# AgentReady

> AgentReady measures whether AI crawlers and agents can read your site, generates the files that let them, and tracks per-engine citation rates with their confidence intervals.

## What this is

AgentReady answers one question: can an AI crawler read this site, and is it
being cited as a result. Those are separate measurements and are reported
separately — readiness is deterministic and scored out of 100; citation is
sampled per engine and reported with a Wilson interval and its sample size.

There is deliberately no single blended "AI visibility score". Roughly 11% of
domains are cited by both ChatGPT and Perplexity, so averaging across engines
discards the only signal that distinguishes them.

## Pages

- [AgentReady — can machines read your site?](https://agentready-fn.ayushbaluni.in/): What the product measures, how the seven readiness categories are weighted, and why citation rates are reported per engine with intervals.
- [Free scan](https://agentready-fn.ayushbaluni.in/scan): Run a readiness scan on any URL without an account. Returns the category breakdown and the ranked list of what to fix.
- [Frequently asked questions](https://agentready-fn.ayushbaluni.in/faq): What AI readiness measures, why it is kept separate from citation, what blocking GPTBot actually does, and why there is no single blended visibility score.
- [Terms](https://agentready-fn.ayushbaluni.in/legal/terms): What the service does, what it does not promise, and the limits of the measurements.
- [Privacy](https://agentready-fn.ayushbaluni.in/legal/privacy): What is stored, what the beacon records, how third-party tokens are encrypted, and how to have it deleted.

## How the measurements work

- Readiness is scored across seven weighted categories: rendering (30),
  crawler access (20), structured data (15), content structure (15),
  technical hygiene (10), llms.txt (5) and freshness (5).
- Citation rates are sampled per engine. Every rate carries its sample size
  and a 95% Wilson confidence interval; below 100 runs the run count is shown
  instead of a percentage, because a point estimate from a small sample is
  noise presented as fact.
- A mention is not a citation, and the two are never merged.
- Missing data is reported as a gap, never as a zero. Search Console and Bing
  report only queries a site already appears for, so a day with no row means
  no impressions were reported — not a rank of zero.

## Limits

- Training-set membership cannot be reliably measured. We track a slow proxy —
  whether a model describes the brand correctly with browsing off — quarterly,
  not as a live metric.
- AI referral counts are a lower bound. A large share of AI traffic arrives
  with no referrer and is attributed to Direct.
- Relationships between deterministic signals and sampled outcomes are
  described as preceding or correlated, never as causing.

## Contact

https://agentready-fn.ayushbaluni.in/scan runs a free scan on any URL without an account.