Methodology·Public document
The Caliber AI-Readiness Standard.
v1.0 · May 2026 · a living document, revised as AI search evolves.
What this is — and what it isn’t.
We’re publishing this because most discussion of AI search is either vendor hype or guesswork. This is our reasoning, in writing, where anyone can argue with it.
What it is
Our published, transparent, versioned methodology for making websites readable and citable by AI search engines — the discipline the industry is starting to call generative engine optimization (GEO).
Six dimensions, each with a plain-English explanation of what we check and why it matters. When the landscape changes — new crawler, new schema, new ranking signal — the version increments and the change is recorded.
What it isn’t
Not an industry standard, an official standard, or a certification. We didn’t form a committee. We don’t represent anyone but ourselves.
Not a guarantee of placement. We engineer and verify readiness; we never promise an AI citation, exactly as we never promise a Google ranking. Anyone who promises either is selling you something.
How we score AI-readiness.
Each dimension has a clear test and a plain-English reason. Pass-fail is documented in writing. Nothing is graded on vibes.
-
Crawler Access
What we check. The
robots.txtexplicitly allows the retrieval bots that power AI search answers:OAI-SearchBot,ChatGPT-User,Claude-SearchBot,PerplexityBot,Google-Extended, andBingbot. Training-bot policy (GPTBot,ClaudeBot,CCBot) is a separate, deliberate, documented choice — we don’t default it; we discuss it with you.Why it matters Accidentally blocking retrieval crawlers is the single most common reason businesses are invisible in AI search. A site can be technically excellent and still excluded because a five-line robots.txt slammed the door before any of the work was read. -
Server-Rendered & Parseable
What we check. Core content — headlines, body copy, contact details, structured data — arrives in the initial HTML response, not after a client-side JavaScript bundle finishes booting. The page can be read by something that has never executed a line of JS.
Why it matters Most AI retrieval crawlers don’t execute JavaScript. If your content lives behind a render that only happens in a browser, the AI sees a blank shell. This is the structural edge our builds are designed around — static, served, readable. -
Structured Data
What we check. Valid JSON-LD for the appropriate types —
Organization,LocalBusiness,FAQPage,Article— that matches what is visible on the page. Schema is validated, not just present.Why it matters Schema is how a machine learns what your page is about without inferring. But broken or mismatched schema — claiming five-star reviews you don’t have, or describing a service you don’t offer — is worse than none. Both Google and AI engines penalize it. -
Answer-First Content
What we check. Important questions appear as
<h2>or<h3>headings, with a direct, self-contained answer of roughly forty to sixty words immediately beneath — before any storytelling, history, or marketing wind-up.Why it matters AI engines summarize. They grab the cleanest answer to the user’s actual question and quote it. Pages that bury their answer behind two paragraphs of preamble get skipped; pages that lead with the answer get cited. -
SEO & Performance Foundation
What we check. One populated
<h1>per page. A cleancanonicaltag that actually points at the right URL. No accidentalnoindex. A mobile layout that doesn’t collapse. A page that loads quickly under realistic network conditions.Why it matters AI search isn’t a separate internet. It rides on the same crawl infrastructure that’s indexed the web for two decades. The pages that win in AI answers are, almost always, the pages that were already structurally sound for Google. -
Entity & Trust Signals
What we check. Consistent Name, Address, Phone (NAP) across your website and the public profiles that point at it. Claimed and accurate listings where they matter. Honest third-party mentions — press, partner pages, citations — that align with what your site says about you.
Why it matters AI engines build a model of who you are by triangulating mentions across the web. When the signals contradict each other — wrong phone here, different name there — the engine’s confidence drops, and confidence is what decides who gets cited.
What we don’t claim.
The space around AI search is full of confident sales pitches that can’t be verified. We’d rather state our limits up front.
- No placement guarantees. We can verifiably make a site readable, parseable, and citable. We cannot make ChatGPT, Perplexity, or Gemini choose to cite you on any given query — nobody who’s honest can.
- No unverified vendor hype stats. You will not find “sites with our standard get 4x more AI citations” on this page. We don’t publish numbers we can’t source, and we don’t recycle other vendors’ numbers either.
-
On
llms.txt: the proposed standard is emerging and unconfirmed by any major AI engine as of this version. We deploy it as cheap insurance — it costs nothing and may matter someday — but it isn’t graded in the scorecard, and we don’t charge for it. - No moving goalposts. AI search will change. When the dimensions change, we publish the next version of this document with a dated changelog — not a quiet edit that pretends the old version never existed.