This is an erratum to the research article "How do social media feed algorithms influence election campaign attitudes and behavior?" published in Science, Volume 391, Issue 6791, March 2026.[1] The Erratum clarifies that the introduction of 63 "break-glass" changes to the Facebook Meta algorithm during the experiment did not affect its internal validity, that is, the ability to draw accurate causal conclusions from the data.[1] Experiment by Guess et al. (2023) ran from September 24 to December 23, 2020 and compared a reverse chronological feed with an algorithmic Facebook feed during the 2020 US election campaign and aftermath.[1] These temporary changes were intended to reduce the visibility of content from untrusted sources after the election, thereby altering the control condition of the experiment, but did not affect its internal validity.[1] In the discussion, the authors emphasized that the algorithms of Facebook and Instagram strongly influenced the user experience on social networks at the end of 2020.[1] They cautioned that the results could be different with different algorithms or outside of a polarized election campaign.[1] The erratum confirms the cumulative causal effect of switching to a reverse chronological feed versus the algorithm over the entire study period.[1]