Skip to content
AlgorithmWatch·Indexed by this site · Originally published

AlgorithmWatch tests four chatbots: anti-abortion websites treated as authoritative sources in abortion consultations

How Popular AI Chatbots Recommend Pro-life Websites to Pregnant People

AI summary

AlgorithmWatch's cross-national investigation in Italy, Germany, and the UK posed 270 sets of abortion-related questions in three languages and three personas to ChatGPT-5, Gemini 3, Grok 4.3, and Claude Sonnet 4.8, finding that all four models commonly cited official health institutions alongside advocacy organizations and forums, with at least a quarter of responses containing links to anti-abortion websites. Profemina, a counseling organization linked to Heartbeat International, appeared in about 17% of responses, rising to 33% and 29% respectively when asked about personal experiences in German and Italian, and in most cases the models did not indicate its anti-abortion stance. In the German scenario, models often recommended Caritas, which does not issue mandatory counseling certificates, potentially affecting the process within the 12-week limit. Researchers noted that advocacy organizations heavily deploy niche content through SEO and AI optimization, causing models to identify them as professional sources, while public institutions and pro-choice parties have a weak online presence.

Read originalalgorithmwatch.org