Notes on AI visibility: how answer engines pick who they name, and what it takes to get your brand cited before your rivals do.
Which page type earns citations from which AI engine. Koalr's full archetype-by-engine matrix, including Copilot's year-stamp bias and Perplexity's odd habit of citing weak pages nobody else touches.
Koalr's citation data: 1-3 word slugs earn 9.6 citations per URL at a 48.3% mention rate, against 5.6 and 19.5% for 10+ word slugs. Why the gap exists and what to fix.
Koalr tracked 25 AI-visibility pages across seven engines. Six cite them. Claude cites zero. Here's why, and what to build instead.
Which page type earns citations from which AI engine. Koalr's full archetype-by-engine matrix, including Copilot's year-stamp bias and Perplexity's odd habit of citing weak pages nobody else touches.
Koalr's citation data: 1-3 word slugs earn 9.6 citations per URL at a 48.3% mention rate, against 5.6 and 19.5% for 10+ word slugs. Why the gap exists and what to fix.
Koalr tracked 25 AI-visibility pages across seven engines. Six cite them. Claude cites zero. Here's why, and what to build instead.
GPTBot, ClaudeBot and Google-Extended feed model training, not citations. Here is every AI crawler's real job, and how to check your own robots.txt.
A scoring method for prioritising GEO fixes: which citation gaps and weak pages to close first, ranked by buyer intent, competitor displacement, brand status and effort.
A buyer's checklist for AI visibility tools: real prompts vs sample lists, engine coverage across all seven, citation vs mention, and the multi-brand maths agencies miss.
GA4, Similarweb and Semrush measure AI referral traffic. None can see a citation that never produced a click. The difference, and how to track both.
What a GEO audit should actually measure, how to build the prompt set behind it, and how to turn findings into fixes instead of a report nobody reads.
Scan your site free and get your AI Visibility Score, plus your first ranked fix list, in minutes. No card, no demo call.
One scan, 25 real prompts, 7 AI engines, under five minutes, zero setup.
