Methodology · v1.0.0

AIRS — AI-Readiness Score

One number, six axes, fully open. The same engine that powers every Synelo audit — published here so any auditor can reproduce a score by hand.

Formula version 1.0.0·Updated when weights or checks change

The composite

AIRS is a weighted average of six axis-scores (each 0–100). Heaviest weights go to AI Readiness and Compliance Risk — the two categories nobody else measures, and the ones EU SMBs face hardest deadlines on in 2025–2027.

AIRS = round(
    Performance       × 0.20  +
    Accessibility     × 0.15  +
    SEO Visibility    × 0.10  +
    AI Readiness      × 0.25  +
    Compliance Risk   × 0.20  +
    Security Hygiene  × 0.10
)

Each axis score is itself a weighted average of its checks. For PSI-anchored axes (Performance, Accessibility, SEO), the PSI category score is blended 70/30 with the check pass-rate so the AIRS axis stays correlated with Lighthouse but adds the EU-specific layer.

Score labels

Range	Label
`0–39`	Critical exposure
`40–59`	Below median
`60–74`	Industry standard
`75–89`	Strong
`90–100`	Reference-grade

Performance · weight 20%

PageSpeed Insights mobile run + CWV thresholds. Real users feel speed first; everything else is downstream.

lcp-good
LCP < 2.5s
Optimise hero image, defer non-critical JS, enable a CDN.
cls-good
CLS < 0.1
Set explicit width/height on images and embeds; avoid late-injected ads above the fold.
inp-good
INP < 200ms
Reduce main-thread JavaScript; split long tasks; defer hydration.

Accessibility · weight 15%

PSI accessibility category + WCAG 2.2 AA / EAA-specific checks (text alternatives, contrast, focus visibility, target size).

image-alt
Images have alt text
Descriptive alt text on every image; alt="" only for decorative.
color-contrast
Colour contrast ≥ 4.5:1
Raise text/background contrast to WCAG AA threshold.
button-name
Buttons have accessible names
aria-label or visible text on every interactive control.
link-name
Links have discernible text
Replace "click here" with descriptive link copy.
lang-attr
HTML lang attribute set
<html lang="…"> declares page language for assistive tech.

SEO Visibility · weight 10%

PSI SEO category + presence of structured data and social-graph metadata. AI search engines and traditional crawlers consume the same signals.

json-ld
JSON-LD structured data
Add Organization + WebSite + WebPage JSON-LD on the homepage.
og-meta
Open Graph metadata
og:title, og:description, og:image, og:url.
twitter-card
Twitter card metadata
twitter:card, twitter:title, twitter:image.
canonical
Canonical URL declared
<link rel="canonical" href="…"> to prevent duplicate-content penalties.

AI Readiness · weight 25%

Whether AI bots (GPTBot, ClaudeBot, PerplexityBot) and AI search engines can read, summarise, and recommend this site. The single biggest growth lever for 2025-2027 — and the differentiator AIRS is built around.

robots-allow-ai
AI bots allowed in robots.txt
Remove Disallow: / for GPTBot / ClaudeBot / PerplexityBot / Google-Extended. AI engines that can't crawl you can't cite you.
llms-txt
/llms.txt manifest published
Publish /llms.txt summarising your site's purpose, key URLs, and content hierarchy for LLM context. See anthropic.com/llms-txt.
json-ld
JSON-LD schema (AI-readable)
AI summarisers consume JSON-LD as primary signal — add Organization, Service, Product, FAQPage where relevant.
og-meta
Open Graph (AI summarisers)
AI search results lean on og:title / description heavily — these are your "snippet" in the AI era.
semantic-html
Semantic HTML structure
Replace <div> soup with <header>, <nav>, <main>, <article>, <section>. AI crawlers extract structure, not styles.
sitemap
Sitemap discoverable
Reference sitemap in robots.txt: "Sitemap: https://…/sitemap.xml".

Compliance Risk · weight 20%

EU regulatory exposure: AI Act (Art. 50 transparency), EAA (28 Jun 2025 effective), NIS2 (transposed Oct 2024), GDPR (cookie + privacy).

ai-act-transparency
AI Act Art. 50 transparency (if AI features used)
Add visible "Powered by AI" / "AI-generated" notices near every AI-driven surface.
eaa-accessibility
EAA-grade accessibility (WCAG 2.2 AA)
Address all WCAG AA failures listed in the Accessibility axis.
nis2-security-txt
NIS2 — incident contact published
Publish /.well-known/security.txt with contact + policy URL (RFC 9116).
gdpr-cookie
GDPR / ePrivacy cookie consent
Compliant consent layer (Cookiebot, Iubenda, Usercentrics).
privacy-policy
Privacy policy linked
Footer link to /privacy or /datenschutz.
imprint
Imprint / legal notice (TMG §5)
/imprint or /impressum is mandatory in DE/AT for commercial sites.

Security Hygiene · weight 10%

Response-header analysis: HSTS, CSP, X-Frame-Options, X-Content-Type-Options, Referrer-Policy, Permissions-Policy. Cheap to ship; signals operational maturity.

hsts
HSTS header
Strict-Transport-Security: max-age=31536000; includeSubDomains.
csp
Content-Security-Policy
Whitelist trusted script/style origins; prevents XSS escalations.
x-frame-options
X-Frame-Options (clickjacking)
X-Frame-Options: DENY (or use CSP frame-ancestors).
x-content-type-options
X-Content-Type-Options
X-Content-Type-Options: nosniff.
referrer-policy
Referrer-Policy
Referrer-Policy: strict-origin-when-cross-origin.
permissions-policy
Permissions-Policy
Lock down camera, mic, geolocation by default.

What AIRS does NOT measure

Conversion rate. Even a score of 100 won't save bad copy or weak product–market fit. AIRS measures technical foundation; conversion is a separate audit.
Brand quality. A reference-grade score on a Comic-Sans landing page is still a reference-grade score. Visual design is intentionally out of scope.
Backend reliability. AIRS is an external audit — uptime, database integrity, billing logic are not observable from the homepage.
Industry-specific compliance beyond EU AI Act / EAA / NIS2 / GDPR. HIPAA, PCI-DSS, SOC 2 etc. are domain-specific and require dedicated audits.

Disputing a score

If you believe a check returned a false negative (e.g. you have a CSP but our parser missed it), email audits@synelostudio.com with the URL and the failing check ID. We rerun the audit and update the spec if our detection logic was wrong.