# The Half-Life of a Tweet

arXiv: https://arxiv.org/abs/2302.09654 · HTML: https://mlsanity.com/paper/2302.09654/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 19 Feb 2023.

## Built on by (most cited first, 4 of 4)

- [Facts are Harder Than Opinions -- A Multilingual, Comparative Analysis of LLM-Based Fact-Checking Reliability](https://mlsanity.com/paper/2506.03655/) (2025)
- [Inductive detection of Influence Operations via Graph Learning](https://mlsanity.com/paper/2305.16544/) (2023)
- [Beyond Content: Behavioral Policies Reveal Actors in Information Operations](https://mlsanity.com/paper/2602.02838/) (2026)
- [Language Mutations Sustain the Persistences of Conspiracy Theories on Social Media](https://mlsanity.com/paper/2605.20050/) (2026)

## Abstract (arXiv)

Twitter has started to share an impression_count variable as part of the available public metrics for every Tweet collected with Twitter's APIs. With the information about how often a particular Tweet has been shown to Twitter users at the time of data collection, we can learn important insights about the dissemination process of a Tweet by measuring its impression count repeatedly over time. With our preliminary analysis, we can show that on average the peak of impressions per second is 72 seconds after a Tweet was sent and that after 24 hours, no relevant number of impressions can be observed for ~95% of all Tweets. Finally, we estimate that the median half-life of a Tweet, i.e. the time it takes before half of all impressions are created, is about 80 minutes.