Making your AI sound less like an AI

30 Jul 2025

5 min read

“Our chatbot is OK, but the responses sound too much like AI.”

We hear this a lot from companies starting to use AI for customer service. As we look for customer service and customer experience to be about human connections, having something that feels off can be an issue. Especially as we hear stats saying that 75% of people prefer talking to a human

But the same proportion (75%) of customers over the age of 65+ demand quick resolutions stating they would leave due to long wait times. 

So clearly some automation is required, but how can you make your AI seem more human? Let’s take a look at some ways, but let’s first ask the question: should you?

Should you make your AI sound human?

This seems like a no-brainer. Of course you should. But, ask yourself why? 

People are becoming more and more comfortable engaging with Large Language Models (LLMs) like ChatGPT or Gemini. This means that people are starting to understand the nuances of their chatting style, and still going back for more. 

If you are open and declaring that the thing your customers are talking to is an AI or a bot – which we think you should be – then why does it need to be human-like? 

Some customer service issues truly need a human touch. But there are plenty where AI performs just as well – often faster and more reliably.

So does your AI really need to sound human? That’s worth exploring.

There is an argument that perhaps you should lean into it and make it more like an AI so you are not fooling your customers. 

Will customers notice if they’re talking to a bot?

Many of the ecommerce brands we work with report back to us that they get customer comments and reviews praising the bot by name and giving feedback as though they were a human. “James was really friendly and helpful.”, “Alice sorted my problem out really quickly.” and so on. 

Even if it’s made very clear that it’s a bot, sometimes customers just don’t notice – or care.

On the other hand, we’ve heard from businesses that use no AI or automation that customers have complained about robotic answers from customer service agents. That’s because the agents are just using scripts and not deviating from them.

So people are getting more used to talking to AI, often they don’t notice or care if it’s AI, and sometimes they think humans are bots. So maybe you shouldn’t worry about your bot sounding AI?

How to Make Your Bot Sound Less Robotic

Even if customers don’t mind talking to a bot, you still want your AI to reflect your brand’s voice. That’s the power of generative AI – you can train it to respond with the right tone, style, and phrasing.

Here are 3 slightly different ways to respond to a similar customer message: “I’m a bit confused about which [product] to get for my wife lol”

The first, from Honeylove says: “I'd be happy to help you choose a bra for your wife's birthday! Could you provide a bit more information about what she might prefer?”

Another one from On says: No worries, I’ve got you covered! Here’s a quick guide to help you pick the perfect bottle for your wife:

Both feel similar but subtly different. That’s the power of generative AI to have similar messages so things don’t feel repetitive and hence, robotic. 

But you can take it a step further with much more personality, as shown in this example from air up. 

Notice all the emojis and the cheeky responses from the bot, mirroring the customer. This is partly because air up have chosen to make its bot playful given it has a younger audience. Hopefully you love it, but even if you think it’s too much, you have to concede that it exudes style. 

On the other hand, Charles Tyrwhitt use a much more formal tone, which suits their brand as a men’s formalwear brand.

These different approaches are all possible with generative AI because you can prompt the AI to respond in a particular way. 

But at the same time, some queries may need very precise phrasing or handling. Such as needing legal wording for certain answers, or anything that needs to go through compliance. At that point you can use scripted answers rather than ones generatively created. 

Training your AI agent to sound like your people

One option is to train the AI to speak how you want it to speak. A way to do this is to use good examples from your existing agents. Bloom & Wild asked their Head of Content to train Willow, their AI Agent, to sound just like one of their Customer Delight agents. You can hear them explain it here: 

If you have brand guidelines, or a guidelines for your human agents you can use that as a basis to write a prompt. With that in hand, you can then write prompts to explain how to speak and the tone you want to strike. 

If you're stuck or need inspiration, here are some ways to think about it. 

Questions to ask to build your prompt

  • Do you want your bot to be formal, or playful?

  • Do you want it to be verbose or succinct?

  • Should it use simple words and phrases? Or is it OK to use more complex language?

  • Should it use emojis?

  • How should it address your customers?

  • What questions does it need to answer precisely, and which ones can be more freeform?

If you want help building your AI Agent and getting the right tone of voice, then speak to us today.