How the AI Brand Truth Check uses AI

How this tool uses AI

Most of the AI Brand Truth Check tool results come from fixed rules, not AI. Your scores, readiness labels, risk level, top gaps, recommendation, first-90-days outline, prices, and every number in your result come from fixed rules applied to your answers. No AI is involved in any of them.

AI does three narrow jobs, each with a fixed checklist or a fixed set of facts:

Rating a pasted voice guide.
If you paste a brand voice guide, AI marks which of 8 checklist items it covers. It returns only those marks, never text from your guide.

Rating a pasted claims list.
If you paste a claims list, AI marks which of 7 checklist items it covers. It checks how the list is organized, not whether any claim is true or allowed.

Wording your report summary.
AI writes three short paragraphs from the labels in your result. It sees only fixed labels from your result, never your paste, your email, or anything you typed, and a check rejects any summary with a number, a price, or a gap your result doesn't hold. Status: off at the start of the beta. Your report uses a fixed summary until it turns on.

What the AI sees.

AI sees pasted content after emails, phone numbers, and links with tracking codes have been removed by the tool. The paste is used for one rating and isn't stored, logged, or put in your report.

Which AI.

Claude Sonnet 5.5, made by Anthropic, through Anthropic's API. Anthropic says it doesn't use API data to train its models and deletes it within 30 days, apart from limited cases its policy describes.

What it doesn't do.

AI doesn't decide whether a claim is true, proven, or allowed. It doesn't rewrite your documents. If the AI step is off or fails, your result still shows, built from your answers.

How it's checked.

Every change to an AI step runs against a fixed set of test cases before it goes live, and I read real results every week.

Questions?

Email kara.lineal@gmail.com.