How We Count Our Numbers
This page defines every statistic SiteCandor publishes about itself. Below you will find what goes into each number, what is left out, and how often it is worked out again — so if one ever looks wrong, this is what you hold us to.
The Three Rules Behind Every Number Here
No statistic we publish about SiteCandor starts anywhere but zero, and none has ever been set by hand.
Nothing is ever padded. We do not estimate or extrapolate, and we round only to the nearest whole number, or to one decimal place for the average score. If our database says 37, we publish 37 or nothing at all.
Small numbers are labelled as small. Below 100 audits we say plainly that we are new rather than dressing a small figure up as scale.
Every Number, Defined Exactly
Audits Run
A straight count of every audit we have run and stored — free audits and the paid plans' monthly audits alike. If the same site is checked three times, that is three audits, because each one was real work. The count goes up only when an audit actually finishes, and it falls only when an audit is deleted at someone's request or under our retention rules. An audit that could not read a single page of its site measured nothing and has no score, so it is not counted here or in any number below.
Average Overall Score
The average of the out-of-100 score across every audit counted above — the same score your own report shows. Low scores are included, repeat scans are included, and nothing is left out to flatter the average. We show it to one decimal place at most.
Most Common Failed Check
For every website whose audits we still store, we look at its most recent audit only, and count which single check failed on the most websites. We count each website once, so scanning one site fifty times cannot tip the result. A check that did not apply to a site — looking for images on a page that has none — is left out rather than counted as a failure.
Sites Blocking Automated Readers
Of the websites we have checked, the share that turn away at least one of the 14 automated readers we look for: GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, Claude-User, anthropic-ai, Google-Extended, PerplexityBot, Perplexity-User, CCBot, Bytespider, Applebot-Extended and meta-externalagent. Most of these are robots that fetch pages; Google-Extended and Applebot-Extended are permissions Google and Apple read rather than robots that fetch. Either way it is a different set from the AI assistants a subscription puts your questions to. We read the site’s own robots.txt to find out. Sites where we could not tell are left out of the count entirely rather than assumed one way or the other. Turning these tools away can be a perfectly deliberate choice; this number says how common it is, not that it is wrong.
How Fresh, and What Can Go Stale
These numbers are worked out again once a day, and wherever we show one we show the date it was last worked out beside it. If that ever fails, we keep showing yesterday’s real numbers rather than anything invented. If we cannot read them at all, we show none rather than a made-up figure.
Spot a Number That Looks Wrong?
Tell us, and we will fix either the number or the explanation — whichever of the two is lying. Every page has a Contact link in the footer.