The Ground rules

Sometimes an analysis answers "can't tell from the post" instead of giving you a confident number. That's the honesty floor working as designed.

It's a fixed set of rules — the Ground rules — sitting underneath every AI analysis, and nothing overrides them, not even your own custom prompt. You can read them in full in the Settings custom-analysis builder, so you always know how each run begins.

What the floor holds to

• Observable only. Every statement is grounded in what is actually in the post's media. No invented facts, intentions, or audience reactions.
• "Can't tell" over a guess. When the post doesn't show something, the AI says so rather than fabricating an answer.
• No performance or virality claims. The AI holds no reach, views or performance data on the post, so it never predicts that one will "work", "go viral" or "succeed". Asking it in your own prompt to rate a hook 1–10 is fine — that's a craft judgment, not a claim about the real world.
• No trend claims. It won't call an audio "trending" or a format "viral" — no live trend data exists for it to know that.
• Patterns, not campaigns. It deconstructs and generalizes to portable patterns; it doesn't write finished campaign copy or personalize to your client's brand. That work stays yours.
• Your AI language. An analysis comes back in your agency's AI language; the AI's reads of a post follow that client's own. See Huntit and your language.

Why this is a feature, not a limit

An AI that always sounds confident is an AI you can't trust.

The floor is what makes a Huntit analysis worth reading: when it does make a claim, that claim is anchored to something visible in the post, not to a guess dressed up as insight. A short "can't tell from the post" is the system refusing to mislead you.