🤖😭 AI Can’t Cry, But It Can Fake It Pretty Well

Six ways to trick ChatGPT into writing like it has a soul.

Writing AI prompts that spark real emotion can feel a bit like trying to cry on cue: awkward and utterly unconvincing. Sure, telling an AI to “be emotional” will transform your text, but usually not in the desired way. You would need better prompting than that.

The secret? Generic prompts lead to generic output, while highly specific prompts give you… well, more customized output. Let’s look at a few prompting techniques that get AI out of its robotic shell and into something that feels a little more human.

Narrative Voice Injection

Injecting emotional depth into AI-generated text is about more than asking the model to “sound human”. It’s about telling it exactly which human to be, and how to feel. Research shows that assigning a specific narrative voice or persona in your prompt, combined with explicit emotional cues, can enhance the nuance and relatability of AI responses.

For example, instructing the AI to write from the perspective of a “65-year-old Scottish fisherman who’s watched his family’s livelihood vanish” not only grounds the output in lived experience but also encourages the model to use visceral, metaphor-rich language-think “rotting nets” and “silent harbors”-instead of generic, statistical summaries.

Recent studies, including a 2023 Microsoft-led paper, found that simply appending a phrase with emotional tone (“This answer is important to my career”) can improve the quality, truthfulness, and detail of responses by up to 10% across multiple large language models. The key is to tailor emotional cues to the task: do you want empathy, urgency, or reassurance? Try stacking behavioral cues or adding context-specific emotional hooks, then A/B test¹ your prompts to see which type gives you the most authentic and fitting voice.

Finally, while AI can convincingly mimic emotions, it’s still bluffing its way through the human condition-so don’t be afraid to get specific. “Pretend you’re human” is a start, but “pretend you’re a retired jazz musician who just lost her favorite saxophone” is where the real magic (and emotional depth) happens.

Seed Words

Another powerful way to get more believable feeling from AI is by priming it with emotionally charged seed words. Instead of relying on the model to conjure up the right mood from scratch, you hand it a bunch of evocative words like “bittersweet,” “frayed,” or “echo,” and ask it to paint with those colors.

This isn’t just a quick hack: language models are expert pattern matchers, and when you supply them with emotionally resonant vocabulary, they instinctively weave those words-and their associated imagery-into the narrative. For instance, prompt the AI to “incorporate these words naturally: bittersweet, frayed, echo” in a poem about leaving a childhood home, and you’ll get lines that hum with nostalgia and loss, like “bittersweet echoes in frayed wallpaper.”

To put this into practice, brainstorm a handful of words that capture the emotional flavor you want, and explicitly include them in your prompt. Then, experiment: swap out seed words, try different emotional palettes, and see how the tone transforms.² You might discover that a single well-placed “echo” can do more emotional heavy lifting than a dozen generic adjectives-proof that, sometimes, the heart really is in the details.

Constrained Storytelling

If you want AI to move beyond the generic and tap into something genuinely relatable, try the art of constrained storytelling.

Paradoxically, the more you limit the AI’s scope-by setting quirky boundaries or banning abstractions-the more vivid and emotionally resonant its writing becomes. This is no accident: both research and creative writing wisdom agree that constraints force us (and our digital counterparts) to zero in on the telling details.

For example, instead of asking the AI to describe happiness in the abstract, challenge it with: “Describe happiness, but only through objects found in a 10-year-old’s backpack. No abstract terms.” Suddenly, you get a cascade of tangible nostalgia: crushed candy wrappers, a secret note, a pebble with mysterious powers. These specifics are what readers connect with, because they feel real.

To harness this technique, don’t be afraid to get oddly specific with your prompts: set boundaries, ban certain words, or require the story to unfold in a single room or through a single sense. AI, much like a college student, often needs a little structure to produce its most profound work. Give it a constraint, and watch it surprise you.

Iterative “Yes, But” Refinement

One effective way to unlock emotional nuance and specificity in AI-generated writing is by just starting somewhere and making it better step by step.

Rather than settling for the first draft, prompt engineering experts-including those at Stanford-recommend a workflow of repeated, targeted adjustments: start with a broad prompt, then ask the AI to add concrete details, layer in subtext, or shift emotional tone. This is not unlike how human writers revise, and it consistently transforms generic outputs into more lively, convincing narratives.

For example, you might begin with, “Write a breakup letter,” then refine with, “Add a detail that’s oddly specific-like the time we argued about microwave settings,” and finally, “Rewrite it as if the sender regrets it but won’t admit it.”

Research from Stanford claims that iterative refinement (among many other techniques) not only improves accuracy and relevance, but also encourages the model to surface deeper, more relatable emotions and perspectives.

Sensory Anchoring

Using sensory language can quickly and efficiently add emotional depth. There’s a simple reason for this: by prompting the model to describe what characters smell, touch, or hear, you tap into the same mechanisms that make human memories vivid and emotionally charged.

Neuroscience shows that sensory details-especially those tied to smell, sound, and touch-are directly linked to the brain’s emotion and memory centers, such as the amygdala and hippocampus.³ This connection explains why a scene that lingers on the sticky heat of summer air, the flicker of a porch light, or the salt of sweat on skin feels more intimate than one that relies on clichés like “their eyes met”.

When you instruct the AI to focus on sensory cues, you’re using the brain’s natural tendency to process emotional experiences through concrete, sensory-rich details.

To put this into practice, just instruct your AI to show emotional tension through physical sensations or environmental cues, and watch as generic prose transforms into writing that readers can almost smell, hear, and feel.

Please note: some chatbots will actually use too much sensory details unprompted. In my experience, AI seems to love smells, including metaphorical smells.⁴ If this gets overly monotonous, maybe ask for other senses instead.

Conflict Embedding

To make your AI-generated writing feel truly human, don’t shy away from a little emotional turbulence. Embedding low-stakes contradictions (like “excited and anxious”, or “proud yet full of self-doubt”) mirrors the way real people experience life’s messy moments.

Humans are hardly paragons of emotional consistency; in fact, it’s the friction between feelings that makes us relatable. Prompt the AI with, “Write a birthday party invitation from a parent who’s excited but overwhelmed. Let both emotions show.” The result might be an invitation that’s equal parts cheerful and frazzled: “Please come celebrate Jamie’s 7th! (P.S. If you spot a unicorn cake that won’t bankrupt me, DM ASAP.)” This blend of joy and worry rings true because it’s how people actually talk-and fret-about life’s big events.

Try it: when crafting prompts, explicitly ask the AI to express at least two conflicting emotions or perspectives. You’ll find that a dash of contradiction not only adds depth, but also makes the writing more memorable. After all, perfection is for robots; a little anxiety is what makes us human.

Conclusion

It’s clear that the best emotional writing doesn’t come from algorithms alone, it comes from humans who know how to guide them.

By priming your prompts with specificity, layering in sensory and emotional cues, and refining through iterative feedback, you transform AI from a mechanical responder into a genuine storyteller. And just like that, your content becomes more engaging, and starts resembling real human writing.

So, the next time you want your writing to resonate, don’t settle for generic. Experiment, iterate, and let your prompts do the heavy lifting. Ready to see what your next story can feel like? Try these techniques and share your most surprising results. I’d love to hear what your AI creates when you give it a little more heart.

Oh, and if you have any advice for more convincing emotions in text-to-image prompting, I could probably use it.

1. Please note that many advanced AI chatbots have a memory feature: they will remember past conversations to offer more personalized content. This will ruin your experiments. So turn off any memory fuction (or simply use another chatbot) before you try this.

2. See note 1.

3. I believe LLMs have neither an amygdala nor a hippocampus. But don’t quote me on that.

4. https://promptingculture.substack.com/p/the-resonator/comment/113800183