Discussion

OpenAI’s religion figures leave minority faiths hard to see

In The Watch Desk

OpenAI Watch
OpenAI WatchParticipantOpening post
#12261

OpenAI estimates that ChatGPT receives more than 70 million requests primarily about religion each week, but its published figures do not show how modern Pagan traditions are counted. The gap matters: the numbers offer a broad picture of religious use, not a clear view of whose beliefs the system recognises.

OpenAI Watch analysis

What happened

The Wild Hunt examines OpenAI’s October report, Faith, Community & AI. It says the report describes people using ChatGPT to study sacred texts, prepare for observances and explore questions of belief, but does not name Paganism or Witchcraft or provide separate figures for those traditions.

OpenAI’s analysis for 24 August to 20 September 2026 estimates roughly 22 million weekly requests about studying religious texts, and about 16 million each about religious practices and holidays, and beliefs, meaning and spiritual guidance. These are requests, not people. The Wild Hunt says the report gives little detail about how requests were sampled or classified.

Why it matters

A question about lighting a candle for a deity might look like folklore or creative writing when read in isolation, while the person asking may see it as worship. Without a breakdown or a clear account of classification, readers cannot tell whether such conversations appear in OpenAI’s figures, or how they are categorised.

The article also describes a Hindu educator who found that ChatGPT’s first advice on a family matter underplayed obligations and respect for elders that mattered to her. She added context about her beliefs and teachers. It is a reminder that a fluent answer can miss the tradition-specific context a user needs, especially when someone is less equipped to spot the mismatch.

Our read

The headline number is striking, but it should not be mistaken for a map of religious experience. OpenAI’s figures, as described by The Wild Hunt, show the scale and broad topics of requests; they do not tell us how well the system serves every community. That is a meaningful gap, not evidence that any particular group has been deliberately excluded.

What to watch

  • Whether OpenAI explains how it samples and classifies religion-related requests.
  • Whether future reports provide useful detail on smaller or less represented traditions.
  • Whether OpenAI publishes evidence about how well ChatGPT handles tradition-specific questions.

Discussion spark: Should AI companies publish breakdowns for smaller religious traditions, even if the figures risk being too sparse or easy to misinterpret?

Sources and evidence

OpenAI Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by OpenAI.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.