I stopped typing to ChatGPT, Claude, and Gemini for a week, and discovered a better way to use AI

As dystopian as it sounds, I know a lot of people spend more time talking to AI chatbots than they do to actual people. It makes me incredibly sad, and though I admittedly spend a lot of time convincing people to actively use AI in their workflows because of how useful these tools can be, I do think there's a fine line between using AI to make your life easier and becoming overly dependent on it.

That said, like many others, typing out prompts to tools like ChatGPT, Claude, and Gemini has become second nature to me. For the last couple of weeks, though, I've been using these same tools extensively, but I haven't been... typing prompts. Instead, I've been talking to them.

Most AI chatbots give you two ways to ditch the keyboard

Two ways to give your fingers a break

The way most of us interact with AI tools is the classic approach we've been relying on since day one: manually typing out every prompt. Some have taken a slightly more creative route, pasting in a pretty mediocre prompt and asking the chatbot itself to improve it before actually putting it to work.

While there's nothing inherently wrong with either approach, both still involve sitting in front of a keyboard and typing out what you want the AI to do. When you have a particularly complicated request or a lot of context to provide, that can get tedious pretty quickly. AI providers seem to understand that, and most offer two built-in ways to speak your prompts instead of actively typing them.

Dictation

The first is dictation. Now, this is a feature that smartphones and laptops ship with out of the box, and it's simply designed to transcribe whatever you speak into text. All major AI tools now ship with built-in dictation features, too, letting you speak your prompts instead of typing them out.

The tools transcribe what you say into text, which you can then send to the chatbot. Depending on the transcription, you might also get some punctuation and minor cleanup along the way, although these tools don't explicitly promise to rewrite messy speech into polished prompts.

That's where dedicated dictation tools like Wispr Flow come in. Rather than simply converting speech into text, they're designed to clean up what you're saying, removing unnecessary filler words and making the final result more readable. Flow is the one I personally use, simply because its transcripts are extremely accurate, well-formatted, and the tool works across all my devices and apps. I'm not going to talk too much about the tool right now, but you can read all about it in this article where I shared my entire experience.

Voice Mode

The next feature is one you've likely experimented with at least once: Voice Mode, or Gemini Live, as Google calls it. ChatGPT, Claude, and Perplexity all have their own versions of Voice Mode, and while there are some differences in how they work, the fundamental idea is pretty much the same.

Instead of speaking a prompt, waiting for it to be transcribed, and then reading through the chatbot's response, you can have an actual, back-and-forth conversation with AI. You speak, the chatbot responds out loud, and you can continue the conversation without ever touching your keyboard.

You can ask follow-up questions, interrupt the chatbot halfway through an explanation, or change the direction of the conversation entirely. It's a lot closer to having a discussion with someone than exchanging a series of carefully worded prompts.

Dictation helps me get my thoughts out before I forget them

I can speak a much better prompt than I can type

After spending a week experimenting with both, I realized that the real benefit isn't simply being able to use AI without typing. It's knowing which approach makes more sense for the task at hand.

For me, dictation has been particularly useful when I'm working on something that requires a lot of context. I'm a pretty fast typist, but I've always found that my thoughts move a lot faster than my fingers can keep up with. When I'm typing a particularly long prompt, I'll sometimes leave out details simply because I can't be bothered to type everything out.

Speaking my prompts eliminates that problem almost entirely. I can ramble about what I'm trying to accomplish, explain what I've already tried, mention specific requirements, and give the chatbot all the context it needs without worrying about how long my prompt is getting. This has been especially useful when I'm working with coding tools like Claude Code, where leaving out a seemingly minor detail can make a pretty big difference to the final result.

It's also great for brainstorming and planning. I can get anything and everything on my mind out as quickly as it comes to me, without having to worry about organizing my thoughts into a perfectly structured prompt first. I can jump between ideas, go off on tangents, and remember something important halfway through without losing my train of thought.

The best part is that I don't have to spend time figuring out how to phrase everything. I can simply explain what I'm thinking, even if it's a little all over the place, and let the chatbot make sense of it. More often than not, I end up giving it far more context than I would've bothered typing out manually.

Voice Mode helps me figure things out as I go

Feels less like prompting and more like brainstorming

While dictation has made it much easier to get my thoughts across to AI, I've found Voice Mode useful for an entirely different reason. Sometimes, I don't necessarily have a specific question in mind or even know exactly what I'm trying to accomplish. I just have a rough idea that I want to explore, and that's where having an actual conversation with AI comes in handy.

For example, there are times when I have a very vague idea for an article I'd like to write, but I'm not entirely sure which direction to take it in. Instead of asking a chatbot to come up with an outline or suggest a bunch of ideas, I can switch to Voice Mode and ask it to challenge my thinking. I can have it ask me questions about the topic, point out angles I might have overlooked, or question some of the assumptions I'm making.

What I particularly like about this is that I have to actually think about the questions and respond to them in real time. One question might lead me down a completely different path, and before I know it, I've explored angles that probably wouldn't have crossed my mind otherwise. It feels less like asking AI to brainstorm for me and more like having someone to bounce ideas off.

I've found this approach equally useful when preparing for presentations and interviews. Whether I'm the one interviewing someone or the person being interviewed, I can have the chatbot play the other role and ask me questions I might not have anticipated.

Another place where I've found Voice Mode incredibly useful is studying. Google's NotebookLM (now Gemini Notebook) recently introduced a feature called Live Voice Chat, which essentially lets you have a real-time conversation with your uploaded sources.

Since I already use NotebookLM extensively for university, I have notebooks filled with lecture slides, readings, and notes. Instead of having to type out a question every time I'm struggling to understand something, I can simply ask it out loud and get an immediate spoken response.

What I particularly like is that I can interrupt the chatbot halfway through an explanation if something doesn't make sense, ask it to explain a concept differently, or immediately follow up with another question.

It changes things in ways you wouldn't expect

At the end of the day, I'm still using the exact same AI tools and models I was using before this experiment. You could argue that replacing my keyboard with my voice is just a fancier way of doing the same thing, and that it shouldn't really make much of a difference.

However, I've come to realize that how you interact with AI can be just as important as which AI tool you're using. There's a phrase that's thrown around a lot when it comes to AI: "garbage in, garbage out." Essentially, the quality of what you get out of an AI tool depends heavily on the quality of what you put into it.

With dictation, I'm giving the chatbot far more context than I would've bothered typing out, simply because speaking my thoughts aloud takes so much less effort. With Voice Mode, I'm able to explore ideas through conversations, ask questions I hadn't initially thought of, and work through problems instead of expecting a single prompt to give me the perfect answer.

Neither approach magically makes the underlying AI model any better. What they do is help me get more out of the capabilities that were already there!