AI is Losing India's Storytelling Magic—Here's How

If you’ve ever asked an AI to tell you a story, you might not realize it’s probably blending Indian folklore into a generic, Western-sounding tale—even when you ask in Hindi, Tamil, or Bengali. A new research paper tested two popular AI models (Claude Sonnet and Gemini) on three distinctly Indian storytelling traditions: the Rajasthani Pabuji epic, classical Tamil Sangam poetry, and Bengali folk tales. The study found that while AI outputs stayed closer to their own tradition than others, they still became surprisingly similar across traditions (like a tech company’s generic “inspirational” poster). Even worse, prompting in regional languages *reduced* accuracy by up to 27% compared to English, making the AI’s versions feel more distant from the real thing.

Why does this matter outside of academia? Because AI is increasingly used to generate content for education, tourism, and even cultural preservation. If AI can’t reliably preserve the nuances of India’s oral traditions—stories passed down for centuries—it risks turning diverse, living cultures into a single, stale script. The study’s author suggests this is like translating a spicy Rajasthani folk song into a bland pop tune: it loses its soul.

The catch? This was a small pilot study with just 42 passages and 54 prompts. It doesn’t mean all AI storytelling is broken—but it’s a red flag for anyone using AI to document or share cultural heritage. The researcher calls for bigger studies and better training data, but for now, if you’re relying on AI to teach your kids about Indian folklore, you might want to double-check its work.