50 ChatGPT Prompts for Side Hustlers (Tested 2026)

The "ChatGPT prompts for X" search results are mostly prompt-pack landing pages. Many of the prompts in those packs sound impressive but produce mediocre results — partly because they were generated by other AI as filler, partly because they don't reflect actual workflows. The list below is different. Every one of these 50 prompts has been used in real side-hustle work — client copy, niche research for blog posts, ghostwriting drafts, customer-support replies, product positioning. J.R. compiled them from 6 months of running AI-augmented service work, removed the ones that didn't reliably produce good output, and kept the 50 that earned their place. Each prompt has a one-sentence note explaining what makes it work, so you can adapt the structure to situations these specific prompts don't cover.

How to actually use these prompts

Three operating principles first, because the prompts work much better with them than without.

Replace the bracketed placeholders with specific details. Every prompt has placeholders like [product/service], [audience], [pain point]. Generic placeholders produce generic outputs. The more specific your replacement, the better the output. "Email subject line for a productivity app" produces mush; "Email subject line for a focus-timer app aimed at ADHD freelancers who keep abandoning timer apps after week two" produces something usable.

Treat the first output as a draft, not a deliverable. None of these prompts produce ready-to-ship output. They produce a strong starting point that needs human editing for tone, accuracy, and context. The 70-30 rule: AI does 70% of the work, you do the final 30% that makes it actually good.

Verify factual claims. Especially for niche research and competitive analysis prompts, AI confidently produces fabricated statistics, misattributed quotes, and citations that look real but aren't. Fact-checking is the most-skipped, most-important step. We covered this in detail in how to make money online with AI.

Category 1: Marketing copy (10 prompts)

Headlines, ad copy, landing pages, email subject lines — the highest-volume use case in side-hustle work. These prompts pair well with the writing-and-editing services we cover in our AI side hustles weekend playbook.

01. The headline variants prompt

Write 10 headline variants for [product/service] aimed at [specific audience]. 
Mix these formats: how-to, listicle, question, benefit-driven, problem-aware. 
Tone: [conversational / professional / playful]. 
Avoid clichés like "ultimate guide" and "everything you need to know."

Why it works: Asking for variants forces the model to explore the space rather than commit to its first idea. The format mix prevents repetition; the cliché ban removes the worst defaults.

02. The pain-point opener

List 5 distinct pain points that [audience] experiences when [doing task]. 
For each, write a 2-sentence opener for a marketing page that names the pain 
without exaggerating, then transitions naturally to a solution.

Why it works: Forces the model to articulate the user's mental state before pitching. Specifying "without exaggerating" stops the dramatising AI tends toward.

03. The before/after page block

Write a "before vs after" content block for [product] in 60 words total. 
Format: [Before] paragraph, [After] paragraph. The Before should describe 
[audience]'s current frustrating reality. The After should describe their 
state once they've used [product] for [time period]. Concrete details, 
no vague phrases like "easier" or "better."

Why it works: The word limit forces concision; the "concrete details" instruction blocks generic puffery.

04. The skeptical-customer rebuttal

A skeptical customer reads my landing page and thinks: "[specific objection 
from [audience]]." Write a 100-word section that addresses this objection 
without being defensive. Acknowledge the legitimate part of their concern, 
then explain how [product/service] handles it. End with a small action they 
can take to verify.

Why it works: Naming the objection upfront builds trust. The "legitimate part" framing prevents dismissive copy.

05. The email subject line A/B set

Generate 8 email subject lines for an email about [topic] going to [audience]. 
Mix: 4 question subjects, 2 benefit-led, 2 curiosity gaps. Each under 50 
characters. Tone: [conversational/professional]. Don't use clickbait.

Why it works: Character constraint matches inbox preview reality; format mix gives A/B test material.

06. The cold-outreach personalisation

I'm reaching out to [name/role] at [company]. They [public detail or recent 
post]. Write a 3-sentence cold email opener that references their work 
specifically, makes one observation that adds value, and ends with a 
soft CTA. No flattery. No "I just wanted to reach out." No "I came across 
your profile."

Why it works: The forbidden phrases eliminate cold-outreach clichés. The "value observation" forces genuine engagement.

07. The benefit pyramid

For [feature] of [product], create a 3-level benefit pyramid: 
1) the feature itself, 2) what it lets the user do, 3) what that means 
for their life or work. Each level under 20 words. Use plain language.

Why it works: Forces feature → benefit → meaning translation, which is the missing layer in most product copy.

08. The landing-page CTA generator

Generate 6 CTA button variants for a [product type] page targeted at [audience]. 
Mix: 2 specific-action CTAs, 2 outcome-focused CTAs, 2 risk-reversal CTAs. 
2-4 words each. No "Click here" or "Learn more."

Why it works: Forbidden defaults force specificity. Format diversity gives material to test.

09. The ad copy condense

I have this longer description: "[paragraph]". Condense it into a Facebook 
ad that's under 90 characters in the headline and under 125 characters in 
the primary text. Same core promise, no hype words, no exclamation marks.

Why it works: Constraints match the platform reality. "No hype words" blocks the common AI tendency toward "amazing!" and "incredible!"

10. The objection-mapping audit

Read this landing page copy: "[paste copy]". List the 5 most likely 
unstated objections a [audience] reader would have at the moment of 
clicking the CTA. For each, suggest one sentence that could be added 
to the page to address it without breaking flow.

Why it works: Reverse-engineering objections from existing copy is one of the highest-value AI uses for marketing work.

Category 2: Niche research (10 prompts)

Background research for blog posts, market research for pitches, competitor analysis for client work. Always verify the factual claims — niche research is where AI hallucinates most confidently.

11. The competitor landscape

List the top 7 [type of business] competing for [audience] in [region]. 
For each, give: 1) their core positioning, 2) their pricing model, 
3) one perceived weakness based on customer reviews. Mark any claim 
where you're inferring versus citing — I'll verify the cited claims.

Why it works: "Mark any claim where you're inferring" gets the model to flag its own uncertainty, which makes verification far cheaper.

12. The audience persona builder

Build a detailed persona for [audience description]. Include: typical day, 
top 3 frustrations, what they currently use to solve [problem], what would 
make them switch, what would prevent them from switching. Be specific. 
Don't use generic personas.

Why it works: Pushing for specificity forces the model past stock personas. The "switch / prevent switching" framing surfaces real friction.

13. The pricing comparison

Compare the pricing tiers of [3-5 competitor products]. For each: list 
the entry price, what's included, what's locked behind the next tier, 
and what kind of customer the entry tier is designed for. Format as a 
table. Note where pricing has changed in the past 12 months if you know.

Why it works: Tables force structured comparison; explicit "what's locked" clarifies tier strategy.

14. The keyword cluster

For the topic [topic], generate 30 keyword variations grouped into 5 
clusters by search intent (informational, navigational, commercial, 
transactional, local). Show the underlying intent for each cluster.

Why it works: Clustering by intent matches how Google ranks content — better than flat keyword lists.

15. The trend check

What are the 5 most discussed [topic] trends in [audience]'s online spaces 
over the past 6 months? For each, give: 1) the trend, 2) where it's being 
discussed, 3) whether it's accelerating or decelerating. Caveat anything 
you're not certain about.

Why it works: "Caveat anything you're not certain about" elicits explicit hedging instead of false confidence.

16. The objection inventory

List the 10 most common objections [audience] has when considering 
[product/service]. Group them by type: trust objections, cost objections, 
fit objections, timing objections, and skill objections. Don't make any up.

Why it works: Grouping by objection type gives you a strategic map of what to address in copy and content.

17. The opposite-niche scan

For my niche [niche], list 5 adjacent or unexpected niches whose 
audiences might also benefit from [product/service]. For each, explain 
the connection and suggest one entry-point content topic.

Why it works: AI is unusually good at lateral connections — it sees patterns humans miss when too close to a niche.

18. The 5-why drill

[Audience] doesn't currently buy [product/service]. Run a 5-why drill: 
Why don't they buy it? Then why is that? Repeat 4 more times to surface 
the deepest underlying reason. Don't repeat reasons across levels.

Why it works: Forces the model past the surface answer to the structural cause. The "don't repeat" instruction prevents loops.

19. The market-size sanity check

Estimate the rough market size for [product/service] in [region]. Show 
your reasoning: total addressable population, percentage with the relevant 
need, willingness to pay, percentage you can realistically reach. Be 
explicit about which estimates are weak.

Why it works: Showing reasoning makes the estimate auditable and surfaces faulty assumptions.

20. The "what's missing" scan

I'm researching [topic]. I've already covered [list]. What important 
sub-topics or angles am I missing that [audience] would care about? 
Don't repeat what I've covered.

Why it works: Negative-space thinking surfaces gaps in your existing coverage that you'd miss self-reviewing.

Category 3: Content drafting (10 prompts)

Blog posts, newsletter sections, social posts, video scripts. The first-draft volume use case.

21. The blog-post outline

Outline a 1,800-word blog post on [topic] for [audience]. Use this 
structure: hook, the honest framing, 3-5 main sections with sub-points, 
common mistakes section, FAQ, and a "what to do next" close. Each section 
header should preview a specific takeaway.

Why it works: Specifying the structure matches what readers expect from how-to content. Specifying header content prevents vague outlines.

22. The hook variations

Write 5 different opening paragraphs for an article about [topic]. 
Each should use a different hook style: 1) a counterintuitive observation, 
2) a specific personal-sounding scenario, 3) a contrarian claim, 4) a 
data point with implication, 5) a direct address to the reader's frustration. 
Each under 80 words.

Why it works: Format constraint forces variety; word limit forces concision.

23. The newsletter section

Write a 200-word newsletter section about [topic]. Voice: [conversational 
/ analytical / playful]. Include one specific example, one practical 
takeaway, and end with a single sentence that opens curiosity for the 
next email.

Why it works: "Single sentence opening curiosity" directly addresses the email-marketing engagement metric that matters most.

24. The "explain like I'm 12" pass

I have this technical explanation: "[paste]". Rewrite it for a 12-year-old 
who is curious but knows nothing about [field]. Use one analogy. Don't 
patronise. Don't oversimplify to the point of losing accuracy.

Why it works: "Don't patronise" stops the kindergarten-teacher voice AI defaults to. The accuracy clause prevents over-simplification.

25. The Twitter-thread skeleton

Outline a 9-tweet thread about [topic] for [audience]. Tweet 1: hook 
that promises specific value. Tweets 2-7: the core points, one idea each. 
Tweet 8: the surprising/contrarian angle. Tweet 9: CTA. Each under 240 
characters. Don't use hashtags.

Why it works: Mapped to actual platform reality. "Surprising/contrarian angle" tweet 8 is the highest-engaging slot in tested thread structures.

26. The voice-match draft

I'm writing in this voice: "[paste 2-3 sample paragraphs of your or 
your client's writing]." Now write 300 words about [topic] in that 
exact voice. Match sentence length, vocabulary, rhythm, and any 
characteristic phrases.

Why it works: Voice-matching is genuinely hard but works well with explicit samples. The "rhythm" instruction is the under-used lever.

27. The argument-structure check

Read this draft: "[paste]". Identify the central claim, the supporting 
evidence, and any logical gaps where the reader might disagree. Don't 
fix the gaps; just point them out.

Why it works: "Don't fix" preserves your thinking. Identifying gaps is more useful than auto-patching them.

28. The video-script timing

Convert this written content: "[paste]" into a video script for a 
2-minute spoken video. Format with [TIMING] markers, natural speaking 
pauses, and a hook in the first 8 seconds. Voice: conversational, 
not scripted-sounding.

Why it works: Timing markers help recording. "Not scripted-sounding" blocks the formal AI tone that doesn't work in video.

29. The how-to step list

Write step-by-step instructions for [task]. Each step under 25 words. 
Include the failure mode for each step (the most common way someone 
might mess up). Don't number the steps yet.

Why it works: Failure modes are usually missing from instructions and add disproportionate value to readers.

30. The 50/50 split test

Take this draft: "[paste]". Generate a second version that makes the 
opposite stylistic choice — if the original is conversational, make 
it analytical; if formal, make it playful; if long-sentence, make it 
short-sentence. Same content, different voice.

Why it works: Forced contrast surfaces which voice the content actually wants.

Category 4: Customer DMs and support (10 prompts)

Replies, follow-ups, refund handling, onboarding messages. The voice-of-the-business work.

31. The empathetic acknowledge

A customer wrote this complaint: "[paste]." Draft a 4-sentence reply 
that 1) acknowledges their specific frustration, 2) takes ownership 
without over-apologising, 3) explains what happened in plain language, 
4) offers a concrete next step. Don't use "I understand" or "I apologise."

Why it works: Forbidden phrases force specificity. The 4-sentence structure matches what most upset customers want.

32. The polite no

I need to decline [request type] from [type of customer]. Draft a 
3-sentence response that's warm but unambiguous. Explain the reason 
briefly. Don't leave room for renegotiation. End on a note that 
preserves the relationship.

Why it works: "Don't leave room for renegotiation" is the missing piece in most polite-no templates.

33. The refund decision

A customer requesting a refund [their context]. Refund policy is 
[your policy]. Should I refund? Walk me through the considerations: 
1) policy compliance, 2) customer-acquisition value of refunding, 
3) signal it sends to other customers, 4) my actual feeling. Then 
recommend.

Why it works: The explicit "actual feeling" surfaces the human element that should be part of the decision.

34. The follow-up sequence

Draft a 3-email follow-up sequence for someone who [showed interest 
but didn't convert]. Email 1 (1 day later): low-pressure reminder. 
Email 2 (4 days later): address one likely objection. Email 3 (10 
days later): clear last-chance signal. Each under 100 words.

Why it works: Cadence matches what's tested across email-marketing benchmarks. Word limits keep the sequence sustainable to send.

35. The onboarding kickoff

Write the first email a new [customer type] receives when they buy 
[product]. Include: warm welcome (1 sentence), the single most 
important thing they should do first (3 sentences with specifics), 
a soft expectation-setting line, and a real reply-friendly invitation 
to ask questions. Under 200 words.

Why it works: "Single most important thing" forces prioritisation. The "real reply-friendly" invitation generates surprisingly useful customer feedback.

36. The technical-issue triage

Customer describes this issue: "[paste]." Generate 5 follow-up questions 
I should ask to diagnose. Order them by which question, if answered, 
would eliminate the most possibilities. Don't ask things that are 
already in their description.

Why it works: Information theory: ask the question that splits the possibility space best.

37. The boundary-respect close

I had a 3-message conversation with a prospect that ended with them 
saying [their exact words]. Draft a final reply that respects whatever 
they expressed (interest, hesitation, or decline) without pushing 
further. Maximum 2 sentences.

Why it works: "Without pushing further" is the explicit instruction that generates non-pushy replies; AI defaults toward continued sales pressure.

38. The expectation reset

A client thought [their misunderstanding]. The actual scope is [reality]. 
Draft a reply that resets the expectation kindly, doesn't blame anyone, 
and offers two concrete options for moving forward.

Why it works: "Doesn't blame" prevents the reply from becoming defensive. Two options give agency without being open-ended.

39. The thank-you that says something

Write a thank-you message to a customer who [specific positive action]. 
Don't use the words "amazing," "thank you so much," "we appreciate," 
or anything that could be auto-generated. Be specific about what they 
did and what it means.

Why it works: Forbidden generic phrases force the model toward genuine specificity.

40. The FAQ generator from real questions

Here are 8 actual customer questions I've received: [paste]. Identify 
the 4 underlying themes. For each theme, write a clear FAQ entry: 
question, answer, and a sentence pointing to where they can find more 
detail.

Why it works: Theme-based FAQs are more useful than question-by-question lists; AI is good at thematic clustering.

Category 5: Product naming and positioning (10 prompts)

Naming products, writing positioning statements, building brand language. Lower-volume but high-leverage work.

41. The naming brainstorm

I'm naming a [product type] for [audience]. The product does [function]. 
Generate 30 name candidates across 5 styles: 1) descriptive, 2) metaphor, 
3) invented word, 4) compound word, 5) sound-based. Note any names that 
might have brand conflicts I should check.

Why it works: Format diversity prevents the model from converging on one style. The conflict-check note saves a verification step.

42. The positioning statement

Write a positioning statement for [product]. Format: "For [audience] 
who [need/pain], [product] is a [category] that [unique benefit] 
unlike [competitor or alternative], we [differentiator]." Keep it 
to one tight paragraph.

Why it works: The structured template, originally from Geoffrey Moore, forces clarity on every component.

43. The brand-voice document

Generate a brand-voice document for [brand]. Include: 3 voice attributes 
with definitions, 5 "we sound like / we don't sound like" pairs, 3 example 
sentences in voice for: marketing copy, customer support, social media. 
Brand vibe: [description].

Why it works: The "we sound like / we don't sound like" pattern is the single most useful voice-document format.

44. The tagline variations

Write 12 tagline candidates for [product]. Mix: 4 benefit-led, 4 mystery/ 
intrigue, 4 declarative. Each under 7 words. None should use words like 
"empower," "elevate," or "unleash."

Why it works: Word count constraint mimics tagline reality. Forbidden words eliminate the worst defaults.

45. The audience adjective

[Audience] reads about [product]. List 20 adjectives they'd want to 
genuinely associate with [product] — not ones we'd want them to think, 
but ones they'd say to a friend. Then list 10 adjectives that would 
make them lose interest immediately.

Why it works: The friend-language framing surfaces honest reactions. The negative list is often more useful than the positive.

46. The category positioning

[Product] sits in the [broad category]. Within that, define a tighter 
sub-category that we should claim. Argue for the choice: who else is 
already there, why we'd win in that sub-category, and what the audience 
already calls this kind of thing.

Why it works: Forces strategic thinking about category framing rather than default broad positioning.

47. The packaging headline

Write 5 packaging-headline options for [product]. Each must: tell the 
buyer what it is in 4 words or fewer, work at a glance, not require 
reading the rest of the package. Audience: [audience].

Why it works: Word constraint matches how shoppers actually read packaging — at a glance.

48. The about-page voice

Write a 200-word About page section for [brand]. Don't say "founded in," 
"passionate about," or "mission to." Instead, tell a specific story 
about what wasn't working before [brand] existed, and what we believe 
should change.

Why it works: Forbidden phrases gut the worst About-page clichés. The "what wasn't working" framing produces stories with point of view.

49. The 3-word vibe

Describe [product]'s vibe in exactly 3 words. Then explain why those 
3 words, what they exclude, and how they should show up in copy, 
visuals, and customer interaction.

Why it works: The 3-word constraint forces hard prioritisation. The "what they exclude" is the most useful follow-up.

50. The launch-message ladder

Generate a 4-message launch sequence for [product]. Audience: [audience]. 
Message 1: pre-launch tease. Message 2: launch announcement. Message 3: 
mid-launch case study or social proof. Message 4: last-chance close. 
Each tied to a specific date in a [duration] launch window. Each 
under 150 words.

Why it works: Calendar-tied sequences are how launches work in practice. Word limits prevent each message from sprawling.

Verifying outputs: the four-step check

Every prompt above produces a draft. Before any output ships:

Check factual claims. Statistics, citations, named entities, "research shows" claims. Verify each against a real source.

Check voice fit. Read the output as the audience would. Does it sound like your brand or your client's, or like a competent stranger?

Check internal consistency. Long outputs occasionally contradict themselves. Skim for anything that contradicts an earlier sentence.

Check the "AI tells." Em-dashes used aggressively, lists of three when two would do, headlines with "ultimate" or "essential," sentences that begin with "In today's [adjective] world." These are AI defaults that signal AI authorship to readers who've grown sensitive to them.

Frequently asked

Do these prompts work with Claude or Gemini?

Yes, with minor variations. Claude tends to be more verbose; you may want to add "be concise" to outputs. Gemini occasionally needs more explicit structure cues. The core prompt logic transfers across all major models.

Should I use the free or paid tier?

Free tier (GPT-3.5-class or Claude Haiku) is fine for the simpler prompts (categories 4 and 5). For categories 1-3 — especially research and content drafting — the paid tiers (GPT-4-class or Claude Sonnet+) produce noticeably better outputs. A $20/month subscription pays for itself if you're using these prompts in client work.

How do I adapt these prompts to my specific niche?

The bracketed placeholders are the adaptation points. Replace them with niche-specific details: "[audience]" becomes "first-time home buyers in their 30s" instead of just "buyers." The more specific your replacements, the better the output.

What if the output is mediocre?

First check: are the placeholders specific? Generic placeholders produce generic output. Second: did you give the model context it would need? Voice samples, audience details, product specifics. Third: try running the same prompt again — model outputs vary, and the second attempt is sometimes meaningfully better.

Can I sell prompt packs based on these?

The prompts are free for personal and client use. Don't republish them as a paid prompt pack — the licensing isn't designed for that, and the prompt-pack market is saturated anyway. If you want to build a paid product around prompts, niche it heavily (industry-specific, role-specific) rather than re-packaging generic ones.

Do these prompts work for non-English content?

Mostly yes. Output in non-English languages is generally good for major languages (Spanish, French, German, Japanese, Mandarin) and weaker for less-resourced languages. For non-English output, also include "respond in [language]" explicitly in the prompt.

What to do next

Pick three prompts from the list above that match a real piece of work you have in your queue this week. Run them today, with specific placeholders. The single best way to internalise this material is to use it on actual work, not to read it speculatively.

For more on AI in side-hustle workflows, our how to make money online with AI covers the broader strategic landscape, and 12 AI side hustles you can start this weekend covers specific projects where these prompts plug in. If you're choosing between AI tools, Notion AI vs ChatGPT walks through where each one wins.

Drop your email below to get the 50 ChatGPT Prompts Pack — all 50 prompts above as a printable PDF with notes, plus the verification checklist. Free.