Finding a Profitable Niche
"I'm going to sell prompts" isn't a niche — it's a category with thousands of competitors. The sellers who actually make consistent money picked something specific enough that they could become genuinely good at it, and specific enough that a buyer searching for exactly that thing finds them instead of getting lost in a sea of generic listings.
Reading Demand Signals
What's already selling repeatedly, not just what's listed — a listing with many reviews is a real demand signal.
Search a specific use case and see how thin the results are — a popular general category with almost nothing narrow is a genuine opening.
Recurring questions in AI-focused subreddits, Discord servers, and forums — "does anyone have a prompt that actually does X" is a direct signal.
A niche tied to work you've actually done (real estate, HR, e-commerce, teaching) gives you a credibility edge no amount of research alone provides.
Broad vs. Right-Sized Niches
Too Broad (oversaturated)
“Fantasy portrait prompts,” “write me a blog post,” “LinkedIn post generator.” Enormous existing supply, near-zero differentiation, and pricing gets driven down by sheer competitor volume.
Right-Sized (findable, provable)
“Property listing description prompts for UK estate agents,” “LinkedIn thought-leadership prompts for B2B SaaS founders,” “children's-book-style illustration prompts for self-publishers.” Specific enough that a search for it actually surfaces you, and specific enough that you can demonstrate real expertise rather than generic competence.
A Simple Niche-Selection Checklist
- Can I name the specific person who'd search for this (a job title, a hobby, an industry), not just "everyone"?
- When I search this exact phrase on a marketplace, is the result count reasonable — not empty (no demand), not overwhelming (no differentiation)?
- Can I actually produce a convincing worked example in this niche today, or would I be faking expertise?
- Is there a real recurring task here, or a one-time novelty someone would only buy once ever?