A keyword search in the App Store returns your niche mixed with whatever else ranks for those words, and no field tells them apart. The genre is chosen by the developer, so a photo editor files itself under Lifestyle to catch beauty traffic. The rank is a keyword position, so an app called "earn money fast" can sit at number 7 for "skin diary". The first real scan proved it: 5 of the 40 measured apps were not in the niche at all, while 270 candidates that were never got looked at.
So the scan now stops and asks. It shows your Claude every candidate with its name, its maker, its categories and its own pitch, and your Claude answers which ones to measure and how close each one is: a direct rival, adjacent, or another niche entirely. Only then does it spend the measurements, and it spends them where you said.
The table follows that verdict. Apps from another niche move out of the way behind a toggle instead of sitting at the top, adjacent ones stay visible but out of the running, and the momentum bars are scaled between real rivals, so one big neighbour no longer flattens everyone else. The reading that opens the tab is a verdict now, not a report, and the TikTok creatives of each competitor show their thumbnails again.
- Every candidate the search returned is judged, not just the first 40 by Apple's ranking
- Each app is filed as a direct rival, adjacent or another niche, and the table sorts and hides by it
- Momentum bars scaled between direct rivals only
- Needs the companion on 0.9.0 or newer