AI Visibility Grader

How the score is calculated

The tool generates up to six buyer-intent prompts from your brand, category and domain, then answers all of them in one batched call so the measurement stays internally consistent. Each answer is parsed for three things: whether your brand is named, whether it appears as a linked source, and how it is characterised. Every prompt is separately checked against live Google for an AI Overview. The score is arithmetic rather than opinion - 45 per cent presence, 30 per cent citation rate, 25 per cent sentiment. The model writes the commentary; the code sets the number.

What each prompt row shows

Under the engine matrix, every prompt is listed as won or missed, so you can see which buyer questions you already own and which ones name somebody else. Opening a row gives you the answer text the engine produced and the sources behind it, which is usually more useful than the score itself: it tells you which pages, directories and community threads the engine currently treats as authoritative in your category.

• ›Won or missed, per prompt
• ›Your position in the answer when the brand appears in a ranked list
• ›Whether you were cited as a source or only mentioned in passing
• ›The full answer text, so you can read how you are framed
• ›The source URLs the engine leaned on

The Fix Pack and the work it points at

Two artifacts are generated in code so they are always present: an llms.txt scaffold for your site root, and a FAQPage JSON-LD block built from the exact prompts you were measured on. The model adds a comparison-page brief aimed at the questions you missed, plus three talking points for the forums where competitors are being recommended instead of you. Enter up to four competitors before running and the opportunities are written against them by name. The whole pack downloads as a single text file.