Can the crawler read you
isready.ai is an open-source AI-readiness audit for websites. It scores whether an AI crawler can actually read the page.
Chrome can. Googlebot can render JavaScript. GPTBot and ClaudeBot, as their vendors document the training and search crawls, fetch raw HTTP: no JavaScript, short timeouts, an honest user-agent. A React or Vue app that looks finished in a browser can arrive as an empty document. Classic SEO dashboards stay green while that happens.
Thirty-two evidence-based checks across five dimensions, scored 0 to 100. The scanner engine and the CLI are MIT. The web scan is free, no signup. On 13 August 2026 the homepage showed 88.9K checks, 21 AI crawlers tracked, a pass in about five seconds.
The homepage still says it this way: “The future is AI. Is your website ready for AI?”
When to run it
Before you care whether ChatGPT, Perplexity, or Gemini can actually read you. After a SPA or JavaScript-only redesign. In CI, so the score cannot drop in silence.
Google’s AI optimization guide is worth reading once. You do not need special AI files for Google Search. Google Search does not use llms.txt. isready sits next to that work. It is the check for the bots that never execute the bundle.
How to use it
npx isreadyai yourdomain.com
That is a full audit. Add --llm for a resolution plan. Add --md for markdown solutions. Both stay free. --json and --quiet are there for scripts. --deep walks more than the homepage. --smart-ai is a second pass with a real browser, agent-browser from Vercel Labs, for agents that can click.
Or open https://isready.ai and scan from the page. Same question. No account.
There is a GitHub Action, isreadyai/audit-action, if you want the score as a CI gate.
There is a particular relief in a CLI that prints the empty-shell problem in one screen and then writes the markdown for whoever will fix it.
How the 32 checks work
The default fetch imitates the crawler. Raw HTML. The score splits five ways.
Crawler access is 25 percent. robots, challenges, whether the well-known AI user-agents are allowed in.
Rendering is 20 percent. Did the response contain the article, or only a JavaScript bundle a training crawler will not run.
Structured data is 30 percent. JSON-LD, schema, facts a model can quote without guessing.
Trust and security is 10 percent. HTTPS and the basics a bot should not have to infer.
Content GEO is 15 percent. The GEO literature tested quotations, statistics, and citations. Keyword stuffing was not one of the winners.
Each finding carries an observed value, a consequence, and a concrete fix. The score is versioned, so a later methodology change does not silently rewrite old reports.
llms.txt is a community proposal. It is not crawl control. isready reports whether the file exists. It never moves the score. In isready’s server logs, AI crawlers fetch it in about 0.1 percent of visits.
OpenAI documents several bots. GPTBot, OAI-SearchBot, ChatGPT-User. Anthropic splits ClaudeBot, Claude-SearchBot, and Claude-User the same way. A user-triggered fetch often ignores robots.txt. A training crawl usually does not. Mixing those jobs is how a site ends up with a file that does not do what people think.
Cloudflare’s AI Crawl Control is the other thing to read if a challenge sits in front of bots. In May 2025 GPTBot went from about 5 percent to 30 percent of AI crawler traffic on Cloudflare’s network. The March 2025 figure of 50 billion requests a day is about 1 percent of that network, not 1 percent of the whole web.
isready does not promise ChatGPT rankings. A training crawler, a search crawler, and a user-triggered fetch are three different machines.
What the papers measured
The numbers below are from the papers.
Aggarwal et al., GEO, KDD 2024. Quotations improved position-adjusted word count by about 41 percent. Statistics about 32 percent. Cite Sources helped rank-5 sources by +115.1 percent and hurt rank-1 sources by −30.3 percent. Those gains sit inside a fixed 5-document context. The experiment is not organic discoverability. Keyword stuffing did not help.
Liu, Zhang, and Liang, Evaluating Verifiability in Generative Search Engines, Findings of EMNLP 2023. Citation recall 51.5 percent. Citation precision 74.5 percent.
Martinez, Optimizing Visibility in Generative Engines, arXiv:2607.14035. The widely cited ~40 percent GEO figure is conditional on sources already retrieved. In the 45-study set there is no stable longitudinal causal effect on organic traffic.
Puerto et al., C-SEO Bench, NeurIPS 2025. Conversational SEO gains shrink as competitors copy the rewrite.
That is why isready scores fetch and structure first. Citation share inside an answer is a later measurement. A 92 on this scan will not put you in ChatGPT.
Next to the citation products
Most GEO products sell brand citation share inside answers. Prompt panels. How often ChatGPT names you.
isready sells whether the site is even readable to crawlers and agents. Adjacent markets. You can need both.
In Italy, GeoSonar (GeoSonar S.r.l., Rome, founded 2025 by Federico Fancinelli and Chiara Barbieri) is a paid SaaS. Free tier, then Pro at €49 a month, Agency from €199. GEO Score 0–100 across ChatGPT, Perplexity, Gemini, Claude, Copilot. Citation network. A technical GEO audit. Not open source. They track how often a brand is cited. isready checks whether the machine can read the page. The CLI here is MIT and free. Monitoring is what you pay for.
Further out: Profound, Peec AI in Berlin, Goodie at $399 a month, OtterlyAI in Vienna from $29 a month. Paid citation and visibility analytics. Not OSS. Apify’s Agent-Readiness Checker looks at MCP and agent.json endpoints. Another job.
What you actually pay for
The scan, the CLI, --md, --llm, and the GitHub Action are free.
Pro is €19 a month. Team is €49. That money is monitoring, history, a live badge, Ask-your-site, quotas for automated fix PRs. The hosted dashboard is source-available under PolyForm Shield, Smart Squad S.r.l.
Run the CLI. Pay only if you want it watched.