Original measurements, not opinions. Every post is built on queries we actually ran against the AI engines.
2026-08-05
Profound has raised $154M at a $1B valuation. Twenty two days of publishing put us ahead of them on every AI engine that reads the web. 32,908 measured answers, weekly numbers, and the losses included.
2026-07-13
We re-ran the study against grounded engines: 10 buyer-intent prompts, 3 engines, 50 search-grounded responses. HubSpot leads at 66.2, Pipedrive overtakes Zoho for second, and 7 of 20 brands are named zero times.
2026-07-28
Pages that rank across Google's fan-out sub-queries are 161% more likely to be cited. Branded mentions correlate at 0.664, YouTube at 0.740, word count at 0.04. The AI Overviews levers ranked by measured evidence, not vendor claims.
Position one loses 58% of its clicks when an AI Overview appears, up from 34.5% ten months earlier. 68% of US searches now end without a click. And only 37.9% of the pages Google cites in its own AI Overviews come from its own top 10.
2026-07-27
84% of AI citations are earned media, and the only before/after experiment measured a 239% citation lift from third party distribution. How engines resolve brand names without a brand database, where mentions actually live, and the tactics documented to backfire.
2026-07-22
An AI answer is only as good as the web index under it, and only four companies run one: Google, Microsoft Bing, Perplexity, and Brave. ChatGPT rents Bing, Claude is evidenced to rent Brave, DuckDuckGo sources from Bing. What each engine crawls, indexes, and ranks, with every claim linked to a primary source.
A 200 billion URL index updated tens of thousands of times per second, hybrid retrieval, prefilters that drop stale pages, and documents decomposed into individually ranked spans. What Perplexity's own engineering posts reveal about getting cited, with every claim linked to their published words.
ChatGPT search runs no web index of its own. It is a fine-tuned GPT-4o model that rewrites your question, queries third-party providers including Bing and Shopify, and writes up the results. What OpenAI's own docs reveal about getting cited, and why OAI-SearchBot, not GPTBot, is the crawler that matters.
Anthropic documents Claude's web search tool but never names the search engine behind it. The evidence points hard at Brave: a subprocessor added one day before launch, a leaked parameter name, and reproduced matching citations. What is documented, what is inferred, and how to get cited.
2026-07-19
We checked every major vendor's own docs: generation automates behind human gates, community presence (half of AI citations) is automated by nobody, and the two links that would prove any of it works have never shipped.
2026-07-16
AI-referred visitors convert 54% better. Half of B2B buyers start research in a chatbot. Fortune 500s hire AEO managers at $171K. The verified case for GEO, including the famous numbers that turned out to be wrong.
2026-07-15
80% of AI-cited URLs rank nowhere in Google. ChatGPT and Perplexity share only 11% of their sources. The primary research on how answer engines pick brands, and why most are invisible.
Honest, specific breakdowns against the AI-visibility monitoring platforms (Peec AI, Profound) and who each is actually built for.
The Answers section holds 212 tested responses to specific AI-visibility questions: why brands don’t show up in ChatGPT, Perplexity, Gemini, or AI Overviews, and what fixes it. New answers publish daily.
