Skip to content

How to read rank changes without fooling yourself

Most rank change is other apps arriving or leaving, not your app moving. In the 24 hours to October 13, 2026, 224 rows changed position in the top 100 of ChatGPT's ranked lists, yet by cohort rank not one ChatGPT app moved; in Claude's top 100s, 172 raw changes held 125 listings that really moved.

Raw changes, ChatGPT
22424 h, ranked top-100 lists
Cohort moves, ChatGPT
0across 16 ranked lists
Claude listings that moved
125top 100 of each list; 52 by 3 or more places
Slid one place on Oct 9
17 of 18when Google Drive jumped

Two ways to count a move

A raw rank change compares an app's position now with its position 24 hours earlier. A cohort rank change first sets aside every app that wasn't in the list at both times, then compares positions among the apps that remain. If one app arrives above you, your raw rank falls by one and your cohort rank stays put.

A real case from the last 24 hours: HoneyBook was #34 in Business & Operations a day ago and is #35 now. Raw, that is a fall of 1. But the only change above it is Box arriving, and among the apps present both days HoneyBook holds the same place, so its cohort change is 0.

On ChatGPT, three events explain almost all of the 224 raw changes in the last 24 hours:

  • Creativity: 96 apps, from #5 down, each rose exactly one place. The app that held #5 a day earlier is no longer in the top 100. Cohort moves: 0.
  • Business & Operations: Box entered the list at #34, and the 66 apps below it each slid exactly one place. Cohort moves: 0.
  • Productivity: 59 apps, from #42 down, each rose exactly one place. The app that held #42 a day earlier is no longer in the top 100. Cohort moves: 0.

None of those apps did anything, and neither did the apps behind the other 3 raw changes. All 16 ranked ChatGPT lists had zero cohort moves in the window.

Raw moves and cohort moves in 24 hours, by list
  • ChatGPT · Creativity96 raw · cohort 0
  • ChatGPT · Business & Operations66 raw · cohort 0
  • ChatGPT · Productivity59 raw · cohort 0
  • ChatGPT · Education2 raw · cohort 0
  • ChatGPT · Developer Tools1 raw · cohort 0
  • Claude · Most popular25 raw · cohort 25
  • Claude · Connectors trending91 raw · cohort 82
  • Claude · Connectors by popularity23 raw · cohort 23
  • Claude · New (launch order)97 raw · cohort 0
Top 100 of each list. Long bars with a cohort count of 0 are arrivals and departures, not movement. Claude's New list is in launch order, so its 97 raw moves are new listings at the top. Source: OpenApps data, Oct 13, 2026.

One arrival moves everyone below

The cleanest example we have is an insertion wave. On Oct 9, Google Drive went from #49 to #2 in Popular. Of the 18 other apps with a stored Popular series, 17 slid exactly one place at the same time. The one that didn't was Remote Desktop Commander, which stayed at #1, above the slot Google Drive took.

Google Drive and Canva in Popular, the week to Oct 13
#2#49Oct 5Oct 13Oct 9

Google DriveCanva

Google Drive's jump pushed Canva from #13 to #14 at the same moment. Canva did nothing, and its raw rank fell anyway. Source: OpenApps data, Oct 13, 2026.

Over the whole week Canva went from #11 to #17. Read raw, that is a decline. Raw rank alone can't tell you how much of it was Canva's own doing. Cohort rank, computed step by step, can.

Drift below #50 and frozen lists

Rank drift is the slow version of the same effect. Catalogs grow, new apps land above older ones, and positions below about #50 sink a little at a time. Of the 224 raw changes on ChatGPT in the last 24 hours, 153 (68%) were at #51 or lower. Our charts mute raw change below #50 and show cohort change by default.

Frozen lists are the opposite trap. Popular had 0 raw moves in the 24 hours to October 13, 2026. A quiet list is not proof that nothing happened in the store, and a single jump in a list that rarely moves (like Google Drive's) is worth more attention than a hundred one-place slides in a busy one. OpenAI has said the directory order has weighed retention since Aug 21, 2026 (see retention-weighted ranking). Our data shows when the order changes, not why.

Claude's two traps

Claude moves more. In the same 24 hours, 125 Claude listings in the top 100 of a ranked list moved by cohort rank, 52 of them by 3 places or more. In Most popular, 25 of the 25 raw changes were also cohort moves, so on Claude a raw change is more often a real one. Two things still need care.

First, Most popular is three lists stuck together, so rank within kind: a plugin's position means something only among plugins (more in Claude connectors vs plugins vs skills). Second, New is ordered by launch date, not by rank. Its 97 raw "moves" are all steps down, as newer listings arrive at the top (2 new listings in its top 100 since the day before). We show no change column on launch-ordered lists.

The biggest climbers by cohort rank were Taplio (up 35 to #53), Bitrise MCP (up 21 to #78) and bolna (up 16 to #41), and the biggest fallers Jotform Sign (down 13 to #75) and Sonos (down 10 to #87). All 5 were in Connectors trending.

Five checks before you call it a move

  1. Use cohort rank. If the raw change is there and the cohort change is 0, the app didn't move.
  2. Look above. One app entering a list moves every app below it down a place. One app leaving lifts them all.
  3. Discount anything below #50 unless the cohort change agrees.
  4. Check how the list is ordered. Launch-ordered lists (Claude's New) and reshuffled ones (Claude's recommended connectors) have no ranks to compare.
  5. On Claude, rank within kind. Count only listings of the same kind above you.

Then keep the window in mind: a day of movement is news, not a trend. The Movers page ranks by cohort change, and the glossary defines cohort rank and raw rank change.