Best ChatGPT Prompts for Startups (2026)

A monthly investor update that gets read has four things: the key metric that matters most this month, what is working, what is not, and what you need. Most founders avoid writing it because the blank page is intimidating. Give ChatGPT the raw material and let it structure the draft.

**Example prompt:** > You are an experienced startup operator who writes clear, concise investor updates. Write a monthly investor update email using the data below. Tone: direct, no hype. Format: three short sections — Highlights, Challenges, Asks. Keep it under 300 words. > > Company: [Company name], a [one-sentence description]. > Key metric: [e.g., ARR grew from $48k to $61k MoM, +27%] > What worked: [e.g., outbound sequence to HR teams converting at 4.2%] > What didn't: [e.g., enterprise pilot stalled — champion changed roles] > This month's ask: [e.g., warm intro to anyone who has sold into mid-market HR] **How to adapt:** Replace the bracketed fields with your real numbers. If you track multiple metrics, add them in bullet form under 'Key metric' — the model will pick the most salient one to lead with. Ask for a subject-line suggestion as a follow-up prompt.

The result will rarely be final copy, but it gives you something specific to react to rather than a blank doc. One editing pass on a decent draft is faster than writing from scratch every time.

Slide design is separate from story structure. Most pitch deck critiques from investors are really story critiques — the problem is not clear, the market jump is too aggressive, the 'why us' section feels bolted on. ChatGPT can help you audit the narrative before you touch Figma.

**Example prompt:** > You are a seed-stage investor who has reviewed hundreds of pitch decks. I am building a 10-slide deck. Give me a narrative arc — not slide titles, but the story logic that connects each slide to the next. The investor should finish slide 10 feeling that the outcome is obvious, not that they are being sold. > > Company: [Company name] > Problem: [Two sentences on the specific problem you solve] > Target customer: [Who feels this pain most acutely] > Solution: [What you do] > Early traction: [Any signal — revenue, waitlist, pilots, LOIs] > Team: [One sentence on unfair advantage or relevant background] > Market: [How you size it and why it is big] **How to adapt:** After you get the narrative arc, run a second prompt: 'Now play devil's advocate. What are the three weakest transitions in this story and how would a skeptical Series A investor poke holes in them?' That second pass surfaces objections before a real investor does.

For SaaS-specific deck structures, the Best ChatGPT Prompts for SaaS Founders 2026 guide has a dedicated section on ARR-driven pitch narratives.

The most common failure mode in AI-generated cold email is that it sounds like AI-generated cold email. The fix is specificity: give the model enough context about the prospect that the output cannot be a mass-send template.

**Example prompt:** > You are a B2B sales expert who specializes in concise, relevant cold email. Write a cold outreach email from me ([Your name], founder of [Company]) to [Prospect name], [Prospect title] at [Prospect company]. The email should open with something specific about their company, connect it to the problem we solve, and end with a single low-friction ask (15-minute call or a yes/no question). No more than 100 words in the body. Do not use the phrase 'I hope this email finds you well.' > > What I know about them: [e.g., They just raised a Series B, they are hiring 3 SDRs, their current CRM is Salesforce] > Problem we solve: [One sentence] > Best result a similar customer got: [Concrete outcome if you have one] > > Output: Email body + three subject-line variants. **How to adapt:** The 'what I know about them' field is the most important variable. Run a quick LinkedIn or company news search before generating each email and paste what you find. Shallow research produces shallow personalization.

For cold outbound at volume without losing quality, the AI Stack for Agencies 2026 guide covers tooling that combines prompt templates with prospect data enrichment.

After five to ten customer interviews, founders typically have pages of notes and no clear signal. ChatGPT is good at pattern recognition across unstructured qualitative data — as long as you feed it real quotes, not summaries.

**Example prompt:** > You are a product researcher helping a startup synthesize customer discovery interviews. I will paste notes from [N] interviews. For each interview, I have the respondent's role and the key quotes. Your job is: > 1. Identify the top 3 recurring pain themes, with representative quotes for each. > 2. Note any contradictions between respondents — where one person's priority conflicts with another's. > 3. Identify any jobs-to-be-done that appeared in multiple interviews but that we are not yet solving. > 4. Flag any assumptions in our current product thinking that these interviews contradict. > > Interview notes: > [Paste your notes here — raw is fine] **How to adapt:** You do not need polished interview notes — raw bullet points work. The model will impose structure. If you have a hypothesis you want tested against the notes, add it explicitly: 'We believe [X]. Does the interview data support or contradict this?' That targeted framing produces sharper analysis.

Pairing this with a structured follow-up survey and a segment of your most active users gives you a cleaner signal. The Prompt Engineering for Startups guide covers how to chain multiple prompts together for deeper research synthesis.

Positioning is not a tagline — it is a claim about who you are for, what you do, and why that matters relative to alternatives. A useful first draft forces you to be specific about your competitive context, which most founders avoid until a prospect asks why they should not just use a competitor.

**Example prompt:** > You are a B2B positioning strategist. Help me write a positioning statement using April Dunford's Obviously Awesome framework. My answers to the framework: > > Best customers (who values this most): [e.g., Ops leads at 20-200 person SaaS companies who run weekly all-hands] > Market category: [e.g., async meeting summaries, but that is broad — what subcategory should I own?] > Unique attributes: [e.g., we are the only tool that generates actionable follow-up tickets from meeting recordings, not just summaries] > Value for customers: [e.g., reduces meeting overhead by roughly 40% per their self-reported time tracking] > Alternatives: [e.g., Notion AI, Otter.ai, manual notes, nothing] > > Output: A one-paragraph positioning statement, a tagline (under 10 words), and three homepage headline variants. **How to adapt:** If you have not done a formal positioning exercise before, substitute the framework section with three questions: 'Who is this for? What do they do with it? Why not use [top competitor] instead?' The model will still generate usable output and often identifies positioning gaps you had not articulated.

The companion guide Role Prompts for Founders has a section on using strategic advisor roles in ChatGPT to pressure-test positioning claims before you commit them to your website.

A job description written around job titles rather than outcomes attracts candidates who match the title, not the work. The hiring rubric — the scorecard you use to evaluate candidates consistently — should be written at the same time as the JD, not improvised in the debrief call.

**Example prompt:** > You are an experienced startup recruiter and operator. Write a job description and a hiring scorecard for the following role. > > Role: [e.g., First sales hire, Account Executive] > Stage: [e.g., Seed-stage, 8-person team, no existing sales process] > Outcomes expected in first 90 days: [e.g., Run 30 discovery calls, close first 3 paid pilots, document a repeatable outbound sequence] > Outcomes expected in 12 months: [e.g., Own $400k ARR, build the playbook, hire one SDR under them] > Non-negotiables: [e.g., Has sold to SMB, comfortable with ambiguity, can write their own outreach] > Red flags from past hires: [e.g., Needs a full SDR team before they close anything] > > Output: > 1. Job description (400 words max) > 2. Hiring scorecard with 5-7 criteria, each rated 1-3, with behavioral anchors **How to adapt:** The 'outcomes' fields are the most important part. If you cannot articulate what this person will accomplish in 90 days and 12 months, sharpen that before running the prompt — vague outcomes produce vague criteria.

For solo founders hiring their first two or three people, the Best ChatGPT Prompts for Solopreneurs 2026 guide has a condensed version of this workflow tuned for speed over comprehensiveness.

A competitor teardown is only useful if it produces actionable conclusions — not a feature comparison table you already knew. The most valuable analysis identifies positioning gaps: what the competitor is implicitly not claiming, which segments they are underserving, and where their messaging creates an opening for you.

**Example prompt:** > You are a competitive analyst helping a startup understand its market position. Analyze [Competitor name] using the following structure: > > 1. What positioning claim are they making? (Quote their homepage headline and tagline) > 2. Who is their apparent ICP based on their messaging, case studies, and pricing tier? > 3. What are their three strongest selling points based on public reviews (G2, Capterra, etc.)? > 4. What recurring complaints appear in negative reviews? > 5. What customer segment or use case does their positioning NOT address? > 6. If I am building [Your product description], what is the sharpest positioning contrast I can draw against them? > > Context about my product: [Two sentences] > Competitor's homepage copy (paste it here): [Paste] **How to adapt:** Pasting the competitor's actual homepage copy is the key variable — it stops the model from hallucinating their positioning based on general training data. Pull a fresh snapshot before running this prompt.

Running this analysis for your top two or three competitors, then overlapping the results, often reveals a positioning gap that none of them are claiming. That gap is where your messaging should live. See Prompt Engineering for Startups for how to chain this into a full market map exercise.

Product requirement documents are most useful when they surface what you have not yet decided — the edge cases, the constraint tradeoffs, the assumptions that will blow up in sprint three. A good PRD draft from ChatGPT forces those conversations early rather than in a stand-up when the feature is half-built.

**Example prompt:** > You are an experienced product manager at a B2B SaaS startup. Write a one-page PRD for the following feature using this structure: Problem Statement, Success Metrics, User Stories (3-5), Scope (in/out), Open Questions, Edge Cases, Constraints. > > Feature: [Name and one-sentence description] > Job-to-be-done: [What the user is trying to accomplish — not what the feature does, but what they hire it for] > Target user: [Role, context, and what they are doing right before they need this feature] > Constraints: [e.g., Must work in the existing dashboard without a new page, no new backend endpoints, ships in 2 weeks] > What we are explicitly NOT building: [e.g., We are not building a native mobile version, we are not supporting bulk operations] **How to adapt:** The 'Open Questions' and 'Edge Cases' sections are where PRD drafts pay the biggest dividend. If the model's open questions feel obvious, add 'Be adversarial — what would a senior engineer object to in sprint planning?' as a follow-up.

For teams that have non-technical co-founders writing specs, Prompt Engineering for Non-Technical Founders covers how to write prompts that translate business intent into developer-ready language without requiring technical background.

Landing page copy fails when it describes the product instead of the outcome the customer cares about. ChatGPT can generate multiple framings of the same product quickly, which makes it easier to identify which angle your ICP will respond to before you run any paid traffic.

**Example prompt:** > You are a direct-response copywriter who specializes in SaaS landing pages. Write the above-the-fold copy for my landing page. Output: > - Two hero headline variants (under 12 words each) > - One subheadline for each variant (20-30 words, outcome-focused) > - Three bullet-point proof lines (what users achieve, not what features exist) > - One CTA button label for each variant > > Product: [Name and one-sentence description] > ICP: [Role + company type + the specific pain they have] > Core outcome: [What the customer achieves after using it — be specific] > Social proof signal: [e.g., 'Used by 200 teams' or 'Rated 4.8 on G2' — only include if real] > Tone: [e.g., Direct and confident, not startup-hype] **How to adapt:** The 'core outcome' field determines headline quality more than any other variable. 'Faster workflows' is weak. 'Cut your weekly reporting time from 4 hours to 20 minutes' is strong. The more concrete your outcome, the more concrete the headline.

After generating two or three headline variants, run a follow-up prompt: 'Which of these headlines would resonate most with a [specific ICP role] who is skeptical of new tools? What objection would they have to each?' That objection-mapping prompt often reveals a stronger angle than the original drafts.

ChatGPT is not a financial model — it cannot verify your spreadsheet math. What it can do is act as a devil's advocate on the assumptions driving your model: growth rates, churn, conversion rates, hiring timelines. Most financial model problems are assumption problems, not math problems.

**Example prompt:** > You are an experienced SaaS CFO reviewing the growth assumptions of an early-stage startup. I will list the key assumptions in our 18-month financial model. For each one: > 1. Tell me if it is aggressive, conservative, or roughly market-rate based on SaaS benchmarks. > 2. Identify the single biggest risk if that assumption is wrong. > 3. Suggest a sensitivity test — what would the model look like if this assumption is 50% worse? > > Assumptions: > - Monthly new customer growth: [e.g., 15% MoM] > - Gross churn: [e.g., 2% monthly] > - Avg contract value: [e.g., $6,000 ACV] > - Sales cycle: [e.g., 30 days] > - Gross margin: [e.g., 72%] > - Time to first revenue from new hire: [e.g., 60 days] **How to adapt:** Do not paste your full spreadsheet — paste the assumptions in plain text. The model engages with assumptions more sharply when they are isolated. After the sanity check, ask: 'Which two assumptions, if both wrong simultaneously, would be most damaging to runway?' That stress-test framing catches correlated risks that individual assumption checks miss.

If you want to understand what your AI tooling costs are contributing to your burn rate, the AI Prompt Cost Calculator lets you model token costs per workflow so you can include AI spend in your financial assumptions accurately.

These prompts share a common structure: a clear role for the model, a specific output format, and real context from your company. The role framing matters more than most founders expect — 'You are a seed-stage investor' produces a different response than 'You are a CFO' on the same input, because each role has different priors about what matters.

Two habits that improve outputs consistently: First, give the model permission to ask clarifying questions before it drafts. Add 'If you need more context to produce a high-quality output, ask me up to three questions first' to any prompt. This catches vague inputs before they produce vague outputs. Second, run a 'critic pass' after any first draft — paste the output back with 'Now critique this as if you were the most skeptical [investor / customer / engineer] this would reach. What are the three weakest points?'

For deeper coverage of structuring multi-step prompt workflows, the Prompt Engineering for Startups guide covers chain-of-thought techniques, few-shot examples, and how to build reusable prompt templates for your team. If you are running a content or client-services operation alongside your core product, AI Stack for Agencies 2026 covers how to operationalize these workflows at scale.