# How do you write a prompt that changes the output for client work?

**URL:** <https://humanizeaiforum.com/t/how-do-you-write-a-prompt-that-changes-the-output-for-client-work/229>\
**Category:** Tips & Best Practices\
**Created:** [September 22, 2026, 10:59pm UTC](https://humanizeaiforum.com/t/how-do-you-write-a-prompt-that-changes-the-output-for-client-work/229 "2026-09-22T22:59:05Z")\
**Posts on this page:** 6\
**Page:** 1

<div class="post-metadata">

**Author:** ![maya.glitch](https://yyz1.discourse-cdn.com/flex033/user_avatar/humanizeaiforum.com/maya.glitch/32/39_2.png) [@maya.glitch](https://humanizeaiforum.com/u/maya.glitch)\
**Post date:** [September 22, 2026, 10:59pm UTC](https://humanizeaiforum.com/t/how-do-you-write-a-prompt-that-changes-the-output-for-client-work/229/1 "2026-09-22T22:59:05Z")

</div>

specific prompt changes that made a real difference in output quality for client work: not ‘be more specific’ as general advice, actual changes that moved the needle. i’m putting together a workshop and want the concrete ones.

---

<div class="post-metadata">

**Author:** ![ErnestPrice](https://yyz1.discourse-cdn.com/flex033/user_avatar/humanizeaiforum.com/ernestprice/32/24_2.png) [@ErnestPrice](https://humanizeaiforum.com/u/ErnestPrice)\
**Post date:** [September 23, 2026, 3:08pm UTC](https://humanizeaiforum.com/t/how-do-you-write-a-prompt-that-changes-the-output-for-client-work/229/2 "2026-09-23T15:08:29Z")

</div>

adding a negative constraint list changed my output more than any positive instruction. ‘do not summarize at the end, do not use passive voice in the first sentence, do not use noting-as-filler phrases’ cuts the most common AI tells before they appear.

---

<div class="post-metadata">

**Author:** ![rough.Matrix](https://yyz1.discourse-cdn.com/flex033/user_avatar/humanizeaiforum.com/rough.matrix/32/35_2.png) [@rough.Matrix](https://humanizeaiforum.com/u/rough.Matrix)\
**Post date:** [September 24, 2026, 2:06pm UTC](https://humanizeaiforum.com/t/how-do-you-write-a-prompt-that-changes-the-output-for-client-work/229/3 "2026-09-24T14:06:30Z")

</div>

specifying the audience in concrete terms. not ‘professional audience’ but ‘a marketing director at a mid-size B2B software company who reads at speed and skims first.’ the model adjusts assumed knowledge, vocabulary, and detail level. same topic, noticeably different output.

---

<div class="post-metadata">

**Author:** ![BluestHues1986](https://yyz1.discourse-cdn.com/flex033/user_avatar/humanizeaiforum.com/bluesthues1986/32/30_2.png) [@BluestHues1986](https://humanizeaiforum.com/u/BluestHues1986)\
**Post date:** [September 25, 2026, 8:13pm UTC](https://humanizeaiforum.com/t/how-do-you-write-a-prompt-that-changes-the-output-for-client-work/229/4 "2026-09-25T20:13:29Z")

</div>

pasting two or three sentences from your own writing that you’re happy with before asking for the draft. not describing your style, showing it. ‘write in a register consistent with these sentences’ produces more useful output than ‘write conversationally.’

---

<div class="post-metadata">

**Author:** ![d.samuel.brook.29](https://yyz1.discourse-cdn.com/flex033/user_avatar/humanizeaiforum.com/d.samuel.brook.29/32/37_2.png) [@d.samuel.brook.29](https://humanizeaiforum.com/u/d.samuel.brook.29)\
**Post date:** [September 26, 2026, 1:51am UTC](https://humanizeaiforum.com/t/how-do-you-write-a-prompt-that-changes-the-output-for-client-work/229/5 "2026-09-26T01:51:29Z")

</div>

for product content: ‘write this as a recommendation to a friend who asked why you bought it’ produces different material from ‘write a product description.’ the first frame creates voice. the second frame creates a spec.

---

<div class="post-metadata">

**Author:** ![maya.glitch](https://yyz1.discourse-cdn.com/flex033/user_avatar/humanizeaiforum.com/maya.glitch/32/39_2.png) [@maya.glitch](https://humanizeaiforum.com/u/maya.glitch)\
**Post date:** [September 27, 2026, 10:34am UTC](https://humanizeaiforum.com/t/how-do-you-write-a-prompt-that-changes-the-output-for-client-work/229/6 "2026-09-27T10:34:30Z")

</div>

synthesizing from the replies and my own testing for the workshop:

changes that move the needle:

1. negative constraints, list 5-8 things you don’t want. these are more specific than positive instructions and cut AI defaults fast.

2. concrete audience, not ‘professional’ but ‘a [role] at a [company type] who [specific situation].’

3. style example sentences, paste 2-3 sentences you wrote that match the target register. more reliable than describing style abstractly.

4. frame the task differently: ‘recommend’ instead of ‘describe’, ‘explain to a practitioner who’s made this mistake’ instead of ‘provide an overview.’

5. after a [Walter Writes](https://walterwrites.ai/ai-humanizer/) pass on flagged sections, feed the specific detector feedback back into the next prompt. ‘the transitions between sections are still scoring high, rewrite those specifically.’ targeted iteration beats broad re-generation.