Why AI Writing Feels Soulless and How to Add Real Emotion

AI posts feel soulless because the model has nothing to feel about. It hasn't lived the event, it drifts toward a safe and positive mood, and it can only name an emotion. To fix a draft, choose one feeling, find the real fact that earns it, and build the post in the order that feeling needs.

We make VoiceMoat, which handles feeling in its drafts. That comes near the end, and everything before it works with any AI tool.

For the words and rhythm, read our de-AI playbook or why AI LinkedIn posts all look alike. This page is about the mood.

Why AI posts feel flat

The mood stays positive. A 2024 study in Frontiers compared ChatGPT and human answers to 2,000 Stack Overflow questions. ChatGPT was more uniformly positive, while people showed a wider range, including anger and disgust. A smaller 2025 paper on 512 tweets rated about 90% of ChatGPT's replies positive, against about 27% of human comments. Neither looked at posts written to publish, and the second is thin on method, but they point the same way.

There's no event behind it. A feeling has a cause: a number you hit, someone who let you down, a launch that nearly broke. A model doesn't have your Tuesday, so it reaches for labels like "thrilled" and "humbled". That's our reading, untested by any study we found.

One mood repeats. Ask for a post and you often get a calm lesson. When we compared VoiceMoat's own LinkedIn drafts with public creator posts in October, most of ours were insight posts and none read as warm. Fixing that took a rebuild.

Does feeling actually help a post?

Often, though it depends on what you want back. Berger and Milkman's study of New York Times articles found that awe, anger and anxiety drove sharing, sadness reduced it, and positive pieces beat negative ones overall. That's news, and chasing shares can push you toward outrage.

A more useful question: what reaction do you want? LinkedIn defines each one. Celebrate is for a milestone. Support is for a hard time. Love is for something heartwarming. Insightful is for a new idea. Funny is for a laugh. Those are five different posts, and AI usually gives you the Insightful one.

Match the feeling to the fact

No fact? Keep the post light, or go and find one.

How to add emotion to an AI draft

1. Pick one feeling. Two at once turns to mush.

2. Write down the fact first: what happened, when, the number, who was there.

3. Give the order. For proud: the win in line one, then the cost or the credit. For frustrated: the obstacle, its cost, then what you're doing about it.

4. Set a strength: light, medium or strong. If you wouldn't say it to a colleague, turn it down.

5. Delete the labels. Replace "thrilled" with what caused it.

6. Let hard posts stay hard. A sad post shouldn't turn cheerful in the last line.

7. Never invent the feeling. A made-up struggle is worse than a dull post. Here's how to stop AI making things up.

Here's the difference, using two invented examples.

Flat: Thrilled to share that we've reached an incredible milestone this quarter. So grateful for this amazing team. Onward!

Proud, medium: We hit 1,000 paying customers on Tuesday. It took 31 months. For the first 9 we had 12 of them, and I still open that spreadsheet. Priya answered support for all 1,000.

Flat: Setbacks are just setups for comebacks. We didn't get the grant, but we learned so much!

Disappointed, light: We spent 4 months on a grant application and heard on Friday that we didn't get it. I still think the plan was good. It still hurts a bit.

On Twitter you have room for less, so keep the fact and one reaction.

A prompt you can copy

Draft a LinkedIn post that makes the reader feel [one feeling].

The real fact: [what happened, when, the number, who].

Order: [fact first, then the cost or the credit, then one honest reaction].

Strength: [light, medium or strong].

Use only the facts I gave you. Don't name the emotion. Don't end with a lesson. If a detail is missing, ask me and don't make it up.

For the words themselves, training AI on your own posts does more than any single prompt. Our free LinkedIn AI post detector gives a second opinion on the tells.

What other tools give you

None of the pages we read describe tying a feeling to a fact you supply, so that part stays with you.

How VoiceMoat handles feeling

VoiceMoat is an AI social media tool for LinkedIn and Twitter. audenAI, the brain inside VoiceMoat, drafts each post in one feeling, chosen from 26 in 6 families: Play, Warm, Spark, Edge, Tender and Calm.

• You see the choice. Each draft shows its feeling, its strength and the reason, such as "Wry · light".

• It needs a real fact. Proud needs a real result. With nothing to stand on, the post stays light and you're asked: "What really happened, and when?"

• It stays in your range. Humour and strength are held to what your own posts show, per platform. Hard topics such as layoffs or illness get no jokes.

• "Vulnerable, now" is yours alone. It's locked unless you share a hard time yourself.

• One click changes it: Warmer, Funnier, Calmer or More fired up, or any feeling and strength. The post is redrafted for 1 credit, or 2 for a thread.

• It works in Studio, chat and Grow, on both networks, and learns slowly from what you choose and what lands.

What it can't do:

• It can't supply the event. If you give it nothing true, you get a light post and a question.

• Replies, LinkedIn comments and polls don't get a feeling.

• Spotting which facts you've given is automated, and it can be wrong. audenAI, the brain inside VoiceMoat, does the work and you decide what goes out.

What each plan includes

If your drafts keep coming back flat, try VoiceMoat: Creator is $1 for 3 days.