Haaretz reported on October 25, 2026, that old, alarming posts resurfaced in recommendation feeds during a nearly four-month study of six fictitious Israeli personas. The investigation estimated hundreds of such posts, using the 13,000 of 80,000 logged posts whose dates it could establish.
An old alert can look current
Haaretz described an Israel Defense Forces missile announcement that appeared in an Instagram feed two weeks after it was posted, on TikTok three weeks later, and in another feed 90 days after publication. The video had no visible date, so viewers could encounter it without knowing when the announcement was made.
What the study counted
The six fictitious personas all had Facebook and TikTok accounts; four also had Instagram accounts. Haaretz said an AI system logged posts over nearly four months and that the investigation manually checked the system’s findings. It counted a post as recurring if it appeared at least 14 days after its original publication.
Haaretz estimated hundreds of alarming posts had resurfaced, extrapolating from posts with established dates. It also said most resurfaced posts were innocuous; most of the alarming examples appeared on Facebook and TikTok.
Dates were established for 16.25% of logged posts
Haaretz established dates for 13,000 of the 80,000 logged posts, or 16.25%. The remaining approximately 67,000 had no established date in that process. The 16.25% describes date coverage, not the share of posts that recurred. The estimate of hundreds therefore rests on the dateable subset.
TikTok said its recommendations reflect users’ preferences and interactions, and called conclusions drawn from a handful of subjectively observed accounts methodologically flawed, Haaretz reported.
