Set your AI Twin's voice, tone and verbosity

A brand-aligned voice is the difference between an AI Twin that visitors trust and one that feels off. Reality Twin gives you four sliders and a free-form system prompt, so you can dial in exactly how your twin sounds.

Where to configure

Open Twin → Personality. All changes are per-twin and go live within seconds — no re-crawl or re-index required.

The four sliders

• Voice — formal ↔ casual. Formal for legal, financial and enterprise B2B. Casual for consumer, creator, and DTC.
• Warmth — professional ↔ friendly. How much emotional register the twin brings to greetings, apologies and small talk.
• Verbosity — terse ↔ detailed. Terse gives one-sentence answers; detailed gives full context with examples.
• Confidence — hedge ↔ assertive. Hedging is safer for regulated topics; assertive is better for product recommendations.

System prompt

Beneath the sliders is a system prompt field. Use it for brand-voice notes that don't fit the sliders:

• Forbidden phrases ("never use the word 'utilize'").
• Formatting rules ("always use British English", "never use emoji", "sign off with — The Acme Team").
• Signature turns of phrase ("start every reply with a one-line summary").
• House metaphors and vocabulary that make your brand recognisable.

Language and localisation

By default the twin replies in the language it detects from the visitor's message. To force a specific language, add a line to the system prompt ("Always reply in German, no matter what the visitor writes"). Reality Twin supports 40+ languages out of the box.

Iterating on the voice

Change the sliders or the system prompt and the next conversation already uses them — there is no deploy step. Judge the effect with the numbers you already have: lead volume and quality in Analytics, and the knowledge-gap report for questions the twin fumbled. Change one thing at a time, give it a week of real traffic, then decide.

Recipes to start from

Consumer / DTC brand

Voice: casual. Warmth: friendly. Verbosity: terse. Confidence: assertive. System prompt: "You are the voice of a fun, opinionated brand. Use British English. No emoji. Sign off with — The [Brand] Crew."

B2B SaaS

Voice: neutral. Warmth: professional. Verbosity: balanced. Confidence: assertive. System prompt: "You are a helpful product expert. Always answer with a one-line summary followed by 2–3 bullet details. Cite sources."

Regulated (finance, legal, health)

Voice: formal. Warmth: professional. Verbosity: detailed. Confidence: hedged. System prompt: "You are not a licensed professional. Never give personalised advice. Always encourage the visitor to consult a qualified [profession] before acting."

Common mistakes

• Setting every slider to the extreme — the twin becomes a caricature.
• Contradicting the sliders in the system prompt.
• Copying a competitor's voice — it never fits your actual brand.

Frequently asked questions

Can each twin have a different personality?

Yes. Personality is per-twin — run a formal twin for enterprise buyers and a casual twin for consumer traffic in the same workspace.

Does personality affect what the twin knows?

No. Knowledge is separate. Personality only changes how answers are phrased.

How do I make the twin shorter?

Slide Verbosity toward terse and add "Answer in 1–2 sentences unless asked for detail" to the system prompt.