How Agents Spot AI-Written Fiction (Before You Even Query)

AI tools are everywhere in the writing world right now. They can brainstorm, outline, and even rewrite your entire novel in seconds. But here’s the uncomfortable truth: agents and editors can already tell when a manuscript has been heavily written or edited by an AI model—often within the first few pages.

And no, they’re not relying on AI detection software. The patterns are just that obvious.

As more writers lean on tools like ChatGPT and Claude, thousands of manuscripts are starting to share the same emotional tone, rhythm, and stylistic quirks. If you want a traditional publishing deal—or simply want your work to sound like you—you need to know what these tells look like and how to avoid them.

The Real Litmus Test: Who Wrote the Prose?

AI can absolutely be part of a modern writer’s toolkit, but there’s a crucial line you shouldn’t cross if you care about originality and copyright guidance.

The core question is simple:

Who is actually writing and editing the prose?

Ethical, sustainable use of AI for creative writing usually looks like this:

• Using AI for brainstorming, structure, or idea generation
• Asking for high-level feedback on pacing, clarity, or plot holes
• Getting suggestions for comps, blurbs, or marketing angles

Where writers get into trouble is here:

• Letting an LLM rewrite or "polish" their prose
• Using AI to line edit or "make it sound better" at the sentence level
• Feeding in chapters and asking the model to rewrite them in a stronger style

That kind of editing leaves fingerprints all over your manuscript. And agents are already trained to spot them.

If you’re interested in how AI is reshaping creative work more broadly, it’s worth looking at how people are using it to build products and income streams in other fields, like in these AI tools that can actually make you money. But for fiction, the bar for authenticity is much higher.

Tell #1: Nonsensical, "Pretty" Metaphors

One of the biggest giveaways of AI-generated or AI-edited prose is the sudden appearance of metaphors that sound striking at first glance—but fall apart the moment you think about them.

These metaphors:

• Aim for mood and drama
• Stack vivid words together
• Don’t quite make logical or sensory sense

For example, describing silence as a "thunderstorm folded into velvet" might sound poetic, but it doesn’t really hold up. What does that actually mean in the scene? How does it relate to the character’s experience?

Agents are now seeing manuscripts filled with these kinds of metaphors—often in the opening pages, where writers have asked AI to "make this more vivid" or "more literary." Instead of elevating the writing, it signals that the voice isn’t organic.

Used intentionally for a quirky character? That could be brilliant. Sprinkled all over a serious manuscript? It’s a red flag.

Tell #2: Emotional Flatlining Across the Story

Human-written fiction naturally moves through emotional highs and lows. Some scenes are quiet and restrained. Others are intense and explosive. Different characters bring different emotional registers.

AI-edited prose often flattens all of that.

The result:

• Every chapter feels like it’s written at the same emotional pitch
• A simple breakfast scene and a devastating breakup read with the same dramatic intensity
• There’s no sense of modulation, escalation, or release

Good craft involves emotional movement: restraint, build-up, reversal, and relief. When an LLM "tidies" your prose, it tends to normalize everything into one polished, slightly heightened tone. To an agent, that sounds less like a living voice and more like a system output.

Worse, if you had a strong voice to begin with, AI editing can strip it away. Your original draft might be raw but powerful; the AI-polished version becomes smooth, even, and oddly generic.

Tell #3: Adjective and Simile Overload

Another common pattern: nearly every noun and action gets dressed up with adjectives, similes, or both. The prose starts to feel coated in description rather than driven by it.

Signs of this overload include:

• Multiple adjectives before most nouns
• Frequent "X like Y" comparisons for simple actions
• Emotional states described with layered, dramatic imagery

Think of a sentence like:

"The moon hung like a bruised pearl over the trembling, restless, haunted sea as her fractured heartbeat thudded like a broken hymn."

On its own, it might sound intense. But when every paragraph is built like this, the writing feels overworked and artificial. Agents are now seeing the same phrases—like "broken hymn"—repeated across unrelated manuscripts, which is a strong clue that AI was involved.

Tell #4: Recycled LLM Rhetorical Patterns

Large language models have favorite rhetorical tricks they lean on again and again. When you ask them to "elevate" or "intensify" prose, these patterns show up fast—and agents are starting to recognize them instantly.

Common examples include:

• "It was not X, it was Y."
• "She did not just feel it, she became it."
• "Not X. Not Y." constructions for emphasis

On the page, this might look like:

"Not this was fun. Not let’s stay in touch."

Used once in a while, this kind of structure can be effective. But when it appears again and again, it reads like a template rather than a discovered sentence. Multiply that across hundreds of manuscripts using the same tools, and the pattern becomes impossible to miss.

Tell #5: Floating Dialogue With No Setting

Another major giveaway is the lack of concrete scene grounding. AI-generated or heavily AI-edited fiction often leans on dialogue and abstract feelings, but skimps on the physical world.

Typical signs:

• Pages of back-and-forth dialogue with no sense of where the characters are
• Emotions described in vague, internal terms, but no external action
• Very few sensory details: no sounds, textures, smells, or physical movement anchoring the scene

You might read a full page of two characters arguing or flirting and still have no idea:

• What the room looks like
• How they’re moving
• What they’re doing with their hands, faces, or bodies
• How the environment is affecting them

Without lived, specific detail, the scene feels like it’s happening in a void. That’s a strong sign the text was generated or heavily rewritten by a model that’s good at language patterns, but not at embodied experience.

Bonus Tell: The Rule of Three, Everywhere

The "rule of three" is a classic rhetorical device: grouping words or phrases in triplets for rhythm and emphasis. AI models love it—and they overuse it.

On its own, a line like:

"She was furious, frightened, undone."

or

"Cold, cruel, calculating."

can be powerful. But when nearly every emotional beat falls into a three-part list, the writing starts to feel assembled rather than discovered.

Agents are now seeing manuscripts where this pattern appears so often that it becomes a drumbeat. Combined with the other tells—nonsensical metaphors, emotional flatlining, and recycled rhetorical structures—it paints a clear picture of AI involvement.

How to Use AI Without Losing Your Voice (or Your Shot)

None of this means you should avoid AI entirely. Used well, it can make you a more productive and more informed writer. The key is to use it as a support tool, not a ghostwriter or line editor.

Safe, effective uses include:

• Brainstorming plot ideas, character backstories, or worldbuilding details
• Getting high-level feedback on structure, pacing, or stakes
• Summarizing your own chapters so you can see the shape of your story
• Exploring alternative scene approaches without copying the output

Risky uses that leave obvious tells:

• Asking AI to "rewrite this chapter to be more emotional"
• Letting it "polish" your prose line by line
• Using it to generate large chunks of your manuscript, then lightly tweaking

Remember: if you love how smooth and "professional" the AI-edited version sounds, that’s probably because it sounds like everyone else using the same tools. Agents are not just looking for clean writing—they’re looking for a distinctive human voice.

If you’re curious about how these same models are being used to build more advanced systems, it’s worth exploring guides like the essential skills you need to build real AI agents. The tech is powerful—but in fiction, your power is still your voice.

The Bottom Line: AI Is a Tool, Not Your Co-Author

We’re entering a new era for literature, where writers have access to incredibly capable AI assistants. That’s exciting—but it also means the bar for authenticity is rising.

Agents want you to use AI responsibly. They’re not anti-technology; they’re pro-author. They want you to get better at your craft, not outsource it. They want to represent books that only you could have written.

Use AI for structure, ideas, and feedback. Let it help you think—but not speak. The prose itself needs to be yours.

Because when your pages land on an agent’s desk, they’re not just looking for a polished manuscript. They’re looking for a human mind on the page.