AI Petition Writing Tools Failed to Boost Success Rates
A study published September 30, 2026, found AI assistance reduced engagement by 5% on Change.org petitions.
Updated on Sept. 30, 2026 in Artificial Intelligence
Live Poll
Do you believe using AI to draft messages makes your communication more or less effective?
Cornell researchers analyzed 1.5 million Change.org petitions and found that AI-assisted writing failed to improve success rates compared to human-written versions. The study examined data from a tool rollout occurring between October 2 and December 15, 2023.
Why it matters
The findings suggest that AI-generated language often lacks the specific detail and personal commitment required to drive meaningful public action. This research highlights a potential disconnect between automated content creation and the social dynamics of digital advocacy.
The study analyzed 1.5 million petitions using metrics of lexical diversity, readability, and text length. Petitions written with AI assistance consistently displayed greater language homogeneity and increased length compared to human-written counterparts.
The players
Cornell University
A research institution focused on data science and social behavior analysis.
Change.org
A global petition platform facilitating online activism and social engagement campaigns.
The details
Researchers at Cornell University measured the impact of AI tools by leveraging a staggered rollout across the U.S., Great Britain, Canada, and Australia, where Australia received the software 11 weeks after the other nations. They assessed stylistic changes by evaluating lexical diversity—the range of unique words used—and the readability of the petitions. The authors hypothesized that the resulting decrease in engagement occurred because the AI-generated text lacked the necessary nuance and authentic detail that typically captures public interest.
Timeline
Oct. 2, 2023 - Dec. 15, 2023: The AI tool rollout and primary data measurement period.
Sept. 30, 2026: Publication of the research in Nature Human Behaviour.
The Tech Race
This research complicates the broader push to integrate generative AI into every facet of digital communication and advocacy. It establishes a necessary benchmark for measuring the efficacy of automated tools in high-stakes social interactions versus purely commercial content generation.
Campaign organizers using AI writing assistants should be aware that automated content may inadvertently alienate readers due to lack of specificity. The findings suggest that adding personal, human-led detail remains a more effective driver for engagement than relying on AI-generated text.
The takeaway
Advocacy groups should prioritize human-driven specificity to maintain audience trust and engagement in digital spaces. Practitioners can watch for future studies in Nature Human Behaviour to see if adaptive AI prompting can overcome the homogeneity issues identified in this 2026 analysis.
Further reading
Explore more analysis regarding the performance and societal impact of large language models in our Artificial Intelligence section.
Live Poll
Do you believe using AI to draft messages makes your communication more or less effective?
