Open framework · v1.0
The AI Visibility Framework
Your customers stopped scrolling links. They ask machines - and machines don't rank ten options, they recommend one. This framework defines what decides whether that recommendation is you.
Version 1.0 · published 23 July 2026 · maintained by Instly Technologies · changes are versioned on this page
The eight dimensions
What machines actually evaluate.
Every recommendation an answer engine makes is the output of the same underlying questions. The framework names them as eight measurable dimensions. Each carries a signature failure - a pattern we see over and over in live audits.
AI Visibility
Can machines resolve who you are and extract what you claim? Entity clarity, llms.txt, AI-crawler access, answer-ready content structure.
Signature failure: a business whose own structured data pointed at its hosting provider's staging server - identity broken at the root.
Buyer Intent
Are the questions buyers actually ask - cost, comparison, eligibility, process - answered on the site in extractable form?
Signature failure: an elite surgeon with no pricing or recovery content; engines cite competitors who publish both.
Trust & Credibility
Is proof machine-verifiable? Named reviews with schema, dated case studies, credentials and third-party references on-page.
Signature failure: real Jimmy-Fallon-level credits presented as unstructured quotes machines can't validate.
SEO & AEO
Does page structure serve both ranked search and answer extraction? Heading hierarchy, descriptive links, FAQ markup, semantic clarity.
Signature failure: an FAQ page with four H1s and no FAQPage schema - the site's best answers, unreadable.
Technology Stack
Does the technical layer help or hide content? Schema correctness, rendering strategy, canonical integrity, crawlable conversion paths.
Signature failure: every call-to-action a JavaScript-only button - no path an agent could follow to convert.
Security & Privacy
Do browser-visible signals undermine confidence? Exposed keys, mixed identity, missing policy disclosures.
Signature failure: API secrets readable in frontend code - an instant trust ceiling for any cautious engine.
Performance
Is content delivered fast enough for cold-visiting crawlers and agents? First paint, LCP, stability under automated load.
Signature failure: an 858 KB icon font - for seven icons - blocking every first paint by seconds.
Competitive Position
How does all of it compare against the named competitors an engine would otherwise recommend - measured on identical scans?
Signature failure: whole categories where nobody serves llms.txt - and the first mover takes the answer box.
The evidence hierarchy
What machines trust, in order.
Two sites can make identical claims and be treated completely differently. What separates them is the form of the evidence. The framework ranks it in five levels - audits score how high your proof reaches.
Verified third-party proof
Named reviews with Review schema, store listings, press with links, registry entries. Machines can cross-reference it - the gold standard.
Named, dated first-party evidence
Case studies with real names, dates and numbers a reader could verify. "68 → 82 on 23 July 2026" beats "our customers love us".
Structured claims
Facts in schema, tables and definition lists: prices, dimensions, credentials. Extractable, quotable, and attributable to your entity.
Prose claims
True statements buried in paragraphs. Machines may find them - at reduced confidence, and without the structure that earns citation.
Slogans
"Line producer gold." "Best in class." Zero extractable content. Fine for humans, invisible as evidence.
Measurement
One score, honestly banded.
An assessment scans a site's highest-intent pages and scores the evidence available to a machine reader across all eight dimensions, weighted into a 0-100 AI Visibility Score. Scores are designed to be re-measured after implementation - movement, not a snapshot, is the point.
Measure yourself against it.
The free scan scores the surface in seconds. The full audit assesses all eight dimensions against this framework - with every fix ranked by impact and effort. Or read the case study of the framework applied to our own site.