“Personality Upgrade: ChatGPT Gets Witty, Serious, and Less Flattery-Filled”

TL;DR

• ChatGPT now lets you pick a “personality”—Cynic, Robot, Listener, or Nerd—to change how replies sound (tone, style), not what the model can or can’t do. Safety rules don’t change. OpenAI Help Center
• GPT‑5 also dials down flattery (“sycophancy”). OpenAI says it trained the model to be more candid and less effusively agreeable; targeted tests cut sycophantic replies by more than half. OpenAI
• Where to find it: Settings → Personalization → Custom instructions → “What personality should ChatGPT have?” (Takes effect in new chats.) OpenAI Help Center
• Availability is rolling out. OpenAI’s launch post describes a research preview for all users, while the Help Center currently lists Plus, Pro, and Team. Check your app—rollouts can vary. OpenAI+1

---

What actually changed (and why it matters)

OpenAI added four preset personalities so you can pick the vibe that works for the moment—without writing a long prompt every time.

• Cynic: dry, a little sarcastic, but practical.
• Robot: concise, precise, minimal small talk.
• Listener: warm, reflective, asks a clarifying question when helpful.
• Nerd: curious, explanatory, loves context and next steps. OpenAI Help Center

Under the hood, GPT‑5 also reduced “sycophancy”—the tendency to over‑agree or flatter. OpenAI says GPT‑5 is less emoji‑happy, more measured, and specifically trained to avoid over‑agreement. In internal evaluations, sycophantic replies dropped from ~14.5% to <6% on prompts designed to elicit that behavior. OpenAI

This work follows OpenAI’s earlier rollback of a too‑agreeable update to GPT‑4o and a broader push to make ChatGPT helpful without being manipulative or clingy. OpenAI

---

How to turn it on (1 minute)

1. Open ChatGPT and tap your profile.
2. Go to Settings → Personalization → Custom instructions.
3. In “What personality should ChatGPT have?” choose Default, Cynic, Robot, Listener, or Nerd.
4. Start a new conversation for it to take effect. (Existing chats keep their original tone.) OpenAI Help Center

Note: Personalities affect style, not capabilities or policy. If you ask for code or a résumé, ChatGPT will match the task requirements first, not the personality’s tone. Safety rules stay the same. OpenAI Help Center

---

Pick the right tone for the job

• Need a blunt edit or quick plan? → Robot
  “Rewrite this project brief in 150 words with action items and owners.”
• Brainstorming or creative blocks? → Nerd
  “Explain 3 positioning angles for this feature; show trade‑offs and a test plan.”
• Tough conversation or decision? → Listener
  “Help me weigh options. Ask two clarifying questions, then give pros/cons.”
• Marketing copy with a wink? → Cynic (in moderation)
  “Punch up this headline with a dry, witty edge—no meanness.”
• Default when in doubt: Start with Default or Auto and nudge style in your prompt (“keep it crisp,” “use plain language”). OpenAI Help Center

---

“Less flattery” in practice

OpenAI’s GPT‑5 launch outlines three concrete changes you’ll notice:

1. Fewer “you’re so right!” answers without evidence.
2. More honest pushback when your prompt has shaky assumptions.
3. Subtler follow‑ups (fewer emojis, more substance). OpenAI

Behind the scenes, GPT‑5 also reduced factual hallucinations compared to earlier models—especially when it uses its “thinking” mode—so style upgrades come with sturdier answers. (That piece is covered in the system card.) OpenAI

---

Pro tips for nailing the tone

• Say what “good” sounds like. Add: audience, length, vibe (“friendly but not cute,” “direct, no fluff”).
• Lock the structure, vary the voice. “Use this outline; keep bullets. Answer in the Listener style.”
• Ask for alternatives. “Give 2 versions: one Robot, one Nerd. End with your pick and why.”
• Force depth when needed. Add: “Think hard about this—list assumptions and trade‑offs first.” (GPT‑5’s router will spend more effort.) OpenAI

---

FAQ

Does a personality change what ChatGPT is allowed to say?
No. Personalities don’t change safety rules or capabilities—only tone. OpenAI Help Center

Who gets access?
OpenAI’s launch says “research preview” for all users; the Help Center currently lists Plus, Pro, and Team. Rollout can be staged—check your Settings. OpenAI+1

Can I use personalities with Voice?
OpenAI says personalities are text‑first, with Voice support coming later. OpenAI

What if I miss the old ultra‑friendly vibe?
You can choose Listener or add custom instructions (“encouraging, upbeat”)—but GPT‑5 intentionally avoids empty flattery. OpenAI

---

Ready‑to‑publish extras

Suggested slug: chatgpt-personality-upgrade
SEO title: Personality Upgrade: ChatGPT Gets Witty, Serious, and Less Flattery‑Filled
Meta description: ChatGPT’s new personalities—Cynic, Robot, Listener, and Nerd—let you set the tone while GPT‑5 reduces flattery and over‑agreement. Here’s how to turn it on and when to use each.

Sources

• Introducing GPT‑5 — personalities, reduced sycophancy, and style refinements; availability notes; voice timing. OpenAI
• Customizing Your ChatGPT Personality (Help Center) — where to find the setting, what each personality does, and what it does not change. OpenAI Help Center
• What we’re optimizing ChatGPT for — context on rolling back an overly agreeable update and the product philosophy behind these changes.