Why does ChatGPT tell instead of show?

Models default to labels: she was sad, the room was tense, he realized he had grown. That is cheap to generate. A beat (the pan, the limp, the unpaid bill) has to come from you.

Will a humanizer fix show-don’t-tell?

It can vary rhythm after you swap the summary for a fact. It cannot invent the burned pan. One pass. Five passes often become purple prose.

What is purple prose in AI fiction?

Stacked metaphors, weather, and “tapestry of emotions” with no action. Cutting is the fix. Do not ask the model to make it more literary.

Is this the same as the fiction hub?

The hub locks plot and humanizes narration. This post is the sentence job: filter words, abstract emotion, even length. Dialogue is a separate post.

Humanize ChatGPT Prose That Tells Instead of Shows

ChatGPT will tell you the feeling and skip the object. She was devastated. The atmosphere was tense. He knew everything had changed. That is not a scene. That is a report.

Show-don't-tell is craft. Human Writes is the rhythm pass after you put one concrete beat on the page. It is not a class in fiction. Use Story purpose. Do not let a rewriter add moonlight and a single tear.

Hub: How to Humanize ChatGPT Fiction. Sentence habits that apply to any draft: Natural AI Writing. Talk: Humanize AI Dialogue.

The tells (and the purple cousin)

AI habit	What to put instead
Felt / realized / knew	An action that proves it
Was sad / angry / tense	What they do with their hands
Filter: saw, heard, noticed	The thing itself
Even 18-word sentences	One short. Then a longer one.
Purple weather	One sensory fact you already chose
“A mix of emotions”	Pick one and pay for it

If you humanize the left column without filling the right, you get prettier telling.

One beat, then one pass

Circle every abstract emotion in the paragraph.
Replace one of them with a fact from your bible (the limp, the unpaid tab, the voicemail length).
Delete the weather unless the plot needs rain.
Paste the paragraph into Human Writes, Story, once.
If the output added a metaphor you did not ask for, cut it.

You are not showing everything. You are showing the one thing that makes the feeling expensive.

Before and after

AI draft:

Elena felt a profound sense of betrayal wash over her as she realized Tomas had never truly loved her. The room seemed to close in, and a tapestry of memories flooded her mind.

After one beat + one Story pass:

She counted the plates. Four, still. He had said he would take his. The fourth plate was the sentence. She left it dirty.

Same scene job. Second version cannot be filed under “sad woman in kitchen.”

Filter words (quick cut list)

Saw, heard, watched, noticed, seemed, began to, started to, could feel, somehow, just, really, very, quite. AI stacks them. Cut half before you humanize. Keep a few if the narrator is actually uncertain.

Do not replace them with “crystalline” and “elucidated.” That is purple.

Read it with the names covered

If two adjacent paragraphs could swap novels, you are still telling. The beat should be this kitchen, this debt, this limp. That is the same test as dialogue with names off, applied to narration.

Workshop note: “show more” does not mean more adjectives. It means the reader can infer the feeling without the label. If you must keep one “she was tired,” earn it with the third night of the same dish.

Where Human Writes helps

Layer	Use it?
Choosing the beat	No
Plot / names	No
Stiff narration after the beat	Yes, Story, once
Making it “more literary”	No

Chapter-scale paste makes telling worse because the model averages every paragraph. Work scene by scene.

What not to do

Do not prompt “show don’t tell” and accept the new weather.
Do not run five passes until every noun has an adjective.
Do not treat a detector score as craft.
Do not claim a humanizer replaces workshop notes.

Bottom line

Show is one expensive fact. Human Writes is for the sentences around it. Keep the bible. Cut the tapestry.

Paste the paragraph on Human Writes when the plate is still dirty.