Changelog

Seven behavioral GEO checks added to the rubric, targeting how ChatGPT Search, Perplexity, Gemini, and Google AI Overviews actually choose what to quote: first-paragraph brand definition, citation-target schema coverage (FAQPage / HowTo / ProfilePage / ClaimReview), Q&A answer density, fresh-content recency rate, People Also Ask alignment (cross-references your content against real PAA questions on category SERPs via DataForSEO), citable claim density, and brand disambiguation strength. Engine fleet now classified as search-grounded (live web retrieval — ChatGPT, Gemini, Perplexity, Google AIO) vs. knowledge-presence (training data only — Claude, Grok, DeepSeek, Llama, Microsoft Copilot), surfaced in the dashboard with explicit badges so customers can read "absent on Claude" as "not yet in training" rather than "broken scan". Google AI Overviews engine now uses real DataForSEO SERP scrape on pro+ plans instead of a Gemini chat in disguise.

Passage and chunk readiness check added: every paragraph in the main content area is scored on whether it sits in the 40-200 word retrieval sweet spot, lives under a nearby heading, and stands alone enough to quote. Pages whose paragraphs hit the sweet spot get cited at materially higher rates by RAG pipelines, ChatGPT Search, Perplexity, and Claude with search tools.

Five new checks land: Google Knowledge Graph MID detection (we now query Google's entity database directly, not infer from sameAs), tier-weighted press coverage (a Wall Street Journal mention counts more than an aggregator), backlink quality (authority + anchor diversity, not raw counts), local business presence (GMB + local SERP for regional brands), and duplicate-content detection. The JSON-LD walker now descends into Google-recommended nested placements (Review inside Product, Author inside Article). A silent footer-content bug that zeroed about/team/contact/social checks for most sites is fixed. Public /methodology/sources page shipped with versioned, citable snapshots of every AI platform's documentation.

The audit now inspects a representative sample of pages across your site instead of grading every schema type against the homepage alone. Multi-page sites will see fairer scores — schemas that live on inner pages (articles, author bios, tutorials, breadcrumbs) are now credited where they actually exist.

Additional audit checks shipped across crawler access, structured data, content quality, technical foundation, and content authenticity. No weight changes; scores stay comparable to v2.0.0.

Expanded entity recognition coverage, strengthened answer-prominence tracking across the engine fleet, rebalanced category weights to reflect 2026 citation reality. Trend charts render a marker on the changeover date so score shifts are legible.

First published methodology. Established the rubric structure, the engine fleet, and the versioning discipline we carry forward.