Every page you publish now has two audiences: people, and the AI systems that summarize the web for them. The AI Readiness Score is a single 0–100 number that tells you how well a page serves the second audience—whether ChatGPT, Claude, Perplexity, and Google AI Overviews can crawl it, parse it, trust it, and reuse it in an answer. This guide opens the black box: exactly which checks run, how the score is calculated, and what a good result looks like. You can reproduce every number yourself with the free SellOnLLM Chrome extension.
Why a single score for AI search?
Traditional SEO tools grade you for a ranked list of blue links. Answer engines don’t work that way. They ingest your page, decide whether it is machine-readable and trustworthy, and then either reuse it in a synthesized answer or ignore it. A page can rank fine on Google and still be invisible to AI because its key facts are locked in images, its structure is ambiguous, or it lacks the signals models use to decide who to cite.
The AI Readiness Score compresses that reality into one metric you can track over time, then splits it into two sub-scores so you know where the problem is:
- Technical Score — can machines access and understand the page at all?
- Content Score — once parsed, is the content clear, answerable, and trustworthy enough to reuse?
If you have read our guide on how to rank on ChatGPT, this score is the measurement layer underneath that whole playbook.
The Technical Score: can machines read the page?
These checks map to the mechanics of crawling and parsing. Fail them and nothing else matters, because the model never gets clean access to your content in the first place.
Crawlability & access
- robots.txt directives — does your robots.txt accidentally block AI crawlers like GPTBot or PerplexityBot? We cover this in depth in the AI crawlers guide.
- Indexability — noindex tags, canonical conflicts, and redirect chains that stop a page being stored.
- Server-rendered content — whether your main copy exists in the raw HTML or only appears after JavaScript runs. Most AI crawlers read the initial HTML, so client-only content is often missed.
- Status & canonical health — clean 200 responses and a self-referential canonical.
Structured data & metadata
- JSON-LD structured data — valid Schema.org markup (Organization, Product, Article, FAQ, HowTo) that turns prose into machine-readable facts. Google’s structured data guidelines are the baseline here.
- Title & meta description — present, unique, and descriptive of what the page actually answers.
- Open Graph & social metadata — consistent titles and descriptions that reinforce the page’s entity.
- Heading hierarchy — a single clear H1 and logical H2/H3 nesting that models use to map the page’s structure.
AI-specific files & performance
- llms.txt presence & validity — a spec-compliant llms.txt file that points models at your best content. The extension checks this automatically; if you don’t have one, our free llms.txt generator creates one.
- XML sitemap — discoverable and referenced from robots.txt.
- Core Web Vitals & mobile readiness — performance signals (see web.dev Core Web Vitals) that affect both crawl budget and user trust.
- Image alt text — whether facts locked in images are described in text a model can read.
The Content Score: is the content usable and trustworthy?
Once a model can read the page, it decides whether the content is worth reusing. These checks approximate the judgments answer engines make when they choose who to cite.
Clarity & answerability
- Direct answers up top — does the page state what it is and answer the core question in the first paragraph, or bury it below fluff?
- Skimmable structure — short paragraphs, bulleted lists, and descriptive subheadings that map cleanly to sub-questions.
- Question coverage & FAQs — whether the page addresses the follow-up questions a user would ask, ideally with FAQ schema.
- Entity clarity — is it obvious who you are, who you serve, and what you offer, without jargon?
Depth & trust (E-E-A-T)
- Content depth — enough substance to be a credible source rather than a thin stub.
- Author & source signals — named authors, citations, and outbound links to authoritative references. This overlaps with our free E-E-A-T scanner.
- Freshness — visible publish/update dates so models can judge recency.
- Internal linking — contextual links that help models understand your topical footprint.
How the 0–100 score is calculated
The score is not a simple average of pass/fail checks. Each check is weighted by how much it actually affects whether an AI engine can use the page. The logic works in three steps:
- Each check returns a graded result, not just yes/no—for example, structured data can be missing, present-but-invalid, or valid-and-complete.
- Checks are weighted. A blocked crawler or missing main content is catastrophic and weighted heavily; a missing Open Graph tag is minor. High-impact, machine-readability checks carry more weight than cosmetic ones.
- Results roll up into two sub-scores (Technical and Content), which combine into the overall AI Readiness Score.
This is why two pages can both “fail 5 checks” and get very different scores—failing five minor checks is not the same as failing one that stops a crawler cold.
Interpreting the number:
- 80–100 — strong. Machines can read, understand, and trust the page. Focus on content depth and citations.
- 60–79 — fixable. Usually missing structured data, weak headings, or thin answers. Quick wins available.
- Below 60 — at risk. Something structural—blocked crawling, JS-only content, no schema—is keeping you out of AI answers.
Before & after: what raising the score looks like
The score becomes real when you watch a page climb. Here are representative before/after patterns we see repeatedly across the D2C, SaaS, and hospitality pages we audit.
Example 1: A D2C product page (38 → 84)
A skincare brand’s bestseller scored 38. The audit flagged no Product schema, price and ingredients only inside images, and no direct summary of who the product is for. After adding Product and FAQ JSON-LD, moving key specs into text, and adding a two-sentence “best for” summary near the top, the Technical Score jumped and the overall hit 84. See the pattern applied across a category on our D2C & ecommerce page.
Example 2: A SaaS comparison page (46 → 88)
A B2B tool’s “alternatives” page ranked on Google but never appeared in AI answers. The Content Score was the bottleneck: no FAQ coverage, a vague H1, and no author or citation signals. Adding a clear comparison table, FAQ schema, and outbound references lifted it to 88. More on this in our SaaS & B2B playbook.
Example 3: A hotel landing page (31 → 79)
A boutique hotel page was almost entirely JavaScript-rendered, so crawlers saw an empty shell. Server-rendering the core content plus adding LocalBusiness schema and an FAQ took it from 31 to 79. See the hospitality vertical for more.
Why a recognizable, repeatable score matters
AEO is still young, and most teams have no shared vocabulary for “is this page ready for AI search?” A consistent, transparent AI Readiness Score gives you a benchmark you can put in a report, compare against competitors, and track month over month—the same way Core Web Vitals became a shared language for performance. Because the checks are explicit and reproducible, the number is defensible when you show it to a client or a stakeholder.
To go deeper on the strategy behind the score, browse the AEO Hub and compare tools in our best AI search visibility tools roundup.
How to get your score in 30 seconds
You don’t need an account or a crawl budget. Install the free SellOnLLM Chrome extension, open any page—yours or a competitor’s—and click the toolbar icon. You’ll get the overall AI Readiness Score, the Technical and Content sub-scores, and a prioritized list of exactly which checks failed and how to fix them. For a step-by-step walkthrough, see how to audit your site for AI search in 30 seconds.