Inside the Algorithm: How We Score AI Readiness
Full transparency on how AgentReady calculates your AI readiness score. The 8 factors, their weights, the grade scale from A to F, bonus signals, and how it all connects to real AI citation outcomes.
Founder & CEO at AgentReady
Why We Publish Our Methodology
One of the biggest complaints about Domain Authority, Ahrefs Domain Rating, and similar industry metrics is that nobody outside the company truly knows how they work. The black-box approach breeds distrust, gaming, and endless debate about what the numbers actually mean.
We decided from day one that AgentReady would be different. Our scoring methodology is fully documented, and our scoring algorithm is open source. You can read the code, audit the weights, and even contribute improvements.
Transparency is not just a principle — it is a practical decision. If agencies and developers understand exactly how the score is calculated, they can make smarter optimization decisions. If researchers can audit the methodology, they can validate (or challenge) our approach. If the community can contribute, the algorithm gets better faster than any single team could manage alone.
The 8 Scoring Factors
AgentReady evaluates every website across 8 factors. Each factor is scored independently on a 0-to-100 subscale, then combined using weighted averages to produce the overall score. Here is what each factor measures and why it matters.
- Bot Access & Crawlability (25%) — The foundation. We check robots.txt for AI bot directives, meta robots tags, HTTP response codes for key pages, and whether AI crawlers like GPTBot, ClaudeBot, and PerplexityBot are allowed. If AI agents cannot reach your content, nothing else matters. This carries the heaviest weight.
- Structured Data & Schema (20%) — We validate Schema.org markup across your site, checking for type coverage (Article, Product, FAQ, HowTo, Organization), completeness of required properties, and correct nesting. Rich structured data helps AI systems understand your content without guessing.
- AI Protocols (15%) — We check for the presence and validity of llms.txt, NLWeb endpoints, and MCP server declarations. These emerging protocols are the direct communication channel between your site and AI agents. Early adoption is a strong signal of AI readiness.
- Content Quality & Structure (15%) — Heading hierarchy (H1 through H4), paragraph density, content depth (word count per topic), use of lists and tables, presence of cited sources, and readability scores. Well-structured content is easier for AI to parse and quote accurately.
- Topic Authority (10%) — Author attribution, author pages, credential signals, topical depth across the site, and E-E-A-T indicators. AI systems increasingly evaluate source credibility before citing content.
- Speed & Performance (5%) — Server response time (TTFB), page weight, Core Web Vitals where available. AI crawlers have timeouts and resource limits just like human users.
- Internal Linking (5%) — Link density, topical clustering, orphan pages, and navigation depth. A strong internal link structure helps AI systems understand content relationships and topic hierarchies.
- Crawl Health (5%) — Broken links, redirect chains, 404 error rates, XML sitemap validity, and canonical tag consistency. Clean crawl paths ensure AI agents can traverse your site efficiently.
Factor Weight Distribution
The Grade Scale: A Through F
The raw score from 0 to 100 maps to a letter grade that makes the result immediately understandable. The thresholds were calibrated against our study of 5,000 websites and correlated with actual AI citation rates.
We deliberately set the thresholds so that a C grade represents the average website today. This means most sites will initially see a C or D, which is honest — the web is not AI-ready yet. The grade is meant to be a motivator, not a vanity metric.
- A (85-100) — Excellent. Fully optimized for AI discovery. Comprehensive schema, active protocol adoption, strong authority. These sites are cited 2-3x more often than average.
- B (70-84) — Good. Strong foundation with minor gaps. Typically missing one or two protocol files or has incomplete schema coverage. Well-positioned to reach A with targeted improvements.
- C (55-69) — Average. Partially visible to AI but has meaningful gaps. Usually strong on traditional SEO signals but weak on AI-specific factors. Most established websites land here.
- D (40-54) — Below average. Significant gaps across multiple factors. Often blocking AI crawlers unintentionally or lacking any structured data. Immediate action recommended.
- F (0-39) — Critical. Effectively invisible to AI agents. Major issues with bot access, no schema markup, no protocol adoption. These sites need a comprehensive AI readiness overhaul.
Bonus Signals and Score Adjustments
Beyond the 8 core factors, AgentReady applies bonus adjustments for signals that indicate above-average AI readiness investment. These can add up to 5 additional points to the raw score.
Protocol bonus (+2 points): Sites that implement all three major AI protocols (llms.txt, NLWeb, and MCP) receive a bonus because the combination signals comprehensive AI strategy, not just checkbox compliance.
Original research bonus (+1 point): Pages that contain original data, studies, or unique datasets receive a content quality uplift. AI systems strongly prefer citing primary sources over content that merely aggregates existing information.
Multi-language bonus (+1 point): Sites that serve content in multiple languages with proper hreflang tags and translated schema markup demonstrate global AI readiness.
Freshness bonus (+1 point): Sites that have updated their AI-related configurations (schema, protocols, content) within the last 30 days receive a recency signal. Active maintenance correlates strongly with higher citation rates.
How We Validate the Score
A scoring algorithm is only useful if it correlates with real outcomes. We validate AgentReady scores against two primary benchmarks.
Citation analysis: We regularly sample AI-generated responses from ChatGPT, Perplexity, and Google AI Overviews across thousands of queries. We then check whether cited sources have higher AgentReady scores than non-cited alternatives. The correlation is strong and consistent — sites scoring above 70 are cited at roughly 2x the rate of sites scoring below 50.
Expert review: We work with SEO professionals and AI researchers to audit the scoring methodology. When experts disagree with a factor's weight or a signal's inclusion, we investigate, test, and adjust. The algorithm has been through three major revisions since our initial release.
The scoring methodology is documented in full in our scoring documentation, and the source code is available in our open-source repository.
Contributing to the Algorithm
Because the scoring algorithm is open source, anyone can contribute. We accept pull requests for new signal detectors, weight adjustment proposals (with supporting data), bug fixes, and documentation improvements.
If you believe a factor is over- or under-weighted, open an issue with your evidence. If you have found a signal that correlates with AI citation outcomes and is not currently measured, propose it. The best scoring algorithm is one that reflects the collective knowledge of the community, not just the opinions of one team.
Read our contribution guidelines to get started.
Frequently Asked Questions
How often are the scoring weights updated?
We review and potentially adjust weights quarterly, based on new citation correlation data and community feedback. Major weight changes are announced on the blog and in our changelog. Minor calibrations happen more frequently.
Can the score be gamed?
Because the methodology is transparent, someone could theoretically optimize purely for the score. But because the score is validated against real AI citation outcomes, gaming it would require actually improving your AI readiness — which is exactly the point.
Why does Bot Access carry the highest weight at 25%?
If AI crawlers cannot access your content, no other optimization matters. A site with perfect schema, excellent content, and full protocol adoption scores zero on AI visibility if it blocks GPTBot and ClaudeBot. Bot access is the prerequisite for everything else.
Check Your AI Readiness Score
Free scan. No signup required. See how AI engines like ChatGPT, Perplexity, and Google AI view your website.
Scan Your Site FreeRelated Articles
Introducing AgentReady: The AI Readiness Score for Every Website
We built AgentReady because the web is entering its AI era and nobody had a way to measure whether their website was ready. Today we are launching a free tool that scans any site and scores it across 8 factors that determine AI visibility.
ProductOur Algorithm Is Open Source — Here's Why
We made the AgentReady scoring algorithm open source. Not because we had to — because we believe the AI readiness standard should be built by the community, not controlled by one company. Here is why, and how to contribute.
Data & ResearchWe Scanned 5,000 Websites for AI Readiness. The Results Are Alarming.
73% of websites are invisible to AI. We scanned 5,000 sites across 14 industries and the data reveals a massive readiness gap that most businesses don't even know exists.