GeoLook shows you what AI search engines actually say about your brand, then turns that uncomfortable answer into a fix list you can work through and verify. You sample real answers across a wide matrix of AI engines and watch mention rate, ranking, and citation share appear per engine, with raw answer replay when a number looks suspicious. Competitor tables reveal which rival each engine prefers and which citation sources they rely on that you do not. A structured question bank sorts buyer, education, and probing questions, flagging the ones where you are absent, outranked, or described negatively. The site audit checks whether crawlers can reach, understand, and quote your pages, in the order those failures actually block visibility. Findings become tickets with an owner, effort estimate, risk grade, and acceptance check, and later sampling rounds confirm whether a fix worked or reopen the ticket when results slip. A content workbench ranks the topics most worth writing, checks drafts for citability, and prepares assets such as llms.txt and structured data snippets for publishing. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.