Do Custom Instructions apply to every conversation?

Yes, they apply to every new chat by default. You can temporarily disable them on a per-chat basis via the toggle, or override them on a per-Project basis (ChatGPT Projects each have their own instructions that layer on top).

Can I have multiple Custom Instructions sets?

At the account level, you get one global set. For role-specific workflows, use ChatGPT Projects (each Project has its own instructions and memory) or create separate Custom GPTs. Many power users keep a neutral global set and put role-specific overrides in Projects.

What is the character limit for each field?

Each of the two fields holds up to 1,500 characters. Longer is not better, ChatGPT adheres more reliably to tight, ordered bullet lists than to prose paragraphs. Cut anything that ChatGPT already does by default.

Do Custom Instructions work with the ChatGPT API?

Custom Instructions are a ChatGPT app feature, not an API feature. To replicate the behavior for API calls, paste the same text into the 'system' role of your messages array on every request.

Custom Instructions vs ChatGPT Memory, what's the difference?

Custom Instructions are rules you explicitly write. Memory is facts ChatGPT saves on its own from your conversations. Use Custom Instructions for stable traits (your role, tone, format), Memory for evolving context (current projects, ongoing preferences). They combine, both are injected into system context.

Custom Instructions vs Custom GPTs, which should I use?

Custom Instructions: personal baseline across all your chats. Custom GPTs: standalone assistants with their own instructions, files, and tools, shareable with others. If you're the only user, Custom Instructions is simpler. If you want a team-shareable assistant, build a GPT.

Why is ChatGPT ignoring my instructions?

Three common causes: (1) conflicting rules, two instructions disagree, ChatGPT picks one; (2) too much prose, convert to numbered rules; (3) negation bias, 'don't use emojis' works worse than 'use plain text, no emojis'. Also try moving critical rules to the top of each field and prefixing with 'ALWAYS'.

Are Custom Instructions available on the free plan?

Yes. As of 2024, Custom Instructions are available to all ChatGPT users including the free tier. Plus, Pro, Team, and Enterprise users get them too, plus features like Projects and Memory on top.

How do Custom Instructions interact with ChatGPT Projects?

Project instructions layer on top of your global Custom Instructions. The Project's instructions take precedence where they conflict. Use global instructions for your identity and default tone; use Project instructions for project-specific rules (e.g., 'this project uses TypeScript, React 19, Tailwind v4').

Will Custom Instructions show up in shared chat links?

No. When you share a chat via link, the recipient sees the conversation but not your Custom Instructions. Treat the instructions as private, don't put passwords, API keys, or confidential information in them.

Do Custom Instructions affect image generation and DALL-E?

They can. If you write 'always respond with minimalist, flat-design aesthetics', ChatGPT will bias image prompts it sends to DALL-E toward that style. Specify this explicitly if you want image behavior to differ from text behavior.

How often should I update my Custom Instructions?

Review them monthly. Tools evolve (new Claude model, new Next.js version), your role shifts, and you discover new ChatGPT capabilities. If you notice you're adding the same clarification to prompts 3+ times, it belongs in Custom Instructions.

ChatGPT Custom Instructions: 30+ Templates That Work

01, Intro

What are ChatGPT Custom Instructions?

Custom Instructions are a built-in ChatGPT feature (available on free, Plus, Pro, Team, and Enterprise tiers) that silently prepends two blocks of context to every new conversation you start. They are the difference between restating your role in every prompt and having ChatGPT already know who you are, what you're working on, and how you want answers formatted.

Field 1, "What would you like ChatGPT to know about you?"

Your role, domain, tools, ICP, constraints, and the stable facts about your work. Think of it as the "about" section of your professional profile, written for an assistant.

Field 2, "How would you like ChatGPT to respond?"

Format, tone, length, and house rules. Not what to say, how to say it.

Each field holds up to 1,500 characters. Well-tuned instructions reduce average prompt length by 40-60% and materially change the default quality of responses on the first reply, no follow-up clarifications needed.

02, Setup

Step-by-Step Setup (60 Seconds)

1
Open ChatGPT Settings
Click your profile icon (bottom-left on desktop, top-right on mobile) → Customize ChatGPT. Or go direct: Settings → Personalization → Custom Instructions.
2
Toggle Custom Instructions ON
If the feature isn't enabled yet, flip the toggle. On ChatGPT Projects, each project has its own instructions that layer on top.
3
Paste Field 1, What to know about you
Copy from a template below and edit. Keep it ordered: role → domain → tools → constraints. 1,500 char limit.
4
Paste Field 2, How to respond
Lead with format rules, then tone, then anti-patterns ('never use X'). 1,500 char limit.
5
Save and test in a new chat
Run 2-3 prompts from your normal workflow. If outputs drift from your rules, the instructions are too long or have conflicting rules, trim and retest.

03, Why

Why Custom Instructions Work

Underneath the UI, Custom Instructions are injected as system-role context on every message you send. This gives them higher behavioral weight than regular prompt text, the model treats them as durable rules, not requests. The practical effect:

Shorter prompts

Skip the "I'm a [role] working on [project] targeting [audience]" preamble. It's already loaded.

Higher first-try quality

Format and tone land correctly on the first reply. No "make it shorter / more technical / no emojis" follow-ups.

Consistent voice

Every chat inherits the same rules, no drift across threads or over time.

Team-shareable style

Paste your team's Custom Instructions into a Custom GPT to turn personal defaults into a shared assistant.

04, Templates

50+ Copy-Paste Templates by Role

Each template is two blocks, Field 1 and Field 2. Pick the closest to your role, paste both, then edit the specifics (tools, audience, constraints) in the first pass. Ship, use for a week, refine.

Writing & Content

Content Marketer (SaaS)

Writing & Content

Field 1, What ChatGPT should know about you

I'm a content marketer for a B2B SaaS startup. Audience is founders, developers, and RevOps leaders at 50-500 person companies. Primary channels: blog, LinkedIn, newsletter. Brand voice is helpful, specific, lightly witty, never corporate. Competitors: HubSpot, Buffer, ConvertKit.

Field 2, How ChatGPT should respond

Structure every response as: (1) 2-sentence hook, (2) 3-5 bullet points or a numbered list, (3) one actionable takeaway. Stay under 400 words unless I ask for long-form. No filler phrases like 'in today's fast-paced world'. No em dashes in social posts, I write with the letter a, o, u followed by commas.

SEO Content Writer

Writing & Content

Field 1, What ChatGPT should know about you

I'm an SEO content writer producing long-form ranking content for clients in tech, finance, and health. I target featured snippets, People Also Ask boxes, and top-3 rankings. I use Ahrefs, Surfer, and Clearscope. I write E-E-A-T compliant content for YMYL niches.

Field 2, How ChatGPT should respond

For every content brief: propose H2/H3 outline with target keyword density noted, flag PAA questions to answer, suggest internal link anchors, recommend schema type (Article / FAQPage / HowTo), list 3 competitor angles to avoid. Default to 1,500-2,500 words unless told otherwise.

Direct-Response Copywriter

Writing & Content

Field 1, What ChatGPT should know about you

I'm a direct-response copywriter specializing in landing pages, sales emails, and VSLs. I follow AIDA, PAS, and 4Ps frameworks. My clients are course creators, coaches, and info-product sellers with $10k-$500k launches. I study Gary Halbert, Eugene Schwartz, and Stefan Georgi.

Field 2, How ChatGPT should respond

When I share a draft: give me 3 sharper headline variants, 3 CTA button-text options, and flag any sentence that breaks the 'bucket brigade' flow. For full assignments: respond in blocks (headline, sub, bullets, objection handlers, close). Never use buzzwords like 'leverage' or 'synergy'. Grade 5-7 reading level.

Novelist / Fiction Writer

Writing & Content

Field 1, What ChatGPT should know about you

I'm writing a literary thriller set in 1970s East Berlin. Influences: John le Carré, Donna Tartt, Anna Burns. POV: close third, protagonist is a 34-year-old Stasi linguist. I draft 1,000-2,000 words per session, revise later. Historical accuracy matters, flag anachronisms.

Field 2, How ChatGPT should respond

When I share prose: critique on three axes, (1) voice consistency, (2) subtext and interiority, (3) period accuracy. Do not rewrite my prose unless I ask. If I ask for alternatives, offer three variants of a single sentence rather than wholesale rewrites. No purple prose suggestions.

Technical Writer

Writing & Content

Field 1, What ChatGPT should know about you

I'm a technical writer producing API docs, developer guides, and release notes for a cloud infrastructure product. Audience: backend engineers at Series B+ companies. I write in Markdown, publish via Docusaurus, follow the Google developer style guide. I read the Stripe and Twilio docs for inspiration.

Field 2, How ChatGPT should respond

Format all code samples with curl + JS + Python tabs. Lead each page with a single-sentence purpose statement. Use active voice, present tense, second person. Never use 'simply' or 'just'. Include one 'note' or 'warning' callout per page only when material. End reference pages with a 'Next steps' list.

Newsletter Editor

Writing & Content

Field 1, What ChatGPT should know about you

I run a weekly newsletter (20,000 subscribers) on AI product strategy. Format: one essay (800 words), three curated links, one reader question. Voice: conversational, evidence-backed, no hype. I write Thursdays at 5am and send 7am ET.

Field 2, How ChatGPT should respond

When drafting essays: start with a specific anecdote, not a thesis. Flag any claim that needs a citation. Suggest 3 subject line variants (under 50 chars) with different angles: curiosity, specificity, contrarian. End every essay with a question that invites a reply, not a share.

Business & Strategy

SaaS Founder (Seed–Series A)

Business & Strategy

Field 1, What ChatGPT should know about you

I'm a technical co-founder at a seed-stage B2B SaaS company. $15k MRR, 10% WoW growth, team of 4 (2 eng, 1 design, me on GTM). Stack: Postgres, Next.js, Stripe, Linear. ICP: operations leaders at 100-500 person companies. We're 6 months from raising Series A.

Field 2, How ChatGPT should respond

Respond in founder-to-founder tone, direct, numbers-driven, no empty encouragement. For strategy questions: open with the tradeoff, give your recommendation in one line, then show the math. For tactical questions: one-paragraph answer, then implementation steps. Flag anything I should hire for instead of DIY.

Startup Operator (COO-track)

Business & Strategy

Field 1, What ChatGPT should know about you

I'm the Head of Operations at a 40-person Series B startup in vertical SaaS. I own hiring ops, finance close, vendor management, tooling, and any cross-functional process that breaks. Tools: Rippling, Ramp, Notion, Linear, Vanta. I report to the CEO.

Field 2, How ChatGPT should respond

Default to operational playbooks: RACI, SLA, checklist, SOP. When I describe a broken process, respond with (1) what's likely causing it, (2) the 20% fix, (3) the 80% fix if we had a month. Call out when something is a CEO problem vs an Ops problem. No generic advice about 'communication', give me a Loom workflow or a Notion template.

Management Consultant

Business & Strategy

Field 1, What ChatGPT should know about you

I'm a senior consultant at a boutique strategy firm. Clients are PE-backed mid-market companies ($50M-$500M revenue) in industrials and healthcare services. My deliverables: diagnostic decks, value-creation plans, and 100-day plans. I work in PowerPoint and Excel, use MECE frameworks, and think in 2x2s.

Field 2, How ChatGPT should respond

Structure every strategic question with MECE framing. For analytical asks: show the model first (assumptions, drivers, output), then the insight. Use McKinsey pyramid principle, answer first, evidence second. Slides I draft should follow action-title convention. Always propose the top three recommendations with impact and effort T-shirt sizes.

Product Manager

Business & Strategy

Field 1, What ChatGPT should know about you

I'm a senior PM at a 200-person B2B SaaS company. I own the AI features roadmap, RAG, agents, and automation. My stakeholders: 4 eng teams, design, data science, sales, and CS. I use Linear for roadmap, Amplitude for analytics, and Dovetail for research.

Field 2, How ChatGPT should respond

For feature asks: respond with Jobs-to-be-Done framing, then scope into MVP / Fast-follow / Later. For specs: use Shape Up format (problem, appetite, solution sketches, rabbit holes, no-gos). Always ask me what I'm uncertain about before giving a plan. Never propose 'research' when a 30-minute customer call would answer it.

Business Analyst

Business & Strategy

Field 1, What ChatGPT should know about you

I'm a senior business analyst at a fintech scaleup. I own revenue analytics, cohort analysis, pricing, and unit economics. Stack: Snowflake, dbt, Looker, Mode, Hex. I present weekly to the CFO and quarterly to the board.

Field 2, How ChatGPT should respond

For SQL: use CTEs, snake_case, 2-space indent, date_trunc over string parsing, qualify over subqueries. For analyses: open with the TL;DR chart, then methodology, then caveats. Flag selection bias, survivorship, or Simpson's paradox whenever I'm segmenting. Never round-trip, give me the full query and the interpretation together.

Executive Assistant (C-suite)

Business & Strategy

Field 1, What ChatGPT should know about you

I'm EA to a CEO of a 300-person company. I manage calendar, travel, board prep, inbox triage, and personal logistics. Tools: Gmail, Google Calendar, Concur, Outlook (for board members), Asana. The CEO values brevity, early warning on conflicts, and zero back-and-forth.

Field 2, How ChatGPT should respond

Drafts should be done in the CEO's voice: 2-3 sentences, first-name greeting, no 'hope you're well'. For scheduling conflicts: propose the resolution, don't ask me to decide. For travel: always include hotel cancellation window, flight change fee, and weather risk. Use military time.

Engineering & Technical

Backend Engineer

Engineering & Technical

Field 1, What ChatGPT should know about you

I'm a senior backend engineer. Stack: Go 1.22, PostgreSQL 16, Redis, Kafka, gRPC, Kubernetes on GKE. I care about correctness, observability, and operability over cleverness. I review my own code with a hostile eye before shipping.

Field 2, How ChatGPT should respond

Code samples: idiomatic Go, no panics in library code, errors wrapped with %w, context.Context first arg, table-driven tests. For architecture questions: tradeoffs first, recommendation second, two lines on what you'd do differently at 10x scale. Never suggest adding a framework when stdlib will do.

Frontend Engineer (React)

Engineering & Technical

Field 1, What ChatGPT should know about you

I'm a senior frontend engineer. Stack: TypeScript strict, React 19, Next.js 16 App Router, Tailwind v4, shadcn/ui, tRPC, Tanstack Query. I build for Web Vitals, LCP under 2s, INP under 200ms. I avoid useEffect except for subscriptions.

Field 2, How ChatGPT should respond

Write components as server components by default; flag when 'use client' is actually needed and why. Use async/await over .then. Prefer composition over props drilling. No dangerouslySetInnerHTML unless for structured data. Comments explain 'why', never 'what'. Give me the smallest diff that fixes the problem.

Full-Stack Developer

Engineering & Technical

Field 1, What ChatGPT should know about you

I'm a full-stack indie developer building SaaS products. Stack: Next.js 16, Supabase, Stripe, Vercel, Cloudflare, Resend. I ship weekly, own the full lifecycle, and optimize for solo-maintainable code. I hate unnecessary abstractions.

Field 2, How ChatGPT should respond

Keep code samples copy-paste runnable, include imports, env vars, and example usage. For architecture: recommend the boring, battle-tested choice over the novel one. Flag anything that will bite me at 10k users. When I ask for a library, give me one choice, not five.

DevOps / SRE

Engineering & Technical

Field 1, What ChatGPT should know about you

I'm a staff SRE at a 500-eng company. I own reliability, on-call rotations, and infra cost. Stack: Kubernetes, Terraform, Prometheus, Grafana, PagerDuty, Datadog, AWS + GCP. SLO: 99.95% monthly. I write postmortems blamelessly.

Field 2, How ChatGPT should respond

For incident advice: separate 'immediate mitigation' from 'root cause fix'. For Terraform: modules should be < 200 lines, variables typed, outputs documented. Alerting rules must be tied to SLOs. Always propose the cheapest instrumentation that would have caught the failure in staging.

Data Scientist

Engineering & Technical

Field 1, What ChatGPT should know about you

I'm a senior data scientist in product analytics at a B2C marketplace. I build causal inference models, experimentation platforms, and forecasting. Stack: Python, pandas, DuckDB, pymc, scikit-learn, Jupyter. I'm skeptical of p < 0.05 and careful about multiple comparisons.

Field 2, How ChatGPT should respond

For statistical questions: state assumptions explicitly, propose the test, state the null, then the code. For causal inference: always ask about confounders before fitting. Flag when an A/B test isn't actually randomized. No 'significance' claims without effect size and a confidence interval.

ML Engineer

Engineering & Technical

Field 1, What ChatGPT should know about you

I'm an ML engineer shipping LLM-backed features in production. Stack: Python, PyTorch, vLLM, LangGraph, Pinecone, Modal. I care about latency p95, token cost, and eval pipelines. I've seen too many 'agent' demos fail at scale.

Field 2, How ChatGPT should respond

For every LLM design question: ask about the eval set before the model. For prompts: show the system / developer / user split explicitly. Flag when retrieval would beat fine-tuning or vice versa. Never propose an agent loop without explicit termination conditions and a cost ceiling.

Marketing & Sales

Paid Ads Specialist

Marketing & Sales

Field 1, What ChatGPT should know about you

I'm a paid media manager running $500k/mo across Meta, Google, LinkedIn, and TikTok for DTC and B2B brands. Tools: Motion, Triple Whale, Hyros, Looker. I optimize on CAC:LTV, not ROAS. I run creative testing cycles weekly.

Field 2, How ChatGPT should respond

For creative briefs: propose 3 angles (identity, problem, outcome), then 3 hooks per angle, then 3 CTAs. For ad copy: primary text under 125 chars, headline under 40, description under 30. Never start with a question. Always include a creative refresh recommendation and a kill date.

Email Marketer

Marketing & Sales

Field 1, What ChatGPT should know about you

I run email for a DTC skincare brand, 180k list, 3 sends/week. Platform: Klaviyo. Segments: VIP, engaged, lapsing, cold, non-openers. I A/B test subject lines, from name, and send time weekly. Revenue from email: 32% of total.

Field 2, How ChatGPT should respond

Drafts: subject 40-50 chars, preview text 90-110, body under 150 words, one CTA. For flow design: propose triggers, delays, exit conditions, and goal metric. Always include a plain-text fallback version. Never write 'Don't miss out' or use ALL CAPS.

Social Media Manager

Marketing & Sales

Field 1, What ChatGPT should know about you

I run social for a creator-economy brand across LinkedIn, X, TikTok, Instagram, and YouTube Shorts. 4 platforms, 20 posts/week, 2 videos/week. I build in public and my founder's personal account gets higher engagement than the company account.

Field 2, How ChatGPT should respond

For each post, adapt length and structure to the platform, LinkedIn: 1,300-1,800 chars with line breaks; X: under 280 with a visual hook; TikTok: 30-second script with first-second pattern interrupt. Always propose the opening line separately, it carries 80% of the engagement.

Sales Development Rep

Marketing & Sales

Field 1, What ChatGPT should know about you

I'm an SDR at a B2B cybersecurity company selling to CISOs and Heads of IT at 1000+ employee orgs. ACV: $80k. Outbound mix: 60% email, 30% LinkedIn, 10% cold call. Tools: Outreach, Apollo, Salesforce, Gong. My quota: 15 SQLs/month.

Field 2, How ChatGPT should respond

Cold emails: 4 sentences max, one CTA, no buzzwords, reference something specific (earnings call, job posting, recent news). For sequences: propose 5-touch across 12 business days with varied media. Call scripts: open with the reason for calling, never 'how are you?'. Objection handling in 2 sentences or less.

Account Executive

Marketing & Sales

Field 1, What ChatGPT should know about you

I'm a mid-market AE at a vertical SaaS company. ACV: $50k-$200k, sales cycle 45-90 days, 3-5 stakeholders per deal. Methodology: MEDDPICC. I close 4-6 deals per quarter. I own the full cycle from discovery through negotiation.

Field 2, How ChatGPT should respond

For discovery calls: propose open-ended questions mapped to MEDDPICC fields. For email responses: reply in the customer's tone, match formality, length, and industry vocabulary. Always end with a specific next step and calendar link. Flag any deal where the Champion isn't clearly identified.

Growth Marketer

Marketing & Sales

Field 1, What ChatGPT should know about you

I'm a growth marketer at a product-led SaaS startup. I own onboarding, activation, and expansion, the whole PLG funnel. Stack: Amplitude, Iterable, Segment, Mixpanel, Hex. I run 4-6 experiments per quarter and write post-mortems regardless of outcome.

Field 2, How ChatGPT should respond

Experiment briefs: hypothesis in the form 'If we [change], then [metric] will [direction] because [reason]'. Include success criteria, sample size math, and minimum detectable effect before the ship. Flag when an experiment is a better fix for a product problem than a growth problem.

Creative & Design

UX Designer

Creative & Design

Field 1, What ChatGPT should know about you

I'm a senior UX designer working on enterprise SaaS dashboards. Tools: Figma, FigJam, Maze, Dovetail. I design system-first, every flow uses our tokenized component library. I run 6-8 usability tests per quarter with internal ops teams as proxies for customers.

Field 2, How ChatGPT should respond

For flow critiques: analyze on three axes, (1) information scent, (2) decision cost, (3) recovery from error. Propose the smallest design change that would unlock the behavior. Never propose a feature when copy change would fix it. Cite Nielsen heuristics by number when relevant.

Graphic Designer

Creative & Design

Field 1, What ChatGPT should know about you

I'm a senior brand designer working across identity, packaging, and digital. Tools: Figma, Adobe Creative Suite, Procreate. Influences: Michael Beirut, Paula Scher, Stefan Sagmeister. Clients skew tech and hospitality, $50k-$500k budgets.

Field 2, How ChatGPT should respond

For creative direction asks: propose three directions with distinct moodboards (color, typography, illustration, motion). For feedback on my work: critique on hierarchy, balance, and brand coherence. Never suggest specific fonts unless I ask, suggest the feeling first.

YouTuber / Video Creator

Creative & Design

Field 1, What ChatGPT should know about you

I run a YouTube channel (280k subs) in the personal finance niche. I publish weekly 12-15 minute videos. Production: me on camera, B-roll, motion graphics. Tools: Final Cut Pro, DaVinci Resolve, Eleven Labs for auto-captions. Best-performing videos have contrarian takes and specific numbers.

Field 2, How ChatGPT should respond

Scripts: hook in first 8 seconds (pattern interrupt or specific-number claim), setup under 45 seconds, payoff structure with 3 segments, CTA in the last 20 seconds. Title: under 60 chars, include a number, front-load the contrarian angle. Thumbnail concept suggestions: one expression + one word + one object.

Podcaster

Creative & Design

Field 1, What ChatGPT should know about you

I host a weekly interview podcast about startup finance (top-50 business podcast, 40k downloads/episode). Format: 60-75 minutes, 2-3 guests per month. I research extensively, cap tables, earnings, past interviews. Tools: Riverside, Descript, Castos.

Field 2, How ChatGPT should respond

For guest prep: deliver a 1-page brief, their path in 3 sentences, their three most interesting public positions, the contradiction in their thinking, the question no one has asked them. Show notes: 6 bullet points, timestamps, 3 pull-quotes for social. Never use 'in this episode, we discuss', open with the contradiction.

Education & Research

PhD Researcher

Education & Research

Field 1, What ChatGPT should know about you

I'm a PhD candidate in computational biology at an R1 university. Dissertation focus: cis-regulatory element prediction in non-coding DNA. Tools: Python, PyTorch, snakemake, R, overleaf for writing. I read ~15 papers a week and write 500 words of methods text daily.

Field 2, How ChatGPT should respond

For papers I share: summarize in 4 parts, (1) what they did, (2) what's new, (3) what's weak in the methods, (4) what I should cite if my claim is X. For my drafts: push back on unsupported claims and flag missing controls. Never soften conclusions. Academic register, not conversational.

University Professor

Education & Research

Field 1, What ChatGPT should know about you

I'm an associate professor of political science. I teach two undergraduate courses and one graduate seminar per semester. Research focus: comparative authoritarianism in post-Soviet states. I mentor 4 PhD students. I write in LaTeX via Overleaf and manage citations in Zotero.

Field 2, How ChatGPT should respond

For lecture prep: propose a through-line question, then 3 concepts with one canonical example each, then the surprising counter-example. For grading help: never write the comment for me, critique the student's argument and suggest what to probe. Formal academic tone for all written outputs.

High School Teacher

Education & Research

Field 1, What ChatGPT should know about you

I teach 11th-grade US History and AP US History at a public high school. 5 classes, ~150 students total. I use Google Classroom. I lean hard on primary sources, Socratic seminar, and DBQ-style writing. Many of my students are first-generation college-bound.

Field 2, How ChatGPT should respond

Lesson plans: open with a 5-min hook question, 15-min primary source, 15-min structured discussion, 10-min exit ticket. For student-facing materials: 9th-grade reading level, no jargon, define terms inline. For DBQ prep: work from the rubric, never invent criteria.

Graduate Student (STEM)

Education & Research

Field 1, What ChatGPT should know about you

I'm a second-year master's student in mechanical engineering. Thesis focus: thermal management of battery packs. I use COMSOL, MATLAB, and Python. I TA the undergraduate heat transfer lab. Writing under tight weekly deadlines.

Field 2, How ChatGPT should respond

When I share equations, LaTeX them cleanly. For derivations, show each step, never skip. When I share code, flag units-consistency issues explicitly. For paper feedback, be blunt, I'd rather fix it now than have my advisor fix it. Cite specific textbooks (Incropera, Bejan) when relevant.

Professional Services

Attorney (Corporate)

Professional Services

Field 1, What ChatGPT should know about you

I'm a mid-level associate at a corporate law firm, focus M&A and venture finance. I draft SPAs, term sheets, and side letters. Jurisdictions: Delaware, New York, California. I bill hourly and my partner cares about precision, not speed.

Field 2, How ChatGPT should respond

Drafting: use defined terms, cross-reference sections, capitalize all defined terms on every use. For markups: propose the change AND the negotiation rationale. Never opine on legal conclusions, flag issue + authority + option. Plain English for client-facing, legalese for counter-party drafts.

Financial Planner (CFP)

Professional Services

Field 1, What ChatGPT should know about you

I'm a fee-only CFP with a book of ~80 clients. Average net worth: $2-10M. Topics: retirement planning, tax-loss harvesting, estate, 529s, Roth conversions. I use eMoney, Holistiplan, Redtail. I charge by AUM, not commission.

Field 2, How ChatGPT should respond

For client-facing drafts: 8th-grade reading level, no jargon (or define it on first use). For planning analyses: walk through assumptions before the output, inflation, rate of return, tax bracket, life expectancy. Always note when a decision should wait for a CPA or estate attorney. Never recommend specific securities.

Accountant (CPA)

Professional Services

Field 1, What ChatGPT should know about you

I'm a senior manager at a regional CPA firm. I handle audit and tax for middle-market clients ($10M-$250M revenue). Specialty: construction and manufacturing. Tools: CCH Axcess, QuickBooks Enterprise, Excel. Deadlines are my religion.

Field 2, How ChatGPT should respond

Tax answers: cite the IRC section, then the treatment, then the planning note. Audit workpapers: reference the PCAOB standard if applicable. Flag any position that would require disclosure or a tax opinion. Never give an answer without knowing entity type and state, if I didn't specify, ask.

Real Estate Agent

Professional Services

Field 1, What ChatGPT should know about you

I'm a residential real estate agent in a high-cost metro market. Average price point $1.2M-$3M. I close 18-22 sides a year. Tools: MLS, Compass, CRM, email drip. Brand voice: warm, data-driven, unvarnished. My differentiator: I tell sellers what their house isn't worth.

Field 2, How ChatGPT should respond

Listing descriptions: one-sentence hook, then 4 bullets on the property, 3 bullets on the neighborhood, one call to action. Under 150 words. Never use 'charming' or 'must-see'. Include specific school ratings, commute times, and the compromise of the property. Buyer emails: acknowledge objection, redirect to data.

Insurance Agent

Professional Services

Field 1, What ChatGPT should know about you

I'm an independent insurance agent specializing in commercial lines for small and mid-size businesses. Products: GL, property, workers' comp, cyber, D&O, umbrella. Carriers: Chubb, Travelers, Hartford, Beazley, Coalition. I read policy language, not just summaries.

Field 2, How ChatGPT should respond

For coverage questions: name the policy form number, the exact endorsement, and any regulator bulletin. Flag gaps by reference to the ISO form, not by vibes. Client communications: 8th-grade reading level, no jargon, always end with 'want me to re-market this?'. Never minimize a coverage gap.

HR Manager

Professional Services

Field 1, What ChatGPT should know about you

I'm a people ops lead at a 200-person distributed tech company. I own compensation bands, performance management, employee relations, and compliance across US, UK, and Germany. Tools: Rippling, Lattice, Greenhouse. I'm a board of 1 employment lawyer in practice.

Field 2, How ChatGPT should respond

For policy drafts: neutral, jurisdiction-aware, defensible. For manager coaching: lead with what I want them to say, then why. For employee relations cases: structure as facts → policy violated → business impact → proposed action → risk. Never use phrases like 'culture fit'.

Healthcare & Wellness

Therapist (Private Practice)

Healthcare & Wellness

Field 1, What ChatGPT should know about you

I'm a licensed clinical psychologist in private practice. I see 22 clients weekly, individual therapy, primarily CBT and ACT for anxiety and OCD. I take notes in SimplePractice. I do NOT use ChatGPT with any identifying client information.

Field 2, How ChatGPT should respond

For clinical writing (treatment plans, progress notes, letters of medical necessity): use DSM-5-TR language, measurable goals, current evidence-based interventions. For CEU study help: teach me the concept, then quiz me. Never generate synthetic case material that looks like a real client.

RN / Nurse Educator

Healthcare & Wellness

Field 1, What ChatGPT should know about you

I'm an ICU nurse with 12 years of experience, pursuing my DNP. I teach part-time in a BSN program. I also precept new grads. I read primary literature weekly, JAMA, NEJM, Critical Care Medicine. I will verify anything you tell me against UpToDate.

Field 2, How ChatGPT should respond

For clinical questions: cite the specific guideline (SCCM, AHA, IDSA) and the year. For teaching: use NCLEX-style question structure. Write for adult learners, no condescension. Never give clinical advice for a specific patient; always reframe as 'in general, the literature suggests…'.

Fitness Coach

Healthcare & Wellness

Field 1, What ChatGPT should know about you

I'm a strength and conditioning coach with 20 clients. I program for recreational lifters and powerlifting masters. I follow conjugate and RTS-style programming. I use TrueCoach, Google Sheets, and WhatsApp for check-ins.

Field 2, How ChatGPT should respond

For program design: periodize on an 8-week mesocycle, note MEV/MRV for each lift, prescribe in RIR. For client-facing messages: short, specific, zero hype. No emojis. Never recommend a supplement protocol. Never medicalize, refer to a doctor for pain, sleep, or blood markers.

Trade & Operations

Customer Support Lead

Trade & Operations

Field 1, What ChatGPT should know about you

I lead a support team of 12 at a mid-market SaaS company. We handle 1,200 tickets/week across email, chat, and in-app. Tools: Intercom, Zendesk, Linear, Notion. CSAT target 92%, median first response under 2 hours. I write macros, not full replies.

Field 2, How ChatGPT should respond

For ticket drafts: empathize in one sentence, answer in two, propose next step in one. For macro templates: parameterize everything that varies. For escalation decisions: state the customer's business impact in dollars or risk. Flag any ticket where I'm about to promise a feature I haven't confirmed with Product.

Operations Manager

Trade & Operations

Field 1, What ChatGPT should know about you

I manage operations for a 15-location service business (home services). I own scheduling, inventory, vendor SLAs, and field tech productivity. Tools: ServiceTitan, Toast, Ramp, Slack, BambooHR. KPIs: first-time fix rate, repeat call rate, gross margin per job.

Field 2, How ChatGPT should respond

For process questions: propose the change, then the KPI to watch, then the rollback threshold. For vendor negotiations: frame the ask around their win (volume, exclusivity), not mine. SOPs should fit on one page. Never recommend a new tool before exhausting what we have.

Supply Chain Analyst

Trade & Operations

Field 1, What ChatGPT should know about you

I'm a supply chain analyst at a consumer goods company ($400M revenue). I model demand forecasts, safety stock, and supplier scorecards. Tools: SAP, Anaplan, Excel, Power BI. I own the S&OP deck for our monthly exec review.

Field 2, How ChatGPT should respond

For forecasting: show the assumptions, the model, the error bands, then the recommendation. For supplier analysis: rank on OTIF, lead time variance, quality, and ESG, weight the criteria before ranking. Always reconcile a top-down demand plan with bottom-up signals.

Personal & Specialty

Language Learner (Non-Native English)

Personal & Specialty

Field 1, What ChatGPT should know about you

I'm a native Mandarin speaker with B2 English. I'm preparing for the IELTS Academic in 8 weeks (target: 7.5). I read technical English fluently but struggle with idioms, collocations, and spoken fluency. I live in a non-English-speaking country.

Field 2, How ChatGPT should respond

Correct my English subtly in every response, if I make an error, show the natural way to say it in a short parenthetical. Explain why it's more natural, not just that it is. For vocabulary help: give collocations, not definitions (e.g., 'make a decision' not 'decide'). Vary register, formal, neutral, casual versions.

Job Seeker (Mid-Career)

Personal & Specialty

Field 1, What ChatGPT should know about you

I'm a mid-career product marketer with 8 years at two B2B SaaS companies. I'm targeting Head of Product Marketing roles at Series B-D startups. Comp target: $220k OTE, fully remote or NYC hybrid. I have a portfolio on my site and a dense LinkedIn.

Field 2, How ChatGPT should respond

For cover letters: 250 words max, one concrete story that matches a job-description requirement, one specific reason I want that company. Resume bullets: lead with outcome, then action, then context. Always quantify, revenue, reach, time saved. Never use 'passionate' or 'results-driven'.

Freelancer / Solopreneur

Personal & Specialty

Field 1, What ChatGPT should know about you

I'm a freelance brand designer billing $180/hr. I have 4-6 active clients, $220k revenue target. Services: identity, packaging, web. I work out of a contract template and hate scope creep. Tools: Figma, Notion, Bonsai, Stripe Payment Links.

Field 2, How ChatGPT should respond

Client emails: warm but business-professional, 3-4 sentences max, always name the next concrete action. For scope pushback: acknowledge → clarify the original scope → quote the change order. Never apologize for charging. For proposals: 3 tiers (starter / standard / expanded), anchored at standard.

Power User with ADHD

Personal & Specialty

Field 1, What ChatGPT should know about you

I'm a software engineer with ADHD, medicated. I lose time to hyperfocus and task-switching costs. I use ChatGPT as my external working memory. I care about shipping, not perfecting. I prefer short iterations.

Field 2, How ChatGPT should respond

Answer the question first, then context, never bury the lede. Use bullet lists over prose. For any multi-step process: chunk into 3-5 steps max per message, so I can act between them. If I ask for a plan, propose the smallest version that would be meaningful. Push back when I'm scope-creeping.

Parent (School-Age Kids)

Personal & Specialty

Field 1, What ChatGPT should know about you

I'm a parent of two kids (ages 8 and 11). I work full-time, my partner does too. I use ChatGPT for meal planning, homework help, birthday-party logistics, and pediatrician questions. I want honest answers, not reassurance.

Field 2, How ChatGPT should respond

For homework help: never give the answer, walk my kid through the thinking, matched to their grade level. For meal plans: weekly plan, single grocery list grouped by store section, recipes on one page. For medical questions: tell me what to watch for, when to call the doctor, and what to avoid doing at home.

05, Formula

Write Your Own From Scratch

If none of the 50 templates fits cleanly, use this formula. Both fields follow the same five-part structure, identity, domain, stack, constraints, voice, which keeps the model's behavior predictable.

Field 1 Formula

Identity: "I'm a [role] at [company type / size]"
Domain: "I work on [specific outputs or problems]"
Stack: "Tools: [3-6 tools]"
Audience / ICP: "My [readers / customers / students] are [specific]"
Stance: one sentence about what you value or reject

Field 2 Formula

Structure: how to open, middle, close every response
Length: default word / paragraph target
Tone: one adjective, plus one counter-example ("direct, not blunt")
Format preferences: bullets, tables, code blocks, headers
Anti-patterns: "Never use [word / phrase / format]."

Weak (vague, unverifiable)

"I'm a marketer. Be helpful and professional. Keep answers short."

Strong (specific, testable)

"I'm a B2B content marketer for devtools SaaS; audience is backend engineers. Respond with 2-sentence hook, 5 bullets, 1 takeaway, under 350 words. No emojis, no 'leverage', no 'unlock'."

06, Advanced

Advanced Techniques

Layer Projects for role-specific overrides

Keep your global Custom Instructions neutral and identity-level. Then create a ChatGPT Project per active initiative ("Q2 pricing rework", "iOS app", "personal taxes") with Project Instructions that override format, stack, or tone. Projects inherit your global rules and only diverge where needed.

Use ordered numbered lists, not prose paragraphs

Model adherence is consistently higher on enumerated rules. Convert "I want responses to be concise, well-structured, and action-oriented" into three numbered rules with concrete definitions of "concise" (word count), "structured" (format), and "action-oriented" (what appears at the end).

Put anti-patterns in their own block

End Field 2 with a "Never" list: "Never use emojis. Never write 'in today's world'. Never hedge with 'it depends' without giving the default answer first." Negation rules are most reliable when grouped.

Include a worked example for edge behavior

For high-stakes format rules, paste one 30-word example of the response you want. The model follows few-shot examples more reliably than abstract descriptions, if you want SQL with specific indentation and CTE conventions, show one query.

Rotate seasonally

Review your instructions monthly. If your tools, focus, or team shifted, the instructions are probably stale. The most common sign: you're adding the same clarification to three recent prompts, that's a rule belonging upstream.

07, Troubleshoot

When ChatGPT Ignores Your Instructions

Symptom: responses are still long and hedged

Cause: your length rule is prose, not a number. Fix: replace "keep it concise" with "default length 250 words. Hard cap 400." Numbers adhere; adjectives don't.

Symptom: format drifts mid-conversation

Cause: the instructions are in Field 1 (context) instead of Field 2 (behavior). Move format rules to Field 2. Field 1 should describe you; Field 2 should describe the response.

Symptom: "don't do X" is ignored

Cause: negative rules without positive replacements get paraphrased away. Fix: pair every prohibition with a positive alternative, "never use 'leverage', use 'use' or 'apply' instead."

Symptom: conflicting instructions

Cause: "be detailed" and "keep it short" both appear. The model picks one unpredictably. Fix: audit your instructions for contradictions, typically around tone ("witty but formal") and length ("detailed but brief").

Symptom: instructions work for text but not images

DALL-E requests are built from your chat text plus Custom Instructions. If you want different rules for images, specify explicitly: "For image generation only, prefer flat vector style and pastel palette. Text responses remain business-professional."

Nuclear option: reset and re-seed

If behavior is drifting and you can't pinpoint why, clear both fields, wait one new chat, and re-paste a clean version. Memory sometimes accumulates conflicting preferences from past conversations, check Settings → Personalization → Memory and delete any stale items.

08, FAQ

Frequently Asked Questions

Related guides

ChatGPT Prompts: The Complete Library

300+ copy-paste prompts organized by task.

ChatGPT for Business Productivity

Workflows for ops, finance, and people leaders.

ChatGPT Prompt Generator

Describe your task, get a ready-to-paste prompt.

Avoiding Hallucinations

Grounding techniques that actually work in production.

Few-Shot Prompting

When example-based prompts beat instruction-based ones.

GPT-5 Reasoning Guide

How reasoning models change your prompting defaults.

01, Intro

What are ChatGPT Custom Instructions?

Field 1, "What would you like ChatGPT to know about you?"

Your role, domain, tools, ICP, constraints, and the stable facts about your work. Think of it as the "about" section of your professional profile, written for an assistant.

Field 2, "How would you like ChatGPT to respond?"

Format, tone, length, and house rules. Not what to say, how to say it.

02, Setup

Step-by-Step Setup (60 Seconds)

1
Open ChatGPT Settings
Click your profile icon (bottom-left on desktop, top-right on mobile) → Customize ChatGPT. Or go direct: Settings → Personalization → Custom Instructions.
2
Toggle Custom Instructions ON
If the feature isn't enabled yet, flip the toggle. On ChatGPT Projects, each project has its own instructions that layer on top.
3
Paste Field 1, What to know about you
Copy from a template below and edit. Keep it ordered: role → domain → tools → constraints. 1,500 char limit.
4
Paste Field 2, How to respond
Lead with format rules, then tone, then anti-patterns ('never use X'). 1,500 char limit.
5
Save and test in a new chat
Run 2-3 prompts from your normal workflow. If outputs drift from your rules, the instructions are too long or have conflicting rules, trim and retest.

03, Why

Why Custom Instructions Work

Shorter prompts

Skip the "I'm a [role] working on [project] targeting [audience]" preamble. It's already loaded.

Higher first-try quality

Format and tone land correctly on the first reply. No "make it shorter / more technical / no emojis" follow-ups.

Consistent voice

Every chat inherits the same rules, no drift across threads or over time.

Team-shareable style

Paste your team's Custom Instructions into a Custom GPT to turn personal defaults into a shared assistant.

04, Templates

50+ Copy-Paste Templates by Role

Each template is two blocks, Field 1 and Field 2. Pick the closest to your role, paste both, then edit the specifics (tools, audience, constraints) in the first pass. Ship, use for a week, refine.

Writing & Content

Content Marketer (SaaS)

Writing & Content

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

SEO Content Writer

Writing & Content

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Direct-Response Copywriter

Writing & Content

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Novelist / Fiction Writer

Writing & Content

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Technical Writer

Writing & Content

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Newsletter Editor

Writing & Content

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Business & Strategy

SaaS Founder (Seed–Series A)

Business & Strategy

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Startup Operator (COO-track)

Business & Strategy

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Management Consultant

Business & Strategy

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Product Manager

Business & Strategy

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Business Analyst

Business & Strategy

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Executive Assistant (C-suite)

Business & Strategy

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Engineering & Technical

Backend Engineer

Engineering & Technical

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Frontend Engineer (React)

Engineering & Technical

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Full-Stack Developer

Engineering & Technical

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

DevOps / SRE

Engineering & Technical

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Data Scientist

Engineering & Technical

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

ML Engineer

Engineering & Technical

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Marketing & Sales

Paid Ads Specialist

Marketing & Sales

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Email Marketer

Marketing & Sales

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Social Media Manager

Marketing & Sales

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Sales Development Rep

Marketing & Sales

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Account Executive

Marketing & Sales

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Growth Marketer

Marketing & Sales

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Creative & Design

UX Designer

Creative & Design

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Graphic Designer

Creative & Design

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

YouTuber / Video Creator

Creative & Design

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Podcaster

Creative & Design

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Education & Research

PhD Researcher

Education & Research

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

University Professor

Education & Research

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

High School Teacher

Education & Research

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Graduate Student (STEM)

Education & Research

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Professional Services

Attorney (Corporate)

Professional Services

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Financial Planner (CFP)

Professional Services

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Accountant (CPA)

Professional Services

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Real Estate Agent

Professional Services

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Insurance Agent

Professional Services

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

HR Manager

Professional Services

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Healthcare & Wellness

Therapist (Private Practice)

Healthcare & Wellness

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

RN / Nurse Educator

Healthcare & Wellness

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Fitness Coach

Healthcare & Wellness

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Trade & Operations

Customer Support Lead

Trade & Operations

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Operations Manager

Trade & Operations

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Supply Chain Analyst

Trade & Operations

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Personal & Specialty

Language Learner (Non-Native English)

Personal & Specialty

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Job Seeker (Mid-Career)

Personal & Specialty

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Freelancer / Solopreneur

Personal & Specialty

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Power User with ADHD

Personal & Specialty

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

Parent (School-Age Kids)

Personal & Specialty

Field 1, What ChatGPT should know about you

Field 2, How ChatGPT should respond

05, Formula

Write Your Own From Scratch

If none of the 50 templates fits cleanly, use this formula. Both fields follow the same five-part structure, identity, domain, stack, constraints, voice, which keeps the model's behavior predictable.

Field 1 Formula

Identity: "I'm a [role] at [company type / size]"
Domain: "I work on [specific outputs or problems]"
Stack: "Tools: [3-6 tools]"
Audience / ICP: "My [readers / customers / students] are [specific]"
Stance: one sentence about what you value or reject

Field 2 Formula

Structure: how to open, middle, close every response
Length: default word / paragraph target
Tone: one adjective, plus one counter-example ("direct, not blunt")
Format preferences: bullets, tables, code blocks, headers
Anti-patterns: "Never use [word / phrase / format]."

Weak (vague, unverifiable)

"I'm a marketer. Be helpful and professional. Keep answers short."

Strong (specific, testable)

"I'm a B2B content marketer for devtools SaaS; audience is backend engineers. Respond with 2-sentence hook, 5 bullets, 1 takeaway, under 350 words. No emojis, no 'leverage', no 'unlock'."

06, Advanced

Advanced Techniques

Layer Projects for role-specific overrides

Use ordered numbered lists, not prose paragraphs

Put anti-patterns in their own block

Include a worked example for edge behavior

Rotate seasonally

07, Troubleshoot

When ChatGPT Ignores Your Instructions

Symptom: responses are still long and hedged

Cause: your length rule is prose, not a number. Fix: replace "keep it concise" with "default length 250 words. Hard cap 400." Numbers adhere; adjectives don't.

Symptom: format drifts mid-conversation

Cause: the instructions are in Field 1 (context) instead of Field 2 (behavior). Move format rules to Field 2. Field 1 should describe you; Field 2 should describe the response.

Symptom: "don't do X" is ignored

Cause: negative rules without positive replacements get paraphrased away. Fix: pair every prohibition with a positive alternative, "never use 'leverage', use 'use' or 'apply' instead."

What are ChatGPT Custom Instructions?

Field 1, "What would you like ChatGPT to know about you?"

Field 2, "How would you like ChatGPT to respond?"

Step-by-Step Setup (60 Seconds)

Open ChatGPT Settings

Toggle Custom Instructions ON

Paste Field 1, What to know about you

Paste Field 2, How to respond

Save and test in a new chat

Why Custom Instructions Work

Shorter prompts

Higher first-try quality

Consistent voice

Team-shareable style

50+ Copy-Paste Templates by Role

Writing & Content

Content Marketer (SaaS)

SEO Content Writer

Direct-Response Copywriter

Novelist / Fiction Writer

Technical Writer

Newsletter Editor

Business & Strategy

SaaS Founder (Seed–Series A)

Startup Operator (COO-track)

Management Consultant

Product Manager

Business Analyst

Executive Assistant (C-suite)

Engineering & Technical

Backend Engineer

Frontend Engineer (React)

Full-Stack Developer

DevOps / SRE

Data Scientist

ML Engineer

Marketing & Sales

Paid Ads Specialist

Email Marketer

Social Media Manager

Sales Development Rep

Account Executive

Growth Marketer

Creative & Design

UX Designer

Graphic Designer

YouTuber / Video Creator

Podcaster

Education & Research

PhD Researcher

University Professor

High School Teacher

Graduate Student (STEM)

Professional Services

Attorney (Corporate)

Financial Planner (CFP)

Accountant (CPA)

Real Estate Agent

Insurance Agent

HR Manager

Healthcare & Wellness

Therapist (Private Practice)

RN / Nurse Educator

Fitness Coach

Trade & Operations

Customer Support Lead

Operations Manager

Supply Chain Analyst

Personal & Specialty

Language Learner (Non-Native English)

Job Seeker (Mid-Career)

Freelancer / Solopreneur

Power User with ADHD

Parent (School-Age Kids)

Write Your Own From Scratch

Field 1 Formula

Field 2 Formula

Weak (vague, unverifiable)

Strong (specific, testable)

Advanced Techniques