At least it's not AI... yet.
Multi-modal LLMs like Gemini are better than traditional OCR in most ways.