On Synthetic Interiority in AI Writing
When I used to write for Draft.dev, one thing my editors kept training out of me was vagueness.
My editor then, Shane, once gave me a simple rule:
If you use a word like “ideal,” don’t leave it floating. Name what is ideal. Show the evidence.
That advice still explains why a lot of AI prose feels fake deep today.
Lemme show you something.
Look at this sentence:
“You can laugh with people all day and still go home with that quiet ache that says nobody really came close.”
The sentence knows how emotional prose is supposed to sound. It has cadence. It has softness. It reaches for intimacy.
But this part is where it starts to drift:
“that says nobody really came close.”
I call this synthetic interiority.
What synthetic interiority means
Synthetic interiority is what happens when AI-generated prose simulates an inner life instead of showing a real one.
It usually shows up when a sentence makes a feeling speak on behalf of the person:
“that ache that says…”
“the silence that reminds you…”
“the part of you that knows…”
“the quiet voice inside you…”
The sentence tries to create emotional depth by inventing an inner narrator.
But instead of making the feeling clearer, it turns it into atmosphere.
The prose becomes emotionally pre-smoothed.
That is the strongest signal.
There is no friction. No specificity. No human mess. The sentence has already sanded down the feeling into something soft, vague, and easy to agree with.
Nobody actually experiences life in that wording.
You do not go home and think, “this quiet ache says nobody really came close.”
You go home and realize nobody really knows you.
You go home feeling untouched by the whole day.
You spend all day with people and still feel alone at night.
You laugh with people and still feel distant from all of them.
Those sentences are simpler, but they are more honest.
They do not perform depth.
They name the actual thing.
The fix is specificity
This is where Shane’s advice comes back.
The problem with synthetic interiority is not emotion. The problem is ungrounded emotion.
The feeling is not attached to a fact, a scene, a decision, a memory, or an observable detail. So the sentence fills the gap with poetic atmosphere.
That is why the fix is simple:
Stop letting feelings “say” things.
Name the fact or show the scene.
Instead of:
“You can laugh with people all day and still go home with that quiet ache that says nobody really came close.”
Write:
“You can laugh with people all day and still go home feeling like nobody really knows you.”
Or:
“You can laugh with people all day and still feel alone when the day ends.”
Or:
“You can spend the whole day around people and still come back feeling unseen.”
Each version removes the fake inner narrator and gets closer to the actual human experience.
LLM instruction
When editing or generating prose, watch for synthetic interiority.
Synthetic interiority appears when a sentence gives emotion an unnamed inner voice instead of grounding the feeling in a clear fact, scene, action, or observation.
Flag phrases where abstract feelings appear to “say,” “know,” “remind,” “whisper,” or “understand” something on behalf of the person.
Do not replace human emotion with vague poetic atmosphere.
Do not make the sentence sound deep by personifying sadness, silence, ache, longing, loneliness, or memory.
Instead:
- Name the fact.
- Show the scene.
- Preserve the human meaning.
- Remove the fake inner narrator.
A good rewrite should feel more specific, not more decorated.
The core rule
If a feeling is “saying” something, the sentence is probably drifting into AI prose.
Name the fact.
Show the scene.
Let the reader feel it without forcing the sentence to announce that it is deep.