The quote landed on a Tuesday afternoon. My editor had connected me with a freelance indexer she'd worked with for years — experienced, credentialed, highly regarded. The estimate: $975 for my 270-page business book. Four-week turnaround. Payment in advance.
I closed the email. I sat with it for a while.
If you're a nonfiction author trying to figure out what indexing your book will actually cost in 2026 — and whether you have to spend that much — this is the breakdown I needed and couldn't find. No vague ranges, no affiliate disclaimers. Just an honest side-by-side of every realistic option, with real numbers and clear guidance on who each one is right for.
What a Professional Indexer Actually Charges
The short answer: more than most authors expect, with a range wider than most pricing guides admit.
As of 2026, the Society of Indexers (UK) recommends rates of £3.80 per page for non-academic text — roughly $4.80 at current exchange rates. American indexers typically work in the $4–$8 per page range, according to industry pricing guides. Run those numbers on a typical nonfiction book:
| Book length | Low estimate ($4/page) | High estimate ($8/page) |
| 150 pages | $600 | $1,200 |
| 250 pages | $1,000 | $2,000 |
| 350 pages | $1,400 | $2,800 |
That's before complexity adjustments. Academic texts, legal books, and technical manuals with dense terminology typically land at the high end — or above it. A professional index for a serious 300-page reference book can comfortably run $1,500–$2,400.
For context: Reedsy's 2026 publishing cost data puts a full self-publishing production budget at $2,000–$5,000 including editing, cover design, and formatting. Spending $1,000–$2,000 of that on indexing alone — a single component of back matter — is a decision that deserves scrutiny.
There's also the timeline problem. Professional indexers require your final, typeset PDF before they can begin. You can't start the clock until everything else is locked. At an industry-standard pace of 8–10 pages per hour, a 250-page book represents 25–35 hours of work. Most indexers quote two to four weeks minimum, and many have waitlists.
If your print submission window is tight, this timing can be genuinely impossible to accommodate — regardless of what you're willing to pay.
The DIY Option — What It Really Costs You
Two paths exist for DIY indexing: traditional indexing software (Cindex became free in 2023; Macrex and SKY Index remain paid), or Microsoft Word's built-in index tools.
The financial cost is low. Cindex is now a free download. Word is already on your machine. But the real cost is time — and most authors dramatically underestimate it.
Here's what the data shows:
- Professional indexers — people who have done this for years — clear roughly 8–10 pages per hour. A 250-page book takes them 25–35 hours.
- Self-publishing authors report a much wider range. One experienced author documented 5–20 hours on their own projects. Others recommend budgeting three to four full weeks for a first attempt.
The gap reflects the learning curve. Cindex requires training to use effectively; it's built for professional indexers, not for authors who index once every few years. Word's indexing tools are more accessible but more limited — they can build a concordance (a keyword list with page numbers), but they can't generate the hierarchical entries, subentries, cross-references, and double-posting logic that make an index genuinely useful to readers.
The other hidden cost: judgment calls. Every few pages you'll face decisions — does this term get its own entry, or become a subentry under something broader? What do you do with a concept that appears under three different phrasings throughout the manuscript? Do these two tangentially related terms need a see also cross-reference? Professional indexers make these calls intuitively. Authors doing it themselves spend real time on each one.
DIY works well for authors who have indexed before, are indexing a short and conceptually simple book, and have 30+ hours of patience and focus available. For everyone else, the math starts to look less appealing quickly.
Using ChatGPT or Claude: What Actually Happens
The instinct makes sense. AI is embedded in author workflows in 2026, it's fast, and it can process large amounts of text. Why not indexing?
The honest answer is that general-purpose AI tools produce something that looks like an index — and that's the problem.
What they generate is typically keyword extraction with guessed page numbers: a list of terms that appear frequently, with locations assigned based on the model's best approximation of where it encountered them in the text. The structural problems are predictable:
- Hallucinated page numbers: LLMs don't inherently understand pagination. They can guess, but they cannot verify against your actual typeset file.
- Missing cross-references: Linking "revenue recognition" to "GAAP compliance" via a see also requires domain understanding general AI doesn't reliably apply.
- No subentry logic: Grouping twenty page references for "leadership" into meaningful subentries — "leadership, situational," "leadership, transformational," "leadership, in crisis" — requires editorial judgment that keyword extraction doesn't provide.
- Token limits: A full 280-page manuscript is a lot of text. Most general LLMs truncate when processing complete books, which creates gaps and hallucinations in the middle sections.
The editing time required to fix these problems is substantial. Several authors have documented spending 6–10 hours cleaning up a ChatGPT-generated index on a 250-page book. At that point, you've spent more time than many DIY approaches and produced a less reliable result.
This is not a general knock on large language models — they're extraordinary tools for many parts of the writing and publishing process. They're just not built for this specific task.
This category barely existed three years ago. Today there are several tools built specifically for manuscript indexing that approach the problem differently from general AI: they process the complete uploaded PDF, identify page-accurate locations, recognize indexable entities in context, and produce hierarchical entries that follow CMOS or Chicago Manual conventions.
The practical difference matters. A purpose-built indexing tool:
- Processes your entire typeset PDF without truncation — the same file you'd send to a professional indexer
- Identifies entity page numbers accurately from the document structure, not from probabilistic guessing
- Surfaces people, concepts, organizations, frameworks, and key terms with context around each occurrence
- Groups variants and synonyms (so "machine learning," "ML," and "machine-learning" become a single entry)
- Generates hierarchical entries with subentries and cross-references
- Presents everything for human review before finalizing — so you catch errors before they go to print
The result is a first-pass index that's substantially more accurate than anything a general AI model produces, and it compresses what would be 25–35 professional indexer hours into a workflow that most authors complete in under two hours — including their own review pass.
Tools in this category include Onomastic, IndexStudio, and Indexia. Pricing varies by tool and book length, but typically lands in the $25–$80 per book range — a fraction of professional indexer rates for most manuscripts.
The Full Comparison
Here's every option in the same table:
| Option | Cost (250-page book) | Turnaround | Quality | Author time |
| Professional indexer | $1,000–$2,000 | 2–4 weeks | Highest | Minimal |
| DIY (Cindex/Word) | Free | 30–50+ hours | Variable | Very high |
| General AI (ChatGPT/Claude) | Free–$20/month | 4–8 hours (inc. edits) | Inconsistent | High |
| Purpose-built AI tool | $25–$80 | 1–3 hours | Good to high | Low–medium |
A few things the table doesn't fully capture:
Price and quality don't scale linearly. A professional indexer charging $1,800 will almost certainly produce a better index than any software on the market today. For academic press publications, legal references, or any book where indexing quality is formally evaluated and part of the author's contractual obligations — that premium is worth paying.
Time is often the binding constraint, not money. If your print submission deadline is ten days away, no amount of money will get a professional indexer to turn around a 250-page manuscript faster than the queue allows. This is where purpose-built AI solves a real problem that money alone cannot.
The "professional quality" bar for most trade books is achievable with AI. For a business book, a self-help title, a parenting guide, a health and wellness book — the practical standard is an index that serves readers well, passes editorial review, meets library and retailer standards, and doesn't embarrass the author. A purpose-built AI tool with a thorough human review pass consistently clears that bar.
Budget reality check: 46% of indie authors earn $100 or less per month from their writing, according to 2026 self-publishing surveys. Spending $1,500 on indexing alone — one line item in a $3,000–$5,000 production budget — is a meaningful financial choice when an alternative exists at $50–$80 and two hours of your time.
Who Should Use Which Option
Here's the clearest way I can frame it:
Hire a professional indexer if you're publishing an academic textbook, legal reference, or technical manual where index quality is formally evaluated — or if your publisher requires professional indexer credentials by contract.
DIY if you've indexed your own books before, your manuscript is under 150 pages, and you have 30+ hours and a high tolerance for tedious precision work.
Skip general AI for indexing specifically. It's the wrong tool for this job, even if it's the right tool for many others in your workflow.
Use a purpose-built AI tool if you need professional-quality output on a real production timeline at a realistic budget — which describes the large majority of self-published nonfiction authors in 2026.
The math has genuinely shifted. A focused 90-minute workflow — upload your PDF, review the extracted entries, make your editorial calls, export — can produce a 300+ entry index that a publisher marks accepted without revision. That's not the exception. That's what purpose-built indexing tools are designed to deliver consistently.
Try Onomastic free and run your manuscript through it. See what it produces. You'll have real, indexed output to compare against every other option — and you'll know within an hour whether the quality clears your bar.
