A Prompt That Adds Contractions, Fragments, and Casual Language to AI Text

Ask a model to make something "more casual" and it will probably change "utilize" to "use" and call it done. The text is still stiff, still hedged, still reads like a memo. That's not a failure of the model's ability. It's a failure of the instruction. Vague requests produce vague results. If you want contractions, say "replace every instance of 'do not' with 'don't'." If you want fragments, say "allow short incomplete sentences after rhetorical questions." Specific instructions get specific changes.

This guide gives you a copy-paste prompt that enumerates exactly those changes, explains why that specificity matters, and shows you what the output looks like before and after.

Why "Make It More Casual" Fails as an Instruction

"Casual" means different things depending on who you ask. For a model trained to be helpful and precise, casual might mean removing the word "moreover" once. That's technically a shift in register. It just isn't the shift you wanted.

The underlying problem is that register sits on a spectrum, and the model has no way of knowing where on that spectrum you're aiming. Are you writing a Slack message to a coworker? A product page for a streetwear brand? A newsletter that sounds like it's from a real person? Each one calls for a different level of informal language, and "more casual" doesn't distinguish between them.

Prescriptive instructions sidestep this ambiguity entirely. Instead of asking for a feeling, you're asking for a set of mechanical changes: swap these specific phrases, allow this specific sentence structure, delete this specific type of hedge. The model is good at following rules. Give it rules.

This is also why prompts for specific writing patterns (see short one-line prompts that instantly improve AI writing) tend to outperform broad rewrite requests. One targeted rule often does more than a paragraph of general direction.

The Copy-Paste Prompt

Here's a prompt you can paste directly into ChatGPT, Claude, or any other chat interface. Paste it before or after the text you want rewritten, whichever your workflow prefers.

Rewrite the following text to sound casual and natural. Apply these changes exactly:

1. Replace all contractions-avoided forms: swap "do not" → "don't", "is not" → "isn't", "cannot" → "can't", "will not" → "won't", "it is" → "it's", "they are" → "they're", "you are" → "you're", "we are" → "we're", "I am" → "I'm", "that is" → "that's", "there is" → "there's", "would not" → "wouldn't", "could not" → "couldn't", "should not" → "shouldn't", "did not" → "didn't", "does not" → "doesn't", "has not" → "hasn't", "have not" → "haven't", "was not" → "wasn't", "were not" → "weren't".

2. Allow short sentence fragments after rhetorical questions. For example: "Is it worth trying? Absolutely." Keep these to one or two words.

3. Remove modal hedges: delete phrases like "it may be worth considering", "one might find", "it is possible that", "it could be argued", "it is important to note", "it is worth mentioning". Replace them with direct statements.

4. Replace formal connectives with shorter, casual ones: swap "however" → "but", "therefore" → "so", "furthermore" → "also", "in addition" → "and", "nevertheless" → "still", "consequently" → "so".

5. Replace wordy phrases with shorter equivalents: "in order to" → "to", "due to the fact that" → "because", "at this point in time" → "now", "in the event that" → "if", "with regard to" → "about".

6. Use second-person where appropriate. Replace "one should" with "you should", "one can" with "you can".

7. Keep the meaning and all factual content intact. Do not add new information or change the structure beyond sentence-level adjustments.

[PASTE YOUR TEXT HERE]

You don't need to use every rule for every piece. If you're editing a professional newsletter rather than a product description, you might skip rule 2 (the fragments) and dial back rule 4. Treat the list as a menu, not a mandate.

For a more context-specific version built around matching your own writing style, see a ChatGPT prompt that makes copy sound like you wrote it.

Why Prescriptive Instructions Work Better

Models don't struggle with casual tone because they can't write it. They struggle because they're trying to guess what you mean by a vague adjective. "Casual" is an aesthetic judgment. A numbered list of swap instructions is a task.

There are a few reasons prescriptive prompts outperform vague ones:

They reduce interpretation variance. When the model interprets "casual," it might draw on different examples from training than you had in mind. When you say "swap 'however' for 'but'," there's nothing to interpret.

They're reproducible. If you run the same prompt twice on different pieces of text, you get the same kind of output both times. That consistency is hard to get with open-ended instructions.

They're adjustable. If the output isn't casual enough, you add a rule. If it goes too far, you remove one. You're editing a list, not rewriting a paragraph of prose direction.

They teach you something. Going through the process of writing prescriptive rules forces you to identify which specific patterns are making the text feel stiff. That's useful even outside of prompting.

The prompt recipes for making AI text casual, formal, or friendly take this further with modular rule sets you can combine for different tone targets.

Before and After

Here's an example of the prompt in action.

Original AI output:

It is important to note that in order to achieve optimal results, one should consider the various factors that may be relevant to their particular situation. Furthermore, it is possible that some individuals will find certain approaches more effective than others. It would be advisable, therefore, to experiment with multiple methods before arriving at a definitive conclusion.

After applying the prompt:

Worth knowing: to get good results, you need to think about what's specific to your situation. Some approaches will work better for you than others. So it's worth trying a few before you settle on one.

The second version says the same thing. It cuts roughly a third of the word count. And it no longer reads like it was written by a committee.

The key changes at work: "it is important to note" was cut entirely (rule 3), "in order to" became "to" (rule 5), "one should consider" became "you need to think about" (rule 6), "furthermore" became nothing (the sentence was merged), "it is possible that" was dropped in favor of a direct statement (rule 3), and "therefore" became "so" (rule 4).

None of those changes altered the meaning. All of them shifted the register.

Adapting the Prompt for Different Contexts

The base prompt is calibrated for general-purpose casual writing. You'll want to adjust it depending on where the text is going.

For social media or short-form content: Keep rules 1 through 6 but also tell the model to shorten sentences to under fifteen words where possible. Short sentences read faster and land harder in a scroll context.

For email newsletters: Keep the contractions and direct statements but drop rule 2 (the fragments). Newsletters benefit from warmth but usually need to stay grammatically complete. Fragments in a newsletter can read as rushed.

For product copy: Focus on rules 3 and 6. Cutting hedges and switching to second person will do more for product copy than fragments or shortened connectives. You want confidence, not chattiness.

For professional but approachable writing: Use only rules 1, 3, and 5. Contractions and the removal of hedges get you most of the way to human-sounding without making anything feel too informal for a business context.

If you want to go deeper on the relationship between specificity and tone in prompting, the best prompt to make AI writing sound human covers the underlying logic in more detail, including how to layer multiple rule sets.

You can also run this kind of prompt through the humanizer prompt tool on this site, which applies a similar set of targeted rules as part of a broader humanization pass.

Frequently Asked Questions

Will this prompt work the same way in every model?

Close, but not exactly. ChatGPT and Claude both follow numbered rule lists reliably, so you'll get consistent results with either. Smaller or older models may skip a rule or two, especially the more nuanced ones like rule 3 (removing modal hedges). If you notice a pattern being missed, try separating it into its own prompt and running it as a second pass.

Why does the model sometimes add new information when I ask it to rewrite for tone?

This happens when the model treats the rewrite as a content task rather than a style task. Rule 7 in the prompt above ("keep the meaning and all factual content intact") is specifically there to prevent this, but you can reinforce it by adding "do not add any information that was not in the original text" as a standalone line at the top of your prompt.

Can I use this prompt directly in a system prompt for an API integration?

Yes. Convert the numbered list into a standing instruction and set it as your system prompt. Every piece of text the model generates will then apply these rules by default without you having to paste the prompt each time. This works well if you're running a content workflow and want all output to hit a consistent casual register.

What's the difference between this approach and just editing the text myself?

Speed and scale. If you're editing a single paragraph, doing it by hand is faster. If you're editing twenty articles, a paragraph at a time, the prompt saves significant time and produces consistent results across the batch. The other difference is repeatability: a prompt applies the same rules the same way every time, which your own editing might not.

Does removing hedges ever make the writing too assertive or inaccurate?

It can, in certain contexts. YMYL content (health, finance, legal) sometimes needs hedging for accuracy, not style. "This may reduce your risk of X" is not just formal; it's factually appropriate. In those cases, skip rule 3 or apply it selectively. For most commercial and informational copy, though, the hedges are purely stylistic and removing them only makes the writing clearer.