Popular AI chatbots repeatedly linked people asking about unplanned pregnancies to anti-abortion organizations without clearly identifying their position, according to an AlgorithmWatch investigation. Researchers tested ChatGPT, Gemini, Grok and Claude with three fictional personas in English, German and Italian, collecting 270 responses.
At least one in four queries included an anti-abortion website. Profemina appeared in about 17% of all answers, often without disclosure of its links or viewpoint. Some systems later warned about a source they had already recommended. In German-language tests, eleven of twelve conversations suggested Caritas for counseling even though it does not issue the certificate required before many legal abortions in Germany.
This was a structured investigation, not a measurement of every response those products generate. Still, pregnancy guidance is time-sensitive, and a plausible but ideologically undisclosed or procedurally wrong referral can delay care. Providers need stronger source labeling, country-specific verification and tests that examine links, not only the prose of an answer. Users should confirm medical and legal guidance with official health services.