# Manipulating Twitter Through Deletions

arXiv: https://arxiv.org/abs/2203.13893 · HTML: https://mlsanity.com/paper/2203.13893/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 25 Mar 2022.

## Built on by (most cited first, 6 of 6)

- [Embedding Democratic Values into Social Media AIs via Societal Objective Functions](https://mlsanity.com/paper/2307.13912/) (2023)
- [Demystifying Misconceptions in Social Bots Research](https://mlsanity.com/paper/2303.17251/) (2023)
- [Detecting and Reasoning of Deleted Tweets before they are Posted](https://mlsanity.com/paper/2305.04927/) (2023)
- [Preemptive Detection of Fake Accounts on Social Networks via Multi-Class Preferential Attachment Classifiers](https://mlsanity.com/paper/2308.05353/) (2023)
- [Bots Don't Sit Still: A Longitudinal Study of Bot Behaviour Change, Temporal Drift, and Feature-Structure Evolution](https://mlsanity.com/paper/2512.17067/) (2025)
- [Characterising Behavioural Families and Dynamics of Promotional Twitter Bots via Sequence-Based Modelling](https://mlsanity.com/paper/2512.18077/) (2025)

## Abstract (arXiv)

Research into influence campaigns on Twitter has mostly relied on identifying malicious activities from tweets obtained via public APIs. These APIs provide access to public tweets that have not been deleted. However, bad actors can delete content strategically to manipulate the system. Unfortunately, estimates based on publicly available Twitter data underestimate the true deletion volume. Here, we provide the first exhaustive, large-scale analysis of anomalous deletion patterns involving more than a billion deletions by over 11 million accounts. We find that a small fraction of accounts delete a large number of tweets daily. We also uncover two abusive behaviors that exploit deletions. First, limits on tweet volume are circumvented, allowing certain accounts to flood the network with over 26 thousand daily tweets. Second, coordinated networks of accounts engage in repetitive likes and unlikes of content that is eventually deleted, which can manipulate ranking algorithms. These kinds of abuse can be exploited to amplify content and inflate popularity, while evading detection. Our study provides platforms and researchers with new methods for identifying social media abuse.