Wikipedia daily top lists show no acceleration of collective attention in 2016–2025; the English slowdown depends on one-day spikes with low app traffic

Question: did collective attention on Wikipedia become faster from 2016 to 2025, as Lorenz-Spreen et al. (2019) found for other platforms, and is the reported English slowdown (sci:sha256:d88df5ed97e71c69813236e7fc57a74920fb497209326818f8096ed355bf8cdc) robust to automated traffic? Data: Wikimedia daily top-1000 lists (user traffic) for English Wikipedia (all-access, mobile web, mobile app) and German Wikipedia (mobile web, mobile app), plus full daily series for 1,190 sampled English peaks. Method: we calculated top-10 and top-100 turnover and residence, median relative gains and losses, and peak ratios for 11 list variants. One new control is an app-consistency filter: it removes article-days whose mobile-app share is below a fraction rel of the daily median. Trends are OLS slopes of 10 yearly values; yearly intervals come from a 28-day block bootstrap. Results: none of 88 trends showed significant acceleration, except one German rise measure in unfiltered data. The largest upper 95% bound for acceleration was +0.4% per year for top-100 turnover; Twitter top-50 residence time changed by about -12% per year in 2013-2016. Unfiltered English lists became slower (turnover -1.9% per year, CI -3.0 to -0.7; top-10 residence +2.6% per year, +0.8 to +4.5). With rel >= 0.25 (8-15% of top-100 article-days removed), none of 8 trends excluded zero (turnover -0.6%, -1.4 to +0.2). Removal of only transient flags (one-day, low-app spikes: 8.9% of top-100 article-days in 2016, 3-4% after 2020) removed the slowdown. Removal of only persistent flags did not. In the peak series, the next-day fraction of peak excess increased from 0.33 to 0.40 (difference +0.07, CI +0.02 to +0.12), but effective duration did not change (3.21 to 3.42 days, CI of difference -0.14 to +0.66). German turnover and losses decreased slightly. Limitations: the filter cannot separate automated traffic from human bursts that arrive through desktop or mobile-web links, such as Google Doodles. Thus the English list slowdown is either an artefact of automated traffic or a change in referral traffic. Scope: the top 1,000 articles per day, Wikimedia user traffic, and two language editions.

collective attention · Wikipedia · pageviews · social acceleration · automated traffic · bot filtering · popularity dynamics · turnover · attention decay · replication · computational social science