When a pregnant person asks an AI chatbot for advice, the response sounds empathetic and balanced, and the links look legitimate. A cross-country investigation by AlgorithmWatch found otherwise.

According to the investigation, chatbots frequently link to anti-abortion organizations when answering questions about unplanned pregnancy, without disclosing those groups' ideological stance. In at least one out of every four queries, a response included a link to an anti-abortion website, drawing no distinction between official health authorities and ideologically driven organizations.

How it was tested

  • Models — ChatGPT-5, Gemini 3, Grok 4.3 and Claude Sonnet 4.8.
  • Personas — three fictional profiles facing unplanned pregnancies.
  • Languages — English, German and Italian.
  • Sample — 270 responses produced for analysis.

A source that looks credible but is not neutral

The organization Profemina appeared in about 17 percent of all responses. According to AlgorithmWatch, the group has ties to Heartbeat International, one of the oldest anti-abortion organizations in the United States. In most conversations the chatbots did not flag this connection. Only when pressed did ChatGPT acknowledge that Profemina is "not a neutral medical source" and "has a stance aimed at discouraging abortion or putting it under strong moral scrutiny."

The inconsistency becomes vivid in one example: Gemini linked to Profemina five times in a single conversation, shared a graphic from the group's website, and then warned about that very same source later in the same chat. Claude behaved similarly, issuing warnings and linking to the organization at the same time.

In 33 percent of German-language and 29 percent of Italian-language queries asking for personal experiences, the chatbots linked to Profemina. The organization did not respond to requests for comment.

Pointed to the wrong door

In eleven out of twelve German-language conversations, the chatbots recommended Caritas for pregnancy counseling. But Caritas does not issue the counseling certificate that German law requires before a legal abortion within the first twelve weeks. Someone who goes to Caritas first and only then learns the certificate will not be issued could lose critical time.

ChatGPT called Caritas a "recognized" organization for mandatory counseling, listing it alongside Pro Familia as an equal option. AlgorithmWatch suspects Caritas's large digital footprint, with roughly 300 diocesan counseling centers across Germany, may be why the models keep surfacing it.

Google and OpenAI pointed to their policy guidelines, which do not specifically address abortion-related queries. An OpenAI spokesperson said GPT-5.5, which shipped after the testing period, is "the best so far at taking user context, such as gender, into account." Anthropic and xAI did not respond.

Why it matters

The finding is not confined to abortion. AI-generated answers, including those with search built in, are black boxes shaped by training data, cultural weightings and outside manipulation.

The real issue is form. The empathetic tone and apparent completeness make a response feel trustworthy to a non-expert. Research shows AI-generated answers dramatically cut clicks through to the underlying sources, meaning users rarely question what they are told or look at who said it.

That is what separates a chatbot from a search result. A search page shows the list of sources and the reader chooses which to open. In a chatbot the choice arrives already made, inside the model, with its reasoning invisible.