Getting a useful answer from AI shouldn't require learning a new language — but for most people, it does. A 2025 survey of over 1,000 AI users by Rev.com and Centiment found that 34% of respondents identified "phrasing requests in a way the AI understands" as their single biggest challenge when using AI tools — ahead of knowing the right level of detail to include (32%) and tailoring instructions for specific outputs (26%). Only 17% said they never have to rewrite their prompts to correct for false or inaccurate information.
These aren't beginner problems. The same survey found that heavy AI users — those spending six or more hours per week with AI tools — were actually more likely to struggle with phrasing than casual users: 42% versus 33%. The more you use AI, the more aware you become of how much your results depend on how you ask, not just what you ask.
We call this hidden effort the Prompting Tax.
What Is the Prompting Tax?
The Prompting Tax is the cognitive work you do before AI even starts helping you — the time spent figuring out how to phrase your request, what context to include, how much detail to provide, and what format to specify. It's the gap between what you mean and what you have to type to get the AI to understand what you mean.
The concept draws on a well-documented usability challenge. Researchers at the Nielsen Norman Group have described an "articulation barrier" in AI interactions — the difficulty users face translating what they want into language the AI can act on. As NNG's research notes, creating an effective prompt is "cognitively taxing," and users "may not even know or be able to recall the right words" for what they need. While that research focused on image generation, the principle applies to any AI interaction: the burden of clear communication falls entirely on the user.
This barrier is invisible in most discussions about AI. We talk about what AI can do, but rarely about what you have to do to get it there.
Why Does a Simple Question Need So Much Work?
Here's a simple example. Say you ask an AI tool:
"Help me plan a trip to Italy."
You'll get an answer. But it probably won't be very useful, because the AI doesn't know when you're going, how long you have, what your budget is, or what kind of experience you're looking for. Without that context, it has to guess — and its guesses will be generic at best.
A more effective version might look like this:
"Help me plan a 14-day trip to Italy this July. I want to spend time in Rome and the Amalfi Coast, prefer a mix of history and relaxation, and have a moderate budget. Suggest an itinerary with key sights, travel tips, and a rough daily cost breakdown."
The second prompt gives the AI enough to work with. But notice what it required from you: anticipating what the AI needs to know, organizing that information clearly, specifying the output format, and doing all of this before the conversation has even started.
Now scale that effort to every interaction. Need help drafting an email? You'd better specify the tone, audience, context, and length upfront. Troubleshooting a problem? Include every relevant detail in your first message, because the AI probably won't ask follow-up questions. The Rev.com survey found that users who write very long prompts are 32 times more likely to need frequent revisions for AI hallucinations compared to those who use single-sentence prompts — suggesting that adding more detail doesn't necessarily lead to better results. Complexity compounds the problem rather than solving it.
Why Doesn't Experience Make Prompting Easier?
You might expect that regular AI users would get better at prompting over time. The data says otherwise.
The same Rev.com survey found that heavy AI users (six or more hours per week) reported 7% more prompting problems across the board than casual users. Users who spend the most time with AI aren't mastering the prompting skill — they're discovering more of its failure modes. They encounter more edge cases, more hallucinations, and more situations where the AI misinterprets nuance.
Each generation also struggles differently. According to the survey, Gen Z users struggle most with matching the AI's tone and style (35%). Millennials wrestle with providing the right level of detail (33%). Gen X and Boomers have the most difficulty simply phrasing requests in a way the AI understands (33% and 34%, respectively). The prompting challenge isn't one problem — it's a set of related problems that shift depending on who you are and what you're trying to do.
Meanwhile, 28% of frustrated users say their top challenge is "having to rewrite the same prompt multiple times." That's not a learning curve — it's a Retry Loop, where the tool's inability to understand you creates a cycle of rephrasing and retrying that eats up the time AI was supposed to save.
What Does the Prompting Tax Actually Cost?
The most obvious cost is time. While 77% of AI users get a satisfactory answer in under two minutes, that number drops to just 50% for heavy users. Heavy users are 10 times more likely than casual users to spend over 11 minutes wrestling with AI before reaching a satisfactory response. The users who rely on AI the most are the ones spending the most time working around its limitations.
But the deeper cost is cognitive. Every interaction where you have to think about how to talk to the AI is an interaction where your mental energy is going to the wrong place. Instead of thinking about your actual problem — planning the trip, writing the email, solving the challenge — you're thinking about prompt structure. That's extraneous cognitive load: effort that doesn't contribute to your goal but is required by the tool.
This matters because AI was supposed to reduce mental effort, not redirect it. If the tool requires you to become skilled at using it before it becomes useful, it hasn't eliminated complexity — it's just moved it from one place to another.
What Would AI Without a Prompting Tax Look Like?
The alternative isn't a longer tutorial on prompt engineering. It's AI that doesn't need you to be good at prompting in the first place.
Imagine saying what's on your mind in your own words — the way you'd explain something to a thoughtful friend — and having the AI do the work of figuring out what you need. If your request is ambiguous, it asks a clarifying question instead of guessing. If you've left out important context, it surfaces what's missing rather than filling in the blanks with assumptions that might be wrong.
This is the difference between a tool that requires careful instructions and a companion that meets you where you are. The Nielsen Norman Group puts it clearly: "The power of LLM-based AI features is largely mediated by users' ability to write good prompts...users don't always know what they are looking for or what the AI is fully capable of." The implication is that the burden shouldn't rest entirely on the user — the system should help bridge the gap.
Like a Friend AI is built around one idea: the AI should do the work of understanding, so you don't have to do the work of explaining. Say what's on your mind in your own words, and let the AI figure out what you need — asking the right follow-up questions instead of expecting you to front-load every detail. A portion of our profits supports global causes, because removing barriers should be a principle, not just a product feature.
Frequently Asked Questions
What is the Prompting Tax?
The Prompting Tax is the hidden cognitive effort users spend figuring out how to phrase, format, and structure their requests so that AI tools produce useful responses. A 2025 survey by Rev.com and Centiment found that 34% of AI users say phrasing prompts is their biggest challenge, and only 17% never have to rewrite prompts to correct for inaccurate information. The Prompting Tax represents the gap between what you mean and what you have to type to get the AI to understand what you mean.
Why doesn't using AI more make prompting easier?
Research shows the opposite is true. The same Rev.com survey found that heavy AI users (six or more hours per week) reported 7% more prompting problems than casual users and were more likely to struggle with phrasing — 42% versus 33%. Experience reveals more failure modes rather than eliminating them.
What is the articulation barrier in AI?
The articulation barrier, described by the Nielsen Norman Group, is the difficulty users face translating what they want into language the AI can act on. Creating effective prompts is "cognitively taxing," and users often lack the vocabulary or framework to express their needs in a way the AI can process — even when they know exactly what they want.
How is an intuitive AI companion different from a chatbot that requires prompting?
An intuitive AI companion absorbs the complexity of communication rather than passing it to the user. Instead of requiring carefully structured prompts, it works to understand your intent from natural conversation — asking clarifying questions when needed rather than guessing when information is missing.
Ready to stop crafting prompts and start getting answers? Join the Like a Friend AI waitlist for early access and a lifetime 10% discount as a founding member. We're building AI that does the work of understanding — so you can spend your time on the things that actually matter to you.