# How the AI Presence Readiness score works Markdown alternate of . Five pillars. 38 technical checks. One number from 0 to 100. The public score is technical-only — the same site gets the same score, every time. ## The five pillars | Pillar | Weight | The question it answers | |---|---|---| | Machine-readable identity | 30% | Can AI identify your destination? | | Content for AI | 20% | Can AI quote you accurately? | | Crawlability & access | 20% | Can AI access your site? | | Action readiness | 15% | Can AI book or contact you? | | Authority & trust signals | 15% | Can AI trust and recommend you? | ## Scoring Each check produces a numeric score from 0 to 100 and a tri-state outcome (pass / partial / fail). Checks roll up into their pillar by weight, and pillars roll up into the published score. There is no AI judgement anywhere in the loop: fixed weights, deterministic checks, reproducible results. Score bands: Leading (70+) · AI-ready (45+) · Developing (25+) · Emerging (10+) · Invisible (below 10). ## Featured checks **Content freshness** — whether your pages look up to date to AI assistants. Agentic search fan-out queries are dominated by freshness terms, so structured, recent dates help agents trust that seasonal advice, pricing and availability are still valid. Pass: structured dates updated within 12 months. **Structured comparison data** — whether key info (rooms, prices, hours) is in real tables AI can read. Citation research finds answers are far more likely to cite pages with actual HTML tables. Pass: decision-critical data in semantic HTML tables. **Page token budget** — whether your pages are short enough for AI to read fully. Agents work inside finite context windows and truncate long pages; pages beyond roughly 30K tokens risk losing the exact facts an assistant needs. **Markdown alternative** — whether AI assistants can fetch a clean, fast version of your pages. A `.md` route mirroring the canonical page removes navigation chrome and cuts token overhead. Pass: important pages expose clean `.md` versions. **Rich card readiness** — whether AI can show you as a rich card with image and rating instead of plain text. Missing one critical field, often `image` or `aggregateRating`, drops the entity to a plain text mention. **Schema completeness** — whether your schema fills in the properties AI assistants actually use to identify and bucket you. Schema does not feed the AI directly; it feeds what the AI calls (Google Places, Maps, the Knowledge Graph). Properties are graded in three impact tiers and only count in the typed form — a primitive `starRating` of `4` fails the filter that a typed `Rating` object passes. **Entity linking / address structure** — whether AI can place a hotel on a map or link a destination to its identity in the knowledge graph. Entity-aware: for destinations it grades Wikidata QID resolution, Wikipedia `sameAs`, schema `geo` and a locality signal; for hotels it splits into address structure, entity reconciliation and geo coordinates. **Brand hierarchy** — whether AI can tell your property apart from its parent chain or group, so recommendations and booking links are not attributed to the wrong property. **Review signals** — whether AI can see enough recent reviews to trust recommending you. Graded on depth and freshness, not on the mere presence of a rating snippet. **llms.txt quality** — whether your AI guide file is detailed enough to actually help: task-organized routes, descriptions, token expectations. Graded on quality, not existence. Google does not use llms.txt for Search; we grade it because other engines and AI crawlers may, and weight it as a supplementary signal. **Agent spec coverage** — whether machine-readable manifests describe the entities, tasks and canonical routes agents can safely use. **Agent action path** — whether AI can book a room, reserve a table or contact you on a guest's behalf: a programmatic endpoint on every entity, plus structured `ReserveAction` and a cancellation policy URL on lodging. **AI retrieval crawler access** — whether AI assistants are allowed to read your site. This is the bot set that matters for live citation: OAI-SearchBot, ChatGPT-User, Claude-User, Googlebot, PerplexityBot. We test robots.txt policy *and* a simulated fetch. Pass: ≥80% allowed on both axes. **AI training crawler access** — whether AI vendors are allowed to learn about your brand. Both yes and no are valid choices: opting out of training is legitimate, and retrieval-only bots increasingly respect the distinction, so blocking training no longer breaks live citations. We test GPTBot, ClaudeBot, Google-Extended and PerplexityBot. **Agent interaction quality** — whether your booking forms work when an AI fills them out for a guest. A composite of accessibility tree (40), form labels (25), interactive semantics (20) and layout stability (15). A booking widget invisible to the a11y tree is invisible to the agents that book on behalf of guests. ## Critical issues Four conditions are called out at the top of a report as critical issues. They degrade readiness even when the rest of the audit looks fine: - Missing schema.org markup - Missing location data - No agent-readable action path (book, reserve, contact) - No review signals Each is a specific, fixable gap: fix it and the underlying checks recover, so the score moves. No cliffs. Which of the four can fire depends on the entity type — a destination is trivially locatable by name, so "missing location data" never fires on a DMO report. ## One exception — when crawlers are blocked If AI crawlers cannot reach your site at all, the rest of the score is academic. For this one case the public score is held at a cap until access is restored, and the report says so explicitly. Your "raw potential" — what the rest of the audit measured — stays visible so you know what unlocks the moment you allow AI crawlers. This is the only case where the displayed score does not equal the weighted aggregate of the checks. ## FAQ **Why does this matter?** AI agents are taking over discovery. If they cannot find, parse and act on your content, you are invisible no matter how strong your traditional SEO is. **Is the score subjective?** No. It is 100% objective — computed from technical signals (schema.org markup, llms.txt, ai.json, sitemap, robots, action endpoints, content readability) using fixed weights. Running the same audit twice yields the same score. **How is this different from SEO?** Traditional SEO measures whether Google can index a page for human readers. AI Presence Readiness measures whether AI agents can find structured data, parse machine-readable summaries, and trigger actions like bookings or reservations on a guest's behalf.