# Read your report

> What the AI Presence Score means, how to read wins and losses, and how to drill from a headline number to the exact AI response behind it.

- Section: Getting started
- Updated: 2026-07-13
- Canonical: https://knitknot.ai/docs/read-your-report/
- Publisher: KnitKnot, the AI Presence Management platform (https://knitknot.ai)

---

After a run completes, your report is the summary of everything the four engines said. Every number in it drills down to the underlying evaluations — nothing is a black box.

## The AI Presence Score

The headline is a 0–100 composite of how favorably AI represents you in buyer evaluations: how often you show up, how you fare head-to-head, whether what AI says about you is accurate, and how you're framed. It's designed to be tracked over time — the trend across runs matters more than any single reading.

Two supporting numbers to know:

- **Visibility** — the share of evaluations where you were mentioned at all (excluding prompts that name you directly, which would inflate it).
- **Win rate** — of the head-to-head comparisons where the AI picked a vendor, how often it picked you.

## What AI tells buyers

This section is the qualitative core: the recurring claims AI makes about your company, with the captured quotes and engines behind them. Claim conclusions use available company facts, receipts, and validation gates; missing evidence remains unverifiable rather than being treated as false.

## Competitive results

Head-to-head outcomes are broken down per competitor and per feature: where you win, where you lose, and what the AI said when it picked the other vendor. A loss row links straight to the responses behind it, so you can read exactly how the AI justified the pick.

One nuance: if AI recommends one of your own products over another, that's a win for your brand, not a loss — the scoring knows your product family.

## Sources

When a captured response exposes a reliable citation signal, KnitKnot binds the claim to that source. The sources view shows the owned, competitor, and third-party pages present in the captured citations. When no reliable signal can be bound to a claim, attribution remains empty.

## From report to action

Repeated gaps in the report can become **Issues** — persistent records such as a factual problem, missed feature, losing comparison, coverage gap, or access failure. Issues feed **Playbooks**: evidence-grounded briefs or technical fix plans. After you ship, a later full benchmark compares observed answer, citation, score, and issue changes without assuming the intervention caused every movement.

## Sharing

Your report has a public, shareable version at your workspace's `knitknot.io` address — the same numbers and drill-downs, readable by anyone with the link. Useful for getting the rest of your team looking at the same evidence.