AI tool reshapes Change.org petitions but doesn't lift outcomes

A new paper in Nature Human Behaviour studied what happened when Change.org added an in-platform 'write with AI' tool, using 1.5 million petitions and a difference-in-differences design. The authors — Isabel Corpus, Eric Gilbert, Allison Koenecke and Mor Naaman — report that the tool changed how petitions read without changing whether they worked.

"In-platform AI access substantially altered the lexical features of petitions and increased petition homogeneity but did not improve petition outcomes," the paper states. The result held in a secondary analysis of repeat writers who posted petitions both before and after the tool arrived.

The framing in the abstract is blunt about the trade: "AI writing tools can profoundly reshape online content," the authors write, while warning that "their practical utility for improving desired outcomes may be less beneficial than anticipated and introduce unintended consequences such as content homogenization."

The abstract does not break out which petition categories moved and which did not, and does not spell out which outcome metric the authors treated as the dependent variable. It was circulating among a couple of the researchers we track in our Who's Who directory within a day of publication, which is the kind of signal that tends to precede the Twitter-thread cycle on an AI-and-society result.