This answer pertains to the creation of index entries and not to the mechanics of turning them into elements in a word processing or page composition program that will automatically generate an index. (Generated indexes still need to be checked by a human being.)
Some authors do their own indexing, some can’t or won’t, and some publishers hire their own indexer anyway even if the author has presented an index with the manuscript.
I’ve done indexing professionally as an adjunct to editing jobs. I actually enjoyed doing it, although some of my editorial colleagues thought that was nuts. I guess it takes a certain kind of mind.
You have to work to a set of guidelines, either given you by the publisher or adopted by you; for example:
• How deep do you want to go? Will it be thorough and comprehensive or just a light culling of major topics?
• How many levels of topics and subtopics will you allow? Two is typical; four is really deep, and I don’t think I’ve ever seen more than that.
• Will you permit cross-references? If so, how will you handle them? What about “see also”?
• Are names really important in your field? (Is everybody in your field going to want to look and see if their name is mentioned?)
Some considerations include how scholarly the book is and how many levels of heads it has. An index can’t just be an alphabetized TOC. Is there a bibliography? Is there a glossary? Are there notes or chapter notes? Do you have a rule of thumb to follow, such as one page of index for every ten pages of text? every fifty?
One of the things I routinely asked myself while going through the text was “When this is the piece of information I want, what am I going to be looking up?” It isn’t always the same term or the same form of the term that appears in the particular passage of text; maybe instead there’s an umbrella concept—or a term used as an alternative to another term—that should apply to all related entries.
It’s not just about terms, and you probably don’t want a page reference for every use of a given term. That’s part of what the human editor looks for.
There are style questions to decide, such as use of punctuation, display of ranges of numbers, use of capitalization, and treatment of acronyms and their expansions. Also be sure which rules of alphabetizing you’re following: word by word or letter by letter. Does “fishing” come before or after “fish sticks”? Whichever rule you use, follow it consistently.
I’ve done indexing directly on page proofs with a pencil and a highlighter. In a workplace setting where someone else is creating the entries and inputting the tokens, writing on the page proofs lets the inputter know where the tokens go.
How this may all have changed with the technology in the last eight or ten years, I don’t know, but the logic ought to be about the same.
Even now, I often annotate indexes as I’m reading. The same question—when I want to find this part, what will I be looking up?—often yields nothing, and so I pencil it in.