One thing I keep bumping into with the "publish where AI looks" step: how do you know it looked? You can rank and still never get fetched by ChatGPT or Perplexity and even when a bot does fetch the page, half the time it can't actually parse what's on it. Everyone's optimizing for AI citation with zero feedback on whether the agent read the page or bounced.
Curious how you're closing that loop on your own runs... are you checking which crawlers actually hit the published URLs or treating "published + ranking" as the proxy?
Yesterday we deployed agentic traffic analysis for NextJs websites. Now you can adapt interfaces for humans and also understand what agents are looking for in your website.
We're building the most advanced intelligence and personalisation traffic tool out there.
introducing Agentic traffic analysis for #NextJs projects within SentientUI. Now you'll be able to see which #AI agent has visited your website and with what intent.
Our aim is to help you understand your audiences, not just human but also agentic.
@MicroLaunchHQ Building SentientUI, understand your traffic (users vs agents) and display different interfaces for each user persona while displaying the right data for agents. Only 1 goal, improve your conversion
@karpathy Maybe Claude should start working on video analysis. At the end of the day everything works around feedback loops. If it can get time analysis coherency it wouldn't need to rely on screenshots