I do not use large language model AI chatbots to generate indexes, nor do I upload or share client documents with AIs without express permission.
These AIs fail at the indexer’s primary task: to ensure readers find needed information. Research I have done as part of American Society of Indexing's AI Committee show that AI chatbots typically provide inadequate and inappropriate structure and cross-references, both under- and over-index material, and produce inaccurate indexes that require significant human remediation. Each of these failings is problematic:
- Inadequate or inappropriate index structure, especially with regards to cross-references, prevents readers from effectively navigating to subtopics and related topics while misrepresenting the focus of the book
- Under- and over-indexing produces bloated indexes that omit significant content. Over-indexing increases printing costs and requires readers to wade through irrelevant material to reach desired content; under-indexing prevents readers from locating information and misdirects readers into thinking information isn’t included in the book
- Accuracy issues—including failure to pass a standard test for page reference accuracy and an extremely high rate of issues that require manual remediation— waste readers’ time and breaks the trust between readers and the book
The human brain of a professional indexer is still the best tool for analyzing, writing, and editing an index.
I do use AI when developing software utilities, as well as sometimes when drafting and troubleshooting regular expressions and macros, which I use for text replacement and automation of repetitive tasks. In none of these cases does the AI have contact with the index, and I carefully review and test all AI-generated code/regular expressions/macros to ensure they meet my standards.