The Open AEO Standard v1.0

Every AEO measurement tool — Profound, Semrush AI Visibility, Otterly, AthenaHQ, Peec, Brand Radar — uses a proprietary, unpublished scoring algorithm. Customers buy the score on faith and cannot independently audit, replicate, or peer-review the numbers.

That's fine for monitoring tools. It's a problem for a certification. A cert that says “your site scores 87 by the AI Search Ready standard” means nothing if “the AI Search Ready standard” is a closed box.

This document is the open box. The full rule set lives in the scanner's src/lib/scanner/data/signal-rules.ts, schema-rules.ts and ai-crawlers.ts — MIT-licensed, version-controlled, and reproducible from published canonical sources where they exist.

Every signal belongs to exactly one pillar. The pillars are independent inputs to the headline score; their weights are documented in the source code and are themselves subject to empirical recalibration via the corpus (see Calibration policy below).

• Content99 signals
• Schema30 signals
• Robots3 signals
• Llms txt3 signals
• Citation2 signals

Of 137 signals, 119 have their score magnitude backed by a cited public source (Google Search Central docs, schema.org spec, Lighthouse/WebVitals documentation, etc.). The remaining 18 are expert judgment — there is no canonical source that says “a missing H1 costs you 6 SEO points,” so we don't pretend there is.

Every rule's impactRationale field is prefixed EVIDENCE: or HEURISTIC:. Future versions of this standard will gradually move HEURISTIC signals to EVIDENCE as the corpus grows enough to regression-fit the weights against measured citation outcomes.

1. Snapshot suite. 15 deterministic fixtures (8 captured real-site HTML + 7 hand-crafted scenarios) run offline on every change. Any weight tweak that shifts a score by ≥1 point fails the snapshot.
2. Adversarial suite. Range/ordering assertions on hand-crafted fixtures: flawless ≥ 85, all-broken ≤ 25, flawless → all-broken gap > 50. Catches the regression where an accidentally-inverted weight passes snapshots.
3. Corpus regression. The corpus (54 labelled sites across 10 sectors as of v1.0) is run nightly via npm run corpus:evaluate. Per-issue Pearson correlation against operator labels is tracked. Rules with consistently positive correlations (firing on the BETTER sites) are softened in subsequent calibration passes.
4. Citation verification. For sites with at least one Citation Probe run, the corpus label can be replaced by the measured citation rate across Perplexity, Claude, and OpenAI — replacing operator judgment with ground truth. (Work-in-progress for v1.1.)

The first AEO-audit Model Context Protocol server. Any MCP-compatible client (Claude Desktop, Cursor, LM Studio, Continue.dev, Zed) can call the AI Search Ready audit directly from a chat. Free preview tier; no key required.

{
  "mcpServers": {
    "aisearchready": {
      "command": "npx",
      "args": ["-y", "@aisearchready/mcp"]
    }
  }
}

Three tools surface: audit_url (137-signal scan), verify_cert (canonical signed cert state for a domain), and get_standard (this document, condensed). Source at github.com/ed2903-web/aisearchready/mcp-server.

A standard whose maintainer doesn't audit themselves isn't a standard. We publish our own live scan and well-known files alongside the published score so anyone can verify we apply the same rules to our own site that we apply to anyone else's.

• aisearchready.io/scan/aisearchready.io — live score on our own production site, regenerated on every scan. We don't cherry-pick the run; this URL always shows the latest.
• aisearchready.io/aisearchready.txt — the canonical spec for the well-known discovery file certified customers publish at the same path on their own domains.
• Why no "ai-search-ready certified" badge in our own footer? We're the issuer — embedding our own badge would be conflict-of-interest theatre. The live audit IS our self-certification.

• The full standard text on this page is dual-licensed: MIT for the methodology + CC-BY-4.0 for the prose.
• github.com/ed2903-web/aisearchready · the rule registry + corpus + scoring engine, version- controlled and reproducible