Frequently asked questions

Clear answers before opening a dossier.

Review the central concepts, their boundaries and the next verifiable source for more detail.

Frequently asked questions

Clear answers before opening a dossier.

Review the central concepts, their boundaries and the next verifiable source for more detail.

What does it mean for a website to be agent friendly?

An agent-friendly website combines clear content, coherent public routes, structured data, sources, dates and boundaries. Tools are declared only when they exist, while sensitive actions retain identity, permission, audit and human control.

Is progression from AF-0 to AF-5 automatic?

Each level requires verifiable work: publish evidence, implement the capability, test it, declare boundaries, obtain human approval where needed and measure again. Not every website needs to reach AF-5.

What does the public audit inspect?

The audit observes discovery, citable content, machine-readable resources, tools, experimental signals, trust and commerce. The result is a proprietary measurement, not certification or a ranking or recommendation guarantee.

What are robots.txt, sitemap.xml and agent documents for?

robots.txt guides crawling, sitemap.xml lists canonical pages, and documents such as llms.txt or catalogs describe knowledge and capabilities. No single file guarantees that a provider will index, cite or recommend the website.

What is llms.txt and how is llms-full.txt different?

llms.txt works as a concise index while llms-full.txt provides expanded public context. They are not universal standards and do not replace HTML, sitemaps, structured data, APIs or editorial review.

What is AEO and how does it relate to SEO?

SEO and AEO complement each other. SEO helps pages get found; AEO improves answer clarity and citability. Neither can promise a particular position, citation or recommendation.

Can an owner define a different policy for each AI crawler?

GPTBot, OAI-SearchBot, ClaudeBot, Google-Extended and other identifiers have different functions and controls. Search, user-requested retrieval and training should be evaluated separately against current official sources.

What does the public Registry establish?

The Registry separates owner declarations, observations and domain evidence. Verifying a challenge establishes specific technical control; it grants no hosting access, legal ownership, write authorization or general certification.

Do I need to share passwords to get started?

When implementation needs access, use temporary, least-privilege and revocable permissions or a reviewable capsule. Secrets must never be sent through chat, FAQ, Registry, OKF or public documentation.

Which public tools are available today?

Current tools query already-public evidence, use versioned contracts, keep no memory and change no websites. A2A, owner actions, payments and other capabilities retain separate gates and are not presented as active.

How much does improving a website cost?

There is no invented one-size price because a website may need only documents or also content, structured data, APIs, provider coordination and external validation. The dossier gathers context for a reviewable proposal before any payment.

How can an owner verify that the work improved the website?

Agent Friendly Web records URL, date, provider, signals and limitations. Different auditors measure different things, so their results complement rather than replace direct tests, transparent methodology and human review.

Why is Tokenizart the first integral case?

Tokenizart explains identity, provenance and its owner-first philosophy; Atelier is the operational platform. The case publishes only verifiable capabilities and keeps Owner Live, Mint, Certify, transfers, vouchers and payments behind identity, consent and separate gates.