Most authors using AI start by opening Claude or ChatGPT, type "help me outline a cozy mystery," and start building. Six weeks later they have a finished novel that reads fine, looks fine, uploads to KDP fine — and then sits there. No traction. No reviews. They blame the cover, the blurb, the algorithm.

The actual problem happened on day one, before a single chapter was outlined.

They skipped premise validation.

I've built 200+ AI-assisted novel packages across romance, cozy mystery, suspense, cozy fantasy, and a handful of other subgenres. The single biggest pattern I've seen separating books that sell from books that don't is whether the author validated the premise against the actual market before they let AI start generating anything.

Here's what premise validation actually looks like:

1. Know what's selling in your subgenre right now — not what sold three years ago. The tropes that dominated cozy mystery in 2023 are not the tropes dominating cozy mystery in 2026. Bookstagram trends shift. Reader appetites shift. If you outline based on what you loved reading five years ago, you're writing for a market that's already moved on. Spend an afternoon on the top 100 in your subgenre. Read blurbs. Note repeated tropes, settings, hook structures, and what's conspicuously absent. That absence is sometimes where the opportunity is.

2. Pressure-test your premise against the trope stack. A premise isn't "amateur sleuth solves murder in small town." That's a setup. A premise is the specific trope combination that gives readers a reason to one-click. "Bakery owner amateur sleuth + small town she returned to after divorce + murder happens at the town festival she's catering" is a stack. Each layer adds reader pull. If your premise only has one trope doing the work, AI will faithfully outline a book that has no commercial hook.

3. Validate before you prompt, not after. This is the one that hurts. Once you've told AI to start generating, it'll generate. It doesn't know whether your premise has a market. It doesn't care that "cozy mystery set in a yarn shop" has 400 competitors and "cozy mystery set in a struggling drive-in theater" has 12. Your job is to bring the market intelligence to the prompt. AI is a force multiplier for whatever direction you point it in — including the wrong one.

4. Write the back-cover copy first. Before any outline, write the blurb. If you can't write a compelling blurb for your premise, the book underneath it won't be compelling either. The blurb is the premise pressure test. If it reads generic, the book is generic. AI cannot rescue a premise that doesn't work on the back cover.

5. Accept that the validation step takes longer than the outline. This is the part most authors won't do. Outlining with AI is fast and dopamine-rich. Researching the market is slow and feels like procrastination. It isn't. An hour of premise validation will save you weeks of writing a book nobody wants.

The authors I see succeeding with AI aren't the ones with the best prompts or the fanciest workflows. They're the ones who treat AI as the production tool it is — and do the strategic thinking before they hand it the wheel.

(Then there's the marketing piece too - which I'll cover next time)

Sound familiar?

You’ve tried using AI to write your novel. You’ve watched the YouTube videos, joined the Facebook groups, tested different prompts… and the output still reads like it was written by a robot who’s never had an emotion.

Or maybe you haven’t started yet because the whole thing feels overwhelming. Where do you even begin? What do you put in the prompt? How do you keep characters consistent across 25 chapters?

Here’s the thing most people get wrong: they start with the prompts. But the prompts aren’t the magic. The planning that happens BEFORE the prompts is what makes the difference between a generic AI draft and a book that actually reads like a novel.

This free guide shows you exactly how that planning works — and gives you 3 prompts you can use today.

What’s Inside the Free Guide

✓ The 3-layer prompting system that forms the backbone of a complete novel draft — not just random scene generation, but a structured approach that builds a real story

✓ Why most AI drafts fall flat — and the one planning step that fixes pacing, character voice, and emotional depth before you ever start generating text

✓ A clear, step-by-step workflow you can use with Claude, ChatGPT, or any LLM — no technical skills required, just copy, paste, and go

✓ The “story foundation first” approach that I use to write complete first drafts in a single day — without sacrificing quality

Get Your Free Copy Now

https://www.plotandprompt.com/free-plot-to-published-guide/

This is how I've learned to work with AI over the last few as I've been building novel packages for indie authors. I'm going to use cozy mystery as the example, but this is true across all genres.

I see the same pattern show up over and over in cozy mystery: writers pick a great hook, a great setting, and a great sleuth — but the combination doesn't actually create tension. The pieces don't argue with each other.

Here's what I mean:

A cozy mystery needs at least three tropes working together, and the magic happens when they create friction:

1. A protagonist trope that creates a reason to investigate (amateur sleuth with a personal stake, reluctant returnee to hometown, newcomer with something to prove)

2. A setting trope that limits the suspect pool and creates intimacy (small town, isolated inn, family business, tight-knit hobby community)

3. A complication trope that makes the investigation costly (romantic interest is a suspect, the victim was beloved by the protagonist, solving it threatens the protagonist's place in the community)

Here's some examples of ones that work well:

- Returning prodigal + family business in crisis + the prime suspect is the sibling who stayed.

- New-in-town professional + insular hobby community + the victim was her only friend so far.

- Retired investigator running a B&B + locked-location murder + the killer has to be one of her guests, who are also her livelihood.

When I'm building a package and a combo feels flat, I run it through three questions:

- What does the protagonist lose if she investigates?

- What does she lose if she doesn't?

- Who in her life is she going to disappoint either way?

If I can't answer all three, the combo isn't quite right yet.

I actually put a bunch of these into a free PDF — 12 cozy mystery trope combos with the friction built in, organized by sub-style (culinary, craft, pet, paranormal-adjacent, small town). I made it as a lead magnet for my site but I'm just going to link it here because it's the most useful version of this thinking I've put on paper. No upsell on the download itself, just the PDF.

Let me know if you want it.

First time watching - just finished Season 10. Felt like it could have been a series finale. Is it really worth continuing to watch after that? I would seriously miss Cristina and her quippy dialogue.

https://preview.redd.it/vajdegn0rx0h1.png?width=1724&format=png&auto=webp&s=39d504798864967fcc2d87cccbe981ff7df9700e

Story Seeds are the foundation of a novel, ready to plant. 😉

A complete codex, chapter-by-chapter outline, market validation, Amazon listing, and implementation guide — everything you need to start writing.

Write it your way: longhand, dictated, or with the AI tools you already use.

One buyer per title.

https://www.plotandprompt.com/story-seeds-summer/

Many writers come to AI with a premise and ask it to help them write.

That works fine — until you're 30,000 words in and realize your amateur sleuth has no business being involved in the mystery, your suspects are thin, and the community you built in chapter one has no real bearing on the plot.

The lack of structure is the problem.

Here's what I've noticed after building 200+ novel packages: cozy mystery is one of the most structurally demanding genres to write with AI assistance, because the genre has rules that aren't about prose — they're about architecture.

The amateur investigator needs a credible reason to investigate. The community needs to generate suspects naturally, not feel stapled on. The red herrings need to be plausible given what the reader knows about the world. The resolution needs to feel earned, not handed down.

If you don't give AI those structural parameters upfront, it'll give you scenes that don't add up to a cozy mystery.

The fix is boring but it works: before you write a single scene, define the trope combination that's doing your structural work. Not just "amateur sleuth" — but amateur sleuth + what setting, what community dynamic, what built-in reason to investigate.

That combination is where the architecture lives. Once it's locked, AI assistance gets dramatically more useful because every prompt has a framework to land in.

(by request, I put together a free one-pager on cozy mystery trope combos that do this kind of structural work — happy to drop the link in the comments or via DM if anyone wants it.)