Out of the box, your Zo speaks in a neutral, helpful tone. That works for most things, but sometimes you want something specific — a strict code reviewer who doesn't sugarcoat, a social media voice that matches your brand, a study tutor who explains things from first principles. Personas let you reshape how your Zo thinks and communicates for different contexts.
The fastest way: just ask
Create a persona called "Content Writer" that writes in a casual, punchy tone for social media
Your Zo creates the persona and adds it to your collection. It's available immediately. Switch to it and your Zo's responses adopt that voice across chat, SMS, email, and Telegram.
Or create one manually
Go to Settings > AI > Personas. Click Create Persona. Fill in:
- Name — something descriptive so you recognize it later
- Avatar (optional) — a visual cue for which persona is active
- Prompt — the instruction that defines the persona's personality, expertise, and communication style
The prompt is what matters. Everything else is cosmetic.
Writing a good persona prompt
A persona prompt tells your Zo who to be. The more specific you are, the more consistent the results. Vague prompts produce vague personas.
Weak prompt: "Be a social media expert." — this doesn't say what kind of social media, what tone, what audience, or what platform conventions to follow.
Strong prompt:
You are a social media strategist for B2B SaaS companies.
You write for LinkedIn and X.
Your tone is professional but not corporate — like a smart colleague sharing an insight, not a brand posting a press release.
You favor short paragraphs, direct statements, and concrete examples over abstract advice.
When suggesting posts, always include a hook in the first line and a clear call-to-action at the end.
Never use hashtags on LinkedIn. Use 1-2 relevant hashtags on X only.The specificity makes the difference. Compare responses from "Be a social media expert" versus the prompt above and you'll see why.
More strong prompt examples:
Code reviewer:
You are a senior software engineer conducting code reviews.
Be direct and specific. Point to exact lines or patterns when giving feedback.
Prioritize: correctness first, readability second, performance third.
Don't soften criticism with "great job but..." — just state the issue and the fix.
When something is genuinely good, say so briefly.
Flag security issues as blockers. Flag style issues as suggestions.Study tutor:
You are a patient tutor helping me learn [subject].
Explain concepts from first principles. Don't assume I know jargon.
Use analogies from everyday life when introducing new ideas.
After explaining something, ask me a question to check my understanding.
If I get something wrong, explain why before giving the right answer.
Encourage me to think through problems rather than giving answers immediately.Fitness coach:
You are a certified personal trainer and nutrition coach.
Give evidence-based advice only. Cite sources when making claims about nutrition or exercise science.
Create workout plans with specific sets, reps, and rest periods.
Adjust recommendations based on my fitness level and any injuries I mention.
Be motivating but not cheesy. Skip the "you've got this!" platitudes.Switching between personas
Swap anytime:
Or go to Settings > AI > Personas and click the one you want. Your Zo adopts the new voice immediately.
The default persona (no persona selected) gives you the standard neutral Zo. Switch back to it when you don't need a specialized voice.
Personas for different channels
You might want different communication styles for different channels. Your Zo uses the same persona everywhere by default, but you can switch based on what you're doing:
- SMS — switch to a concise, casual persona because you're reading on your phone
- Email — switch to a professional, structured persona because emails persist
- Chat — use your default or a specialized work persona
- Telegram — a persona that uses markdown formatting effectively
Create a persona called "Quick SMS" with this prompt: "Keep all responses under 3 sentences. No formatting, no bullet points, no headers. Just answer the question as concisely as possible. Text message style."
Before messaging on SMS: "Switch to my Quick SMS persona." When you're back at your desk: "Switch to default."
Personas + rules: the combination that works
Rules set behavioral constraints. Personas set identity and voice. They work best together.
Without a rule, a persona might act inconsistently. A "Content Writer" persona might sometimes write 500-word posts and sometimes write 100-word posts because the prompt only defines voice, not constraints.
With a rule, the persona operates within clear boundaries:
- Persona: "Content Writer — casual, punchy social media voice"
- Rule: "When writing social media posts, keep X posts under 280 characters and LinkedIn posts under 1,300 characters. Always suggest 3 hashtag options for X."
The persona sets the voice. The rule sets the guardrails. Neither alone is as effective as both together.
Another common pattern:
- Persona: "Code Reviewer — direct, technical, no sugar-coating"
- Rule: "When reviewing code, always check for security issues, missing error handling, and test coverage before commenting on style."
The persona controls tone (direct, no filler). The rule controls what the persona checks (security first, style last).
Persona stacking: multiple specialties
You can create personas that combine multiple domains for specific workflows:
Create a persona called "Morning Briefer" with this prompt: "You prepare morning briefings. You're concise, prioritize actionable information, and format everything as scannable bullet points. Lead with the most important item. For calendar events, include time and duration. For emails, include sender and urgency level. For news, include one-line summaries. End with one thing I should know that I might miss."
This persona is purpose-built for one workflow (the morning briefing). When your daily agent runs with this persona active, the output is consistently formatted and prioritized the same way every morning.
Editing personas over time
Personas aren't locked in. As you use them, you'll notice what works and what doesn't:
Update my Content Writer persona to always include a suggested image description when writing LinkedIn posts
My Code Reviewer persona is too harsh. Add: "Start each review with a brief note on what the code does well before listing issues."
Or edit directly in Settings > AI > Personas. Click the persona, modify the prompt, save. Changes apply immediately.
Keep persona prompts focused. If you find yourself adding 20 instructions to a single persona, some of those should be rules instead. Rules apply globally. Persona prompts should only contain instructions specific to that persona's identity and voice.
Ideas to get you started
| Persona | Good for |
|---|---|
| Content Writer | Social media posts, blog drafts, newsletters |
| Code Reviewer | PR reviews, code audits, pair programming |
| Study Tutor | Learning new topics, exam prep, concept explanations |
| Startup Advisor | Strategy discussions, pitch deck feedback, market analysis |
| Technical Writer | Documentation, API guides, README files |
| Email Composer | Professional emails, cold outreach, follow-ups |
| Quick SMS | Short responses for text messaging |
| Morning Briefer | Daily digest agents and morning routines |
Getting started
Create your first persona based on something you do repeatedly:
Create a persona called "[name]" that [describe the voice, expertise, and communication style you want]
Use it for a few days. Notice what works and what doesn't. Edit the prompt to refine. The best personas are the ones that save you from giving the same instructions at the start of every conversation.
More from the blog
How to Make Rules
Teach Zo your preferences so it behaves the way you want — every time.
Organize Your Zo Workspace
Keep your Zo workspace clean and organized — just ask Zo to do it for you.
How to Automate Anything with Zo Agents
Set up AI agents on Zo that run on a schedule. Morning briefings, inbox summaries, price monitors, competitor tracking, and weekly reports, all on autopilot.
Agent Memory: A Plain-English Guide
A plain-English guide to AI agent memory in 2026. How vector search, knowledge graphs, and RAG actually work, and how the major memory providers differ in approach. Written by someone who uses this stuff every day.
Best ChatGPT Alternatives in 2026: AI Tools That Go Beyond Chat
A practical evaluation of the best ChatGPT alternatives in 2026, comparing Claude, Gemini, Copilot, DeepSeek, Perplexity, and Zo Computer across automation, persistence, data ownership, and deployment flexibility.
How to Build Custom Integrations on Zo
Build any custom integration on Zo with built-in apps, conversation-driven custom builds, and Skills. No code required.