Short answer. AI is most useful for the parts of book marketing that are research-heavy and template-shaped: building a launch plan, writing your metadata and back-cover copy, scouting keywords and categories, and finding reviewers to pitch. It is least useful, and actively risky, for the parts that depend on genuine human relationships, like the actual outreach to a reviewer or the voice of your newsletter. Use it for the scaffolding and the first drafts; keep the relationships human. WriteLoom's Market and Sell studios are organized around this split.

Start with a plan and a budget

Most self-published launches fail not from bad books but from no plan. Before any tactic, you need a timeline (pre-order, launch week, post-launch) and a budget you won't blow. AI is good at turning "I have 200 dollars and six weeks" into a concrete, dated plan with line items. WriteLoom's Market studio includes a marketing plan and a line-item budget calendar so the launch has a shape before you spend a dollar. This part is available without AI at all on the Spool tier; it's a structured planner first.

Write the metadata that sells

For a self-published book, your metadata is your storefront: title, subtitle, back-cover/description copy, keywords, and categories. This is conventional, high-leverage writing that AI drafts well.

• Back-cover copy needs to hook a browser in three sentences. WriteLoom's Sell studio drafts back-cover copy from your project.
• Keywords and categories determine whether anyone finds the book on retailers. WriteLoom's keyword scout suggests keywords; dedicated tools like Publisher Rocket go deeper on Amazon-specific category and competition data.
• The one-pager (for reviewers, bookstores, or press) can be generated from your project rather than rebuilt from scratch.

Edit all of it. AI metadata is a strong first draft, but the description especially should carry your book's actual tone.

Find reviewers, then pitch them like a human

Reviews are the engine of discovery for indie books, and finding the right reviewers is a research slog AI shortens dramatically. WriteLoom's reviewer finder surfaces named book bloggers, BookTok creators, Goodreads reviewers, and indie-press contacts filtered to your genre, with contact info and outreach templates.

Here's the line not to cross: use AI to find reviewers and to draft the template, but personalize every message you actually send. Reviewers can tell a mass-blasted AI pitch from a real one, and the mass blast burns the relationship permanently. A short, specific note that references a review they actually wrote outperforms a polished generic one every time.

What to keep human

A realistic sequence

A self-published launch using AI well looks roughly like this: build the plan and budget six to eight weeks out; generate and then heavily edit your metadata and back-cover copy; scout keywords and categories; use the reviewer finder to build a shortlist; send personalized pitches with an ARC two to four weeks before launch; and schedule announcement copy you've rewritten in your own voice. The AI removes maybe 70% of the busywork. The 30% it can't touch, the relationships and the voice, is the part that actually moves copies, so that's where your time goes.

The honest caveat

No tool markets a book on its own, and any service promising hands-off "AI marketing" is selling automation that readers can smell. The point of AI here is to give you back the hours you'd lose to spreadsheets and blank pages, so you can spend them on the human work that actually sells books.