{"paper":{"title":"Prosocial Persuasion at Scale? Large Language Models Outperform Humans in Donation Appeals Across Levels of Personalization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"LLM-generated donation appeals produced more donations, higher engagement, and stronger persuasiveness ratings than human-written ones in two experiments.","cross_cats":[],"primary_cat":"cs.CY","authors_text":"Bennett Kleinberg, John Caffier, Olga Stavrova","submitted_at":"2026-04-03T17:25:07Z","abstract_excerpt":"Large Language Models (LLMs) are increasingly regarded as having the potential to generate persuasive content at scale. While previous studies have focused on the risks associated with LLM-generated misinformation, the role of LLMs in enabling prosocial persuasion is still underexplored. We investigate whether donation appeals authored by LLMs are as effective as those written by humans across degrees of personalization. Two preregistered online experiments (Study 1: N = 658; Study 2: N = 642) manipulated Personalization (generic vs. personalized vs. falsely personalized) and Content source (h"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"In both experiments, LLM-generated content yielded more donations, resulted in higher engagement, and was rated as more persuasive than human-authored content.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That distributing a small bonus in an online experiment accurately measures real-world charitable donation behavior and that human and LLM content were produced under equivalent effort and quality constraints.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"LLM-generated donation appeals outperform human-written ones in driving donations, engagement, and perceived persuasiveness across generic, personalized, and falsely personalized conditions.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"LLM-generated donation appeals produced more donations, higher engagement, and stronger persuasiveness ratings than human-written ones in two experiments.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"8e2e14cfe0c84b41f22e6814d3f3d37a013fb759a8ce7e093969fe60c7cf0a83"},"source":{"id":"2604.03202","kind":"arxiv","version":2},"verdict":{"id":"af199e24-aeb1-45f8-a914-2a6ee6b69d75","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-13T17:57:12.351188Z","strongest_claim":"In both experiments, LLM-generated content yielded more donations, resulted in higher engagement, and was rated as more persuasive than human-authored content.","one_line_summary":"LLM-generated donation appeals outperform human-written ones in driving donations, engagement, and perceived persuasiveness across generic, personalized, and falsely personalized conditions.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That distributing a small bonus in an online experiment accurately measures real-world charitable donation behavior and that human and LLM content were produced under equivalent effort and quality constraints.","pith_extraction_headline":"LLM-generated donation appeals produced more donations, higher engagement, and stronger persuasiveness ratings than human-written ones in two experiments."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.03202/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}