What works on X (Twitter), measured account by account
A lot of advice about X comes from one account's experience or from charts that pool thousands of posts, where the biggest accounts decide the result. Here every one of 13 creator accounts was measured only against its own posts, 3,762 in total, and every pattern had to pass a test against chance in that account before it counts.
What this sample is, and what it is not
13 public creator accounts on X, chosen by us, mostly in tech, AI and indie building, with 62 to 3,735 followers (median 1,424). For each we took the newest original posts up to 500 (114 to 499 per account), published between 2 Jan 2026 and 14 Sept 2026, with their public reach, bookmarks and replies. Replies to other posts are left out.
Every account was measured only against its own posts, so a large account cannot outvote a small one. “Reliably higher in 5 of 12 accounts” means the gap passed that account's own test against chance in 5 of the 12 accounts that had enough posts in the group. It is a pattern in these accounts, not a law of X, and it does not prove that the kind of post causes the gap.
Your posts are not the median
The same test runs on your own posts: the free plan reads your latest 50 through the official X API, read-only, and shows whether these groups hold up for you; the Deep Check names every one. About 2 minutes, no payment details.