Show, Don't Tell: Fixing Emotion Labels

What "telling" actually means

The advice "show, don't tell" gets repeated so often it's nearly lost its meaning. So let's be precise. Telling is when you hand the reader a conclusion about a character's inner state — she was sad, he felt afraid, they realized the truth — instead of giving them the evidence and letting them reach that conclusion themselves.

The reader doesn't get to feel anything, because you've already done the feeling for them and handed over the summary. It's the difference between a friend saying "that was terrifying" and a friend whose hands are still shaking as they tell you what happened.

The three forms it takes

1. The emotion label: "felt a [emotion]"

The most direct form. A character feels something, and you name it. "She felt a deep sadness." "He felt a surge of anger." The tell here is the naming — you're labeling the emotion rather than rendering it.

Notice the fix doesn't mention sadness at all. It doesn't need to — the jaw, the looking away, the not-watching-him-leave carry it. Naming the emotion would actually weaken the showing, because it tells the reader what to conclude instead of trusting them to feel it.

2. The abstract emotion noun: "a wave of dread"

This is the form AI reaches for most reflexively, and it's worth learning to spot on sight. It's a container noun — a wave of, a sense of, a pang of, a surge of, a flicker of — bolted to an abstract feeling: dread, unease, foreboding, longing, belonging.

These phrases feel literary because they have a shape and a rhythm — but they're empty calories. "A wave of dread" describes the category of feeling without a single concrete detail. The fix is always the same move: find the physical sign. Dread lives in the stomach, the throat, the held breath — not in an abstract wave.

3. The realization: "he realized that"

A quieter form, and easy to miss. "She realized that she loved him." "He understood that it was too late." You're narrating a cognitive event — the moment of insight — as a flat statement, when the insight would land harder shown through behavior.

The showing version dramatizes the process of realizing — the re-reading, the piece that doesn't fit — instead of announcing the conclusion. The reader realizes it a half-second before the character does, which is far more satisfying than being told.

Why AI over-tells

AI models are, at bottom, summarizing engines. When asked to convey that a character is sad, the highest-probability output is a clear, efficient statement of that fact — "she felt a deep sadness" — because that's what the bulk of their training text does. Emotion nouns like "a wave of dread" are especially common because they read as competent, literary-adjacent prose, which is exactly the register a model gravitates toward.

The result is prose that constantly tells you how to feel and rarely makes you feel it. A human writer tells when they're rushing or haven't found the image yet. An AI tells as its default resting state.

The one question that catches it

When you suspect a sentence is telling, ask: "How would a camera know this?"

A camera can record a clenched jaw, a slammed door, a hand that won't stop moving. It cannot record "a deep sadness" or "a sense of foreboding" — those are interpretations. If the only way to film your sentence is to have a narrator read it aloud, you're telling. Rewrite it as something the camera could see or hear.

• "Felt a [emotion]" → cut the label, show the body. felt a deep sadness → her jaw worked
• "A wave/sense/pang of [abstract noun]" → find the physical sign. a wave of dread → his stomach clenched
• "Realized/understood/knew that" → dramatize the process, not the conclusion.
• The camera test → if a camera couldn't record it, you're telling.
• Trust the reader → if the showing is good, you never need to name the feeling. Naming it on top is the tell.

When telling is the right call

You cannot show everything — a novel that dramatized every emotion would be exhausting and four times too long. Telling is the correct tool for transitions, minor beats, and pace control. "She was still annoyed about the parking ticket when she got to the office" is a perfectly good tell; that annoyance doesn't deserve a dramatized scene, and showing it would over-weight a trivial beat.

The rule of thumb: show the emotions that matter, tell the ones that don't. The problem is never that you told — it's that you told the moment the reader most wanted to experience.

Further reading

Emotional telling is one category in a larger set of AI writing fingerprints. The same instinct — trust the reader, cut the label — fixes dead verbs like "was sitting" and "began to walk", AI crutch phrases like "the weight of" and "a dance of", and overengineered dialogue tags.