listing, CompComp notes whether it analysed listing pages, product pages or both.

A page for a single product, with details such as variants, price, availability and delivery.

PDPs carry most of the evidence used for ACO readiness.