Three questions you cannot answer with one answer

The robots protocol is over thirty years old and asks a single question: may you fetch this address? For search engines that was enough for a long time, because fetching and displaying served the same purpose.

With AI systems the two come apart. A fetch can become training material, a footnote in an answer, or both — and anyone who can only say “yes” or “no” inevitably says yes to something they didn't want.

The three questions

Training: May your text feed into a model? This is the question with the longest reach — what is once in doesn't come back out.

Search: May your page sit in an index and appear in answers? This is usually the question you want to say yes to, because this is where visibility comes from.

Fetching at runtime: May an assistant retrieve your page at the moment someone asks? Here you get the current state into the answer, instead of a training snapshot from the day before yesterday.

Why blocking wholesale is expensive

The reflexive answer to AI crawlers is to block everything. That protects against training — and at the same time costs you the visibility the website exists for in the first place.

Answering the three questions separately is therefore not a technical detail. It is the point at which a business decision becomes expressible at all.

What this site says

findready.ai answers the three questions for itself and writes the answer into its own robots.txt: no training, search permitted, runtime fetching permitted.

That is verifiable, and by you: the file is public. A claim about findability that you can't demonstrate on your own house isn't one.