Why AI Help Can Make You Seem Less Caring

You’re staring at a blank card.

Someone you care about just went through a terrible breakup. You want to say the right thing. You also don’t want to sound like you copied and pasted the right thing.

So you do what a lot of people do now: you open an AI tool, type a prompt, and ask for a draft.

The words it gives you are… good. Maybe even great. You tweak a line. You add their name. You hit send.

Here’s an uncomfortable question, though: even if the message lands, what does the other person think that message says about you?

A 2025 paper by Claessens and colleagues in Computers in Human Behavior looked at that exact problem. Across multiple studies and scenarios, participants read about someone completing a task in one of a few ways: doing it themselves, having AI do it, or having another person do it. Then they rated the messenger.

What they found was a consistent pattern. When someone handed the task off–whether to AI or to another person–observers evaluated them more negatively than when they did it themselves. The biggest hits in judgement showed up in ratings of laziness and competence, with smaller noticeable shifts in perceived warmth, morality, and trustworthiness.

That’s already useful. But the more important detail is when the penalty shows up most.

Two Kinds of Tasks

Some tasks are mainly about the output. If the code works, it works. If the schedule is clear, it’s clear.

Other tasks are about the output AND what the output signals.

Think about an apology, a condolence message, wedding vows, or a love letter. In those moments, the “job” isn’t only to produce words. The job is to show the other person that you showed up because you care.

In Claessens' studies, outsourcing had a stronger negative impact in these relationship-heavy tasks than in more practical ones (like writing computer code or planning a syllabus). In other words: the same shortcut can look like efficiency in one context and like distance in another.

The Effort Signal Tax

When a shortcut removes the evidence that you tried, you pay a social cost. We can call this cost an effort outsourcing tax.

Yes, people read a message for explicit information. But they also read it for implicit hidden clues. A lot of times, effort is one of the loudest clues we have that a message is important.

When someone believes you spent time choosing words, remembering details, or sitting with your feelings, they often treat that time as part of the gift. When it looks like you didn’t, the message loses value even if the sentences are perfectly polite.

This is why two messages with the same words can land differently depending on how the recipient believes they were produced.

Why AI Outsourcing Changes the Signal

The paper also digs into what people seem to be reacting to. Their findings suggest that when observers think less effort went into the work, they’re more likely to see it as:

1. less authentic: it feels less like it came from you, and
2. less caring: it feels like the task mattered less to you.

Think about it like this: imagine a friend texts you at 11:59 p.m. on your birthday. It’s not like you feel ungrateful for the message. But the timing suggests you weren’t top of mind, leaving the message a little hollow.

AI can create the same kind of inference. AI isn’t inherently “bad,” but it can erase the visible signs of time and attention in moments where time and attention are the point.

Artificial Intelligence Essential Reads

Is AI Slop the New Spam?

Medical AI Just Lost to a General Model

How to Use AI Without Paying the Tax

So instead of thinking about banning AI from your life, think about how to avoid the accidental social costs that can come with using it.

Here are three moves you can try that follow the logic of this paper.

First, keep AI away from the messages that are supposed to prove you care. If the main purpose is relational like an apology, condolences, or an expression of gratitude, you should still feel free to use AI for brainstorming (it’s great at that). But write the final version yourself.

Second, when you do use AI, add cues of effort that can’t be mass-produced. Include a specific shared memory. Mention a detail that only you would know. Make a concrete offer (“I can bring dinner Tuesday” beats “Let me know if you need anything”).

Lastly, feel free to use AI without reservation when the task is mainly practical. For planning, summarizing, formatting, or first-pass ideation, the social meaning of effort is low. Claessens's results suggest those contexts come with less judgement.

Takeaway

In relationship-heavy moments, people judge the process as much as the product. So while AI can make communication easier, you also have to think about how it can also change what you’re communicating.

If you want to avoid the effort outsourcing tax, make sure the other person can still feel your voice in the work.