Writing With AI: When to Use It (and When to Run)

The Honest Truth About AI Writing Tools

I use AI tools. I also write stories about AI as the villain. This might seem like a contradiction, but it’s not—I understand what these tools can and cannot do.

After two years of using AI for various parts of my writing process, I’ve developed strong opinions about where it helps and where it hurts. This isn’t speculation—this is what I’ve learned by using these tools daily.

What we’ll cover:

• Where AI genuinely helps (research, brainstorming, editing)
• Where AI actively hurts (voice, originality, emotional truth)
• How to use AI without it using you
• The patterns that reveal AI-generated text (and how to avoid them)

What AI Does Well

Let me start with what works. AI excels at certain mechanical tasks that used to eat hours of writing time.

1. Research and Fact-Checking

Use case: Verifying historical details, technical accuracy, timeframes.

Example: I needed to know when certain AI technologies became public knowledge for The NeuroRender of Dorian Grant. Instead of diving down research rabbit holes, I asked an AI to give me a timeline with sources. Then I verified the sources myself.

The key: Use AI to find information faster. Never trust it to be accurate. Always verify.

2. Breaking Writer’s Block

Use case: Getting unstuck when you know where the story goes but can’t find the words.

Example: Midway through a story, I knew character A needed to confront character B. I knew what they’d argue about. But I couldn’t find the right opening line for the scene.

I asked AI: “Give me 10 ways character A could start this confrontation.”

Seven were terrible. Three sparked something. I took one, rewrote it completely in my voice, and finished the scene myself.

The key: Use AI for options, not solutions. The spark, not the fire.

3. Structural Editing

Use case: Identifying pacing issues, plot holes, inconsistent details.

Example: I fed a draft story to AI with this prompt: “Read this story and tell me where you lost interest, what felt slow, and what felt rushed. Don’t rewrite—just diagnose.”

The feedback was surprisingly accurate about pacing. Useless about voice and emotional impact, but good at structural problems.

The key: Use AI as a first reader for mechanical issues. Ignore its opinions on style and voice.

4. Title and Tagline Generation

Use case: Breaking through the paralysis of naming things.

Example: I had a story about an AI that makes perfect predictions. I spent two hours trying to name it. Finally asked AI for 50 title options.

48 were generic. Two were interesting. I combined elements from both and created something better than either. “The Munchausen Algorithm” came from that process.

The key: Use AI for quantity of options. Use your judgment for quality.

What AI Does Terribly

Now for the dangerous parts. These are where AI will actively hurt your writing if you’re not careful.

1. Voice and Style

AI has no voice of its own. It averages the internet. The result is inoffensive, generic, forgettable prose that sounds like a committee wrote it.

Bad AI output:

“The city stretched out before them, a tapestry of lights and shadows that seemed to pulse with life and energy.”

Why it’s bad: Generic imagery. No specific details. Could describe any city anywhere. Zero personality.

Human revision:

“From the bridge, the city looked like a circuit board someone spilled coffee on—half the lights dead, the rest flickering.”

Why it’s better: Specific, surprising comparison. Implies decay. Has attitude.

Your voice is the only thing you have that’s truly yours. Don’t let AI smooth it into nothing.

2. Emotional Truth

AI can describe emotions. It cannot feel them. This shows in the writing.

AI’s version of grief:

“She felt a deep sadness wash over her as she looked at the empty chair.”

Human version:

“She set the table for three. Again. She got halfway through the meal before noticing the extra plate.”

AI reaches for emotional labels. Humans show the physical manifestation of emotion through habitual action. The difference is everything.

3. Subtext and Implication

AI tends to explain everything. It doesn’t understand the power of what’s left unsaid.

AI ending:

“He realized that the AI had been manipulating him all along, and he had been complicit in his own downfall. This taught him an important lesson about trust and technology.”

Human ending:

“The next morning, Marcus opened his laptop out of habit. His hand hovered over the power button. He left it closed and went for a walk instead.”

AI states the theme. Humans trust readers to infer it.

4. Originality and Risk-Taking

AI generates based on patterns in its training data. It will always give you the most statistically likely next sentence. That’s the opposite of interesting fiction.

AI will never write:

• A truly surprising metaphor
• An uncomfortable truth
• A voice that breaks rules deliberately
• Anything that risks confusing readers

Good fiction requires all of these.

The Patterns That Expose AI Writing

Learn to spot AI-generated text so you can avoid these patterns:

Pattern 1: The Not-But Construction

• “It wasn’t just X, but Y”
• “She didn’t simply feel X, she felt Y”

Pattern 2: Thesis Statement Endings

• “This taught him…”
• “She realized that…”
• “In the end, what mattered was…”

Pattern 3: Empty Intensifiers

• “Truly,” “really,” “deeply,” “profoundly”
• These words do nothing. Cut them.

Pattern 4: Over-Explanation

• AI explains the metaphor after using it
• AI tells you what emotion a scene should evoke
• AI labels themes explicitly

Pattern 5: Generic Descriptions

• “Tapestry of…”
• “Stretched out before…”
• “Seemed to pulse with…”

If your writing hits these patterns, you’re either using too much AI or you’ve internalized AI’s bad habits.

My Actual AI Workflow

Here’s exactly how I use AI in my writing process:

Phase 1: Ideation (AI: 20% / Me: 80%)

• I generate the core idea myself
• I might ask AI for 20-30 variations on specific elements
• I cherry-pick interesting angles
• I develop everything myself from there

Phase 2: Research (AI: 60% / Me: 40%)

• AI finds potential sources and information quickly
• I verify every fact independently
• I use AI to organize research, not to trust it

Phase 3: First Draft (AI: 0% / Me: 100%)

• Zero AI. This is where voice lives.
• All raw writing comes from me
• AI ruins the spontaneous energy of first drafts

Phase 4: Structural Edit (AI: 30% / Me: 70%)

• I ask AI to identify pacing issues
• I ask AI to spot plot holes
• I ignore 80% of its suggestions
• I implement only structural fixes I agree with

Phase 5: Line Edit (AI: 10% / Me: 90%)

• I might ask AI if a metaphor is cliché
• I use AI to check for repeated words
• I do all actual rewriting myself

Phase 6: Final Polish (AI: 0% / Me: 100%)

• I read aloud and make final adjustments
• No AI—this is about how the prose sounds and feels

Total AI influence in final work: ~15%

And that 15% is never prose—it’s research assistance and structural feedback.

Rules for Using AI Without Losing Yourself

Rule 1: Never publish AI’s exact words Always rewrite in your own voice. Always.

Rule 2: Use AI for breadth, not depth Use it to generate many options quickly. Use yourself to create something with depth.

Rule 3: Trust your gut over AI’s suggestions If an AI suggestion feels wrong, it is wrong. Your instinct knows your story better than any algorithm.

Rule 4: Write the first draft yourself The first draft is where you discover what the story actually is. AI can’t do this for you.

Rule 5: Read aloud to catch AI patterns AI prose sounds flat when read aloud. Your voice sounds natural. Trust your ear.

Rule 6: Protect your voice If you find yourself writing more like AI, stop using AI for a month. Reset your instincts.

The Real Danger

The biggest risk isn’t that AI will replace writers. It’s that writers will start writing like AI—smoothing out quirks, removing personality, optimizing for averageness.

I’ve seen writers lose their voice by relying too heavily on AI editing. The tool suggested removing everything distinctive until nothing interesting remained.

Your weird metaphors, your odd rhythms, your uncomfortable observations—these are what make your writing yours. Protect them.

Exercise: Detect AI in Your Own Work

Take a recent piece. Search for:

1. “Not just X, but Y” constructions
2. Sentences that explain rather than show
3. Generic descriptors (tapestry, seemed to, etc.)
4. Over-use of intensifiers (truly, deeply, really)
5. Endings that state the theme explicitly

If you find these patterns, rewrite those sections in your natural voice. Read them aloud. Feel the difference.

The Bottom Line

AI is a tool. Like any tool, it’s useful for specific tasks and dangerous when misused.

Use it to:

• Generate options quickly
• Check facts and research
• Identify structural problems
• Break through temporary blocks

Never use it to:

• Write your first draft
• Create your voice
• Make final prose decisions
• Replace your judgment

The stories that matter come from human experience, observation, and risk-taking. AI can help with the mechanics. It cannot help with the soul.

Want to see human-written fiction that uses AI as subject matter but not as author? Read our collection of AI cautionary tales. Every word is human. The irony is intentional.

---

Joe Kryo writes about AI while using AI tools strategically. He believes the future of writing involves collaboration between human creativity and machine assistance, but only if humans remain firmly in control.