American Society for Indexing statement on "AI" in book indexing

As part of ASI’s continuing commitment to excellence in indexing, ASI has formed a committee to investigate potential impacts of AI on indexing.

The committee’s tests indicate that AI tools based on large language models (LLMs), such as ChatGPT and Claude, do not produce viable book indexes, either in whole or in part. The resulting indexes fail in terms of completeness (providing the reader access to all substantive information), navigability (guiding the reader to subtopics and related topics), and accuracy (containing only correct information and reflecting the author’s terminology). They require extensive review and correction by a human, which is almost always more time-consuming and costly than working with a professional indexer from the start.

We recognize the desire to have good indexes quickly and cheaply. However, AI is not currently capable of producing even adequate indexes. High-quality indexes that meet reader needs and expectations still require the fine-grained human judgment of a professional indexer.

Thus, at this time, we recommend AI-generated indexes not be used in books.

See more