Notes on AI search, measured honestly.
How AI engines decide what to say about brands, what the tools in this category get right and wrong, and what we are learning by measuring the answers themselves. Written by the team building ClarAI, disclosed as such.
Prompt Demand Without the Billion-Conversation Dataset
You do not need a proprietary dataset of a billion AI conversations to estimate what buyers ask AI engines. Triangulate prompt demand honestly from three public signals: Google's People Also Ask and autocomplete, your own Search Console query data, and real community questions, then label it as triangulated demand with its sample and date.
Read the post →
The GSC Generative AI Report: What It Shows and What It Hides
Search Console's 2026 Generative AI report is real first-party proof that Google's AI surfaces put your pages in front of people. But it shows impressions only, offers an export and no API, and is Google-only and silent on prompts, answers, and competitors. What it shows, what it hides, and the cross-engine gap it leaves.
Read the post →
Reading Query Fan-Out Like a Map
When someone asks an AI engine a question, the engine fans it out into many sub-queries and stitches the answer from whatever sources win each one. The practical unit of AI visibility is that map of sub-queries, and how ClarAI's 12-axis Query Fusion Model plus per-prompt reverse-engineering make it operable.
Read the post →
Google-Extended Does Not Do What You Think It Does
Blocking Google-Extended in robots.txt does not remove you from Google AI Overviews or AI Mode. It governs Gemini training and grounding in the Gemini app and Vertex AI, and the training of the models behind Search's AI features, not whether you appear in them. The crawler that actually gates Google's AI answers is standard Googlebot, and confusing the two is the most common self-inflicted AI visibility outage.
Read the post →
Is Your Site Agent-Ready?
AI search crawlers download your HTML but do not run your JavaScript, so client-rendered pages read as empty to them even when they rank on Google. What agent-readiness means, the two raw-HTML failures ClarAI's site audit looks for, and how to check your own site.
Read the post →
GPTBot, ClaudeBot, PerplexityBot: Who Is Crawling You and What to Allow
A plain guide to the AI crawler user-agents ClarAI checks by name, split into the retrieval agents that decide answer visibility (OAI-SearchBot, PerplexityBot, Claude-SearchBot, Claude-User) and the training crawlers that do not (GPTBot, ClaudeBot, plus the Google-Extended policy token), and why Google-Extended does not gate AI Overviews.
Read the post →
AI Shopping Is the 2026 Wave: How Product Cards Get Picked in ChatGPT
AI answers assemble a shortlist of specific products, not a ranked list of category pages, so each product's facts need to be accurate wherever an engine reads them. What Google published about product structured data and Merchant Center (eligibility for product listings, not a citation lever), what ClarAI's AI Shopping feature measures, and how to judge demand without a made-up number.
Read the post →
Snippet Controls Are Your AI Visibility Off Switch
How nosnippet, max-snippet, data-nosnippet, and noindexed sitemap URLs silently remove ranking pages from AI answers, how to check your own site in minutes, and the exact fix for each case.
Read the post →
The 10 Best AI Visibility Tools in 2026, Honestly Ranked
A use-case ranking of the AI visibility market in 2026: Profound, Peec, Otterly, Reach, ClarAI and five more, with real pricing, honest tradeoffs, and a disclosure most lists skip.
Read the post →
Your AI Visibility Score Is Lying to You (Ours Might Be Too)
AI answers are nondeterministic, samples are small, and citation distributions are power-law. What honest AI visibility measurement looks like, and exactly how ClarAI implements it.
Read the post →
Google Finally Told Us How AI Search Works. Here Is What Actually Matters
Google's official guide confirms AI Overviews run on RAG plus query fan-out over the Search index, sets the eligibility bar at indexed, snippet-eligible, and crawlable, and debunks llms.txt, chunking, and special AI schema.
Read the post →
AEO Retainers: 2,000 to 8,000 Dollars a Month in One Published Guide. Here Is What Agencies Actually Do
What an AEO retainer includes, the white-label wholesale layer underneath the market, when a retainer beats self-serve tooling, and the hybrid model agencies are quietly running.
Read the post →
Coming soon
Get a score that shows its work
Run a free check on your domain. You get real answers, real citations, and the sample size printed next to every number.