I’ll be honest about something. Some days my prompts are two lines long. I hit enter and hope the AI can read my mind. It never can. The answer comes back bland, and I end up rewriting half of it myself.
So this post from an AI startup founder hit close to home. The author noticed something odd: writing the prompt now takes longer than getting the answer. Claude replies in seconds, but their prompts were taking minutes. And that small gap was quietly wrecking the quality of everything they got back.
Because typing felt slow, they started cutting corners. They dropped the audience. They skipped the constraints. Two lines, hit enter, hope for the best. As the original poster put it, the output matched the effort: “Generic. Safe. Rewrite-it-yourself.”
The Experiment: Same Task, Two Methods
The author runs an AI startup and prompts dozens of times a day. That shortcut was costing them a lot more than they first thought. So last week they ran a simple head-to-head test.
The task stayed the same for both rounds: a 30-day campaign to grow their LinkedIn followers. The only thing that changed was how they gave Claude the instructions.
Round 1: Typing the prompt
- Time spent: 2 minutes
- What came out: a vague paragraph with no real structure
Round 2: Speaking it into Wispr Flow, a voice-to-text tool
- Time spent: 40 seconds
- What came out: context, audience, five numbered asks, a time budget, and a list of what to rule out
- Bonus: it came back already structured, with no cleanup needed
Read that again. The spoken version took a third of the time and carried way more information. That’s the part that got me.
Typing vs Talking: The Honest Pros and Cons
Typing isn’t useless, and I don’t think the author is saying it is. Here’s how the two methods stack up when you look at them side by side.
Typing your prompts
- Pro: you can edit as you go and polish every word
- Pro: it works anywhere, even in a quiet office or a crowded train
- Con: it’s slow, so you naturally write less
- Con: the first things to get cut are context, audience, and constraints, which are exactly what the model needs
- Con: you often end up with a short, vague request and a generic answer
Speaking your prompts with voice-to-text
- Pro: it’s much faster, about 40 seconds against 2 minutes in this test
- Pro: you keep the details in because adding them costs almost nothing
- Pro: tools like Wispr Flow tidy your speech into clean, structured text
- Con: you need a spot where you can talk out loud comfortably
- Con: it takes a few tries to get used to “thinking out loud” in a clear order
Why Talking Wins
The most useful part of the post is the author’s explanation of why this works. It’s not really about the tool. It’s about friction.
You don’t skip context because you’re lazy. You skip it because typing it out is slow. Talking removes the friction, so the context stays in, and the first answer is actually usable.
I love this framing. Most of us know a good prompt needs background, a target audience, and clear limits. We just don’t bother because our fingers can’t keep up with our brains. Speaking closes that gap. You say what’s in your head, and all that rich detail actually makes it to the model.
The author ends on a line that stuck with me: the founders getting the most out of AI right now aren’t using secret models. They’re just giving the models more to work with.
How to Build a Better Spoken Prompt
So what should you actually say when you talk to the AI? The spoken prompt in the test covered five things. You can use them as a checklist:
- Context: who you are and what you’re working on
- Audience: who the output is for
- Numbered asks: the specific things you want, one by one
- Time budget: how much time or effort you can realistically put in
- What to rule out: tactics, tones, or ideas you don’t want
Here’s an example I put together using that structure. It’s not the author’s exact prompt, just a template you can say out loud and adapt:
“I run a small AI startup and post on LinkedIn a few times a week. My audience is founders and marketers who are curious about AI but busy. I want a 30-day plan to grow my followers. One, give me a weekly theme. Two, suggest three post formats. Three, write five hook ideas. Four, tell me the best times to post. Five, give me a simple way to track results. I have about 30 minutes a day for this. Rule out paid ads, engagement pods, and anything that feels like clickbait.”
Say that out loud and time yourself. It takes well under a minute. Typing it would take a lot longer, and honestly, most of us would skip half of it.
Tips to Get Started
- Talk in order: go through the five-part checklist so your speech has a natural structure
- Number your asks out loud: saying “one, two, three” helps the tool and the AI keep things organized
- Don’t over-edit: send the first spoken version and see how much better the answer is
- Start with your hardest daily prompt: the one you always rush is where you’ll notice the biggest jump
- Keep the rule-outs: telling the AI what not to do saves a whole round of rewrites
My Recommendation
If you prompt AI more than a few times a day, try speaking your prompts for one week. Keep typing for quick edits and short follow-ups. But for anything that needs real context, like campaigns, plans, or strategy, talk it out. This test made it clear: same task, a third of the time, and a first answer you can actually use.
I’ve already started doing this for my longer prompts, and the difference shows up right away. Better input, better output. It really is that simple.
Check out the full LinkedIn post to see the author’s experiment in their own words. It’s worth a read.