Change.org 'write with AI' changed petitions, not outcomes

A natural experiment on Change.org covering 1.5 million petitions found that giving users an in-platform AI writing tool reshaped what petitions looked like but not whether they succeeded.

The paper, by Isabel Corpus, Eric Gilbert, Allison Koenecke and Mor Naaman and published in Nature Human Behaviour, uses a difference-in-differences design around Change.org's rollout of an in-platform 'write with AI' feature. The authors report that AI access 'substantially altered the lexical features of petitions and increased petition homogeneity but did not improve petition outcomes.'

To check the platform-level pattern against individual behavior, the team isolated 4,611 repeat petition writers who posted both before and after the tool existed. Those writers, per the paper's Figure 3 caption, 'write longer petitions with worse outcomes when their second petition is written with access to AI.'

The abstract states the broader read plainly: 'while AI writing tools can profoundly reshape online content, their practical utility for improving desired outcomes may be less beneficial than anticipated and introduce unintended consequences such as content homogenization.'

The published abstract reports no exact tool launch date, no per-cause breakdown, and no signature counts for the compared petitions. The lift that did not appear is what the paper is built around.