ChatGPT Prompts for Startup Planning: 7 That Actually Work (Copy-Paste)

I use AI for startup planning every single day — it's literally my product. So this isn't going to be one of those "10 mind-blowing prompts" posts written by someone who's never shipped anything. These are seven prompts I've refined through real use, with the reasoning behind each, and an honest section at the end about where chatbot-based planning falls apart.

Copy them, edit the bracketed parts, and use them in ChatGPT, Claude, or any capable model.

Prompt 1: The Honest Validation Scorecard

Most validation prompts fail because chatbots default to encouragement. You have to explicitly demand honesty and structure:

"Act as a skeptical startup analyst who has seen a thousand pitches fail. Evaluate this idea: [YOUR IDEA IN ONE OR TWO SENTENCES]. Score it from 0-100 on each of: competition intensity, market fit evidence, scalability, monetization potential, and risk of being commoditized by general AI tools. For each score, give two sentences of reasoning that reference MY specific idea, not the category. Then tell me the single strongest reason this fails, and what evidence would change your mind. Do not soften the scores to be encouraging — a false 80 costs me six months of my life."

The last sentence is the load-bearing one. Without it, everything scores 75 and up.

Prompt 2: The Competitor Teardown

"List the 5 most direct competitors to this idea: [IDEA]. For each: their core offer, their pricing model if publicly known, their most-praised strength, and their most-complained-about weakness based on the kinds of reviews that appear on G2, Capterra, and Reddit. Then identify the one gap across all five that a new entrant could realistically exploit, and the one gap that LOOKS attractive but is actually a trap (something customers complain about but won't pay to fix)."

That trap question is the part almost nobody asks. Plenty of loud complaints are about things users won't switch tools over.

Prompt 3: The Ideal Customer Sharpener

"My product is [ONE SENTENCE]. Generate three sharply different ideal customer profiles for it. For each: their job title or situation, the trigger moment that makes them search for a solution, the exact phrase they'd type into Google in that moment, what they use today instead, and their realistic willingness to pay per month. Then tell me which ONE profile I should build for first and why the other two would dilute the MVP."

The "exact phrase they'd type into Google" line does double duty — it sharpens positioning AND hands you your first SEO keywords.

Prompt 4: The Ruthless MVP Cut

"Here is every feature I think my product needs: [PASTE YOUR FULL LIST]. Cut it to the smallest version where a customer would still pay. Sort into three buckets: MVP (needed for the first paying customer), Fast-Follow (build in months 2-3), and Distraction (feels important, isn't). For every feature you put in MVP, state which specific user action proves it was needed. Be brutal — every extra MVP feature delays my launch by a week."

Prompt 5: The Pricing Tier Designer

"Design subscription pricing for: [PRODUCT, AUDIENCE, AND THE VALUE IT DELIVERS]. Propose a free tier with limits tight enough that a serious user hits them within two weeks, one paid tier priced against the alternative of [WHAT THEY DO TODAY AND WHAT IT COSTS THEM], and an annual option with the actual computed savings. Justify every price against the customer's alternative, not against what similar tools charge. Then predict the single most likely pricing objection and how the page should pre-empt it."

Prompt 6: The Landing Page Skeleton

"Write the copy skeleton for a landing page selling [PRODUCT] to [AUDIENCE]. Give me: a headline that names the outcome (not the technology), a subheadline that names who it's for, three benefit blocks each tied to a pain (not a feature), the exact microcopy for the primary button, and the three objections that belong in an FAQ. Voice: a competent founder talking to a peer — no hype words like revolutionary, supercharge, or unleash."

Prompt 7: The SEO Topical Map

"Build a topical map for a site selling [PRODUCT] targeting [AUDIENCE]. Give me 3 pillar topics with commercial relevance, 5 supporting article ideas per pillar with realistic search phrasing, and for each article label the intent: informational, comparison, or transactional. Prioritize the 5 articles I should write first based on purchase intent, and tell me which pillar to ignore for the first six months."

If you want to see what a complete version of this output looks like, I've published a full SEO structure example — same shape, generated by our tool.

Where Chatbot Planning Breaks Down

Now the honest part. I built an entire product in this space, so I've thought about this more than is healthy. Chatbot prompts — including mine above — have four structural problems:

1. Completeness depends on your memory

Each prompt covers one slice. Nothing guarantees you'll remember to do the monetization slice, or the risk slice, or the data-model slice. Founders end up with brilliant answers to four questions and blind spots on the six they never asked.

2. Every run has a different shape

Ask about two ideas on two different days and you'll get two differently-structured answers. That makes comparing idea A against idea B nearly impossible — you're comparing an essay against a bullet list.

3. Nothing persists

Your planning lives in scattered chat threads. Three weeks later, finding "that pricing breakdown" means scrolling through a hundred conversations. There's no document, no progress tracking, no single place your plan lives.

4. Prompt skill is the ceiling

The prompts above work because they encode planning expertise — the demand for honesty, the trap question, the intent labels. Most founders don't know what to demand, so they get the encouraging, generic version. Your output quality is capped by your prompting skill on a topic you're, by definition, new to.

These four problems are exactly why I built PromptSeenAI: fixed generators that produce the same complete, comparable structure every time, save everything as documents, and let you compare two validated ideas side by side. I wrote an honest comparison — including when you should just stick with ChatGPT — at that link.

Or skip the reading and run the validator on your idea free, no signup — then run Prompt 1 above in ChatGPT on the same idea and compare the two results yourself. That's the fairest test I can offer.

Frequently Asked Questions

Do these prompts work in Claude and Gemini too?

Yes. They're model-agnostic — the structure and the demands for honesty are what matter, not the model. In my experience Claude is slightly better at sustained honesty when you ask for skepticism.

Should I use one giant mega-prompt instead?

No. Mega-prompts produce shallow coverage of everything. Run focused prompts in sequence, and paste relevant earlier answers into later prompts as context.

How do I know if the AI is making things up?

Assume any specific number (market size, competitor pricing) is unverified until you check it. Use AI output for structure and reasoning; verify facts before they go anywhere near a business decision. This applies to every AI tool, including mine.