July made the benchmark record more useful end to end: better company knowledge, clearer site priorities, and stronger evidence when a report turns into work.
Permalink: https://knitknot.ai/changelog/2026-07/
Release ledger
July made the benchmark record more useful end to end: better company knowledge, clearer site priorities, and stronger evidence when a report turns into work.
01
The Company section now brings together facts from your own pages and the pages AI cites, with provenance, receipts, history, proposals, and human curation. Third-party facts remain proposals until they meet the required corroboration standard.
02
The page-readiness score now uses structural signals from a page’s real HTML, including answer-first openings, self-contained sections, extractable facts, source links, and page-type-aware expectations. Meaningful site-health failures become persistent issues with an explanation, a lifecycle, and a clear route to the work that can resolve them.
Page readiness
78/100
03
Reports can be locked to a measurement period, while claim attribution uses only a response’s real citation signal. Cited URLs are now first-class rows with ownership, type, reachability, and source detail—so a report stays reviewable after it is shared.
Claim receipt
Cited“The platform gives QA teams a consistent way to review and coach conversations.”
Source behind this claim
docs.example.com / coaching
Owned page · ChatGPT · current period
04
The Library now has a topic-level view alongside the flat prompt list. It gives teams a faster way to see which areas are covered, where demand sits, and which specific prompts explain a topic’s movement.
Topics
By AI Presence
