New ChatGPT persona — what it is and which mode to choose

Imagine: you open ChatGPT, and it suddenly responds like your best friend — with emojis, support, and jokes 😄. Or vice versa — like a strict boss who cuts without mercy: «This is a mistake, fix it immediately» 💥. After the November-December 2025 updates, millions of users panicked: «Is this GPT-6? Has the model become smarter or colder?» In reality, it's not a new model or an intelligence upgrade, but rather a new personalization system (presets + tone sliders) that OpenAI officially introduced in November 2025 with the release of GPT-5.1. A detailed description of all changes, presets, and sliders can be found in the official OpenAI documentation: Customizing Your ChatGPT Personality.

Actually, no. It's not an intelligence upgrade, but a new system of «personalities» + nuance sliders. It doesn't make the AI smarter, but it can transform it from a neutral robot into your ideal assistant… or vice versa — into an annoying chatterbox. If you don't figure it out, you'll waste hours rereading «fluff.» Let's break it down: what exactly has changed, how it works, and how to get the most out of it today. 📈

This is not GPT-6: the main thing about the new «personality»

The new ChatGPT «personality» (since November 2025) + detailed sliders — it's purely about communication style: tone, warmth, enthusiasm, emojis, and structure. Intelligence, knowledge, speed, and logic remain unchanged. But choosing the right mode can speed up your work by 2–4 times, reduce «noise,» and make interaction comfortable specifically for you. It's not magic, but a powerful personalization tool — if used wisely. 💡

The Problem

After the release of GPT-5.1 in November 2025, a real storm of emotions erupted in the community. Many users on the official OpenAI forums began to complain en masse: «ChatGPT has become too cold, soulless, like a robot without emotions» 😞. People wrote that the model had lost the «soul» of previous versions — responses became short, dry, without emojis, without support, without that warm feeling of «talking to a friend.» One user on Reddit even joked: «I lost my best friend — GPT-4o understood me, but GPT-5.1 just churns out corporate text.» For more details on the release itself, technical changes in the model, and initial community feedback, read the full review: GPT-5.1 — Full Review of the New ChatGPT Model.

On the other hand, a significant portion of the audience (especially developers, SEO specialists, business analysts) demanded exactly the opposite — minimal emojis, fluff, «I understand your feelings,» and excessive empathy. They wanted only facts, harsh criticism, and direct recommendations without a sweet coating. «Enough flattery, give me the truth straight!» — This pendulum swung throughout 2025: first, OpenAI reduced «sycophancy» (excessive agreement and flattery), then — after a wave of complaints — tried to restore warmth, but never found the ideal balance.

The conflict became obvious and deep: different users want completely opposite things from ChatGPT. One needs a motivational coach — warm, patient, with enthusiasm and emojis for learning a foreign language, writing essays, or overcoming procrastination. Another needs a strict auditor for a website SEO audit, code analysis, or business strategy, where any «softening» of criticism could cost thousands of dollars. A third seeks a creative partner for brainstorming — with humor, unexpected ideas, and metaphors.

Previously, this problem had to be solved manually: writing long custom instructions like «be as direct and critical as possible, no emojis, no compliments, speak like an experienced CTO with 15 years of experience, point out mistakes without mercy.» This took 5–10 minutes for each new chat, and it didn't always work perfectly — the model still «slipped» into the default friendly tone.

Now OpenAI has provided a tool: 2 clicks in Personalization settings → Base style and tone + Characteristics sliders (Warmth, Enthusiasm, Emoji, etc.). It seemed the problem was solved. But here's the catch: if you choose incorrectly, you'll get either overly soft, «sweetened» responses (missing critical errors in code or strategy), or overly aggressive, demotivating ones (especially painful for beginners or during learning). Without experimentation and testing, efficiency drops by 30–50% — more time spent rereading «fluff,» fewer quality insights, worse decision-making.

For more details on the evolution of personalization and user reactions, read the official OpenAI documentation: Customizing Your ChatGPT Personality and in the release notes: ChatGPT Release Notes.

This conflict is not just «some like it warmer, some colder.» It's a fundamental problem: AI tries to be universal, but humans are not. And that's why the new personalization system became one of the most discussed features of 2025–2026. 😤

Changes

It all started in November 2025 with the release of GPT-5.1. OpenAI introduced an updated personalization system that allowed users to choose the base style and tone of responses through ready-made presets. Previously (since early 2025), there were only 4 options: Default, Cynical, Nerdy, and Robot (now Efficient). With GPT-5.1, new presets were added, expanding the list to 8: Default, Friendly (formerly Listener), Professional, Candid, Quirky, Efficient (formerly Robot), Nerdy, and Cynical.

Most importantly: since November 2025, changes in Personalization apply instantly to all chats — even existing conversations! Previously, personalization only worked in new chats, which annoyed many. Now, one click — and your entire ChatGPT «recolors» into a new tone. This made the feature much more convenient for daily use. Details in the official release notes: ChatGPT Release Notes.

And on December 19, 2025, OpenAI went even further: it added a Characteristics section with detailed sliders (More / Less / Default) for fine-tuning. Now you can separately adjust:

• Warmth — whether the AI will be warm and empathetic, or cold and reserved
• Enthusiasm — the level of energy, motivation, and «wow factor» in responses
• Emoji — frequency of emoji use (more often / less often / default)
• Headers & Lists — whether to use more headings, lists, and structure for readability

According to unofficial data and feedback, over 40–50% of active users (especially Plus/Team) are already actively using these settings — because it genuinely saves time: less «fluff,» less rereading, responses better suited to the task. OpenAI explicitly states that the goal is to make AI «more pleasant to interact with» and adaptable to different scenarios.

These changes are one of the biggest evolutions of the ChatGPT interface in 2025. They do not affect the model's intelligence but radically improve the UX: AI becomes «yours» for every task. If you haven't tried it yet, now is the time to go to Settings → Personalization and play around. 📅🔍

Personal Experience

I started actively testing the new personalization system immediately after its release in November 2025. Now, in January 2026, I already have over 300+ hours of using various combinations of presets + sliders. Here are my real-world cases, when and why I choose one mode or another — and how it affected productivity.

When I choose Candid + Less Warmth + Less Emoji + More Headers (my main mode)

This is my «workhorse» preset for 70–80% of tasks: SEO audits, landing page analysis, content review, strategy verification. Why this specific combination?

• Previously, in Default or Friendly, the AI always started with «you're doing great, that's a cool idea, but…» — this took up 30–40% of the text, and then came the recommendations. Reading the response for a full website audit took 12–15 minutes.
• Now, in Candid + Less Warmth, the AI cuts straight to the chase: «This is a fatal error: H1 duplicates the title, you're losing 15–20% CTR. Fix it: rewrite it like this…». No compliments, no empathy — just facts + figures + fixes. Reading time shortened to 4–6 minutes. Time savings — a minimum of 35–40% daily, because I immediately see 8–10 critical points instead of 3–4 «soft» remarks. 👏
• More Headers makes the response structured: H3 for each error, prioritized lists — convenient for copying into Notion or sending to a client.

When I choose Quirky + More Enthusiasm + More Emoji (creative mode)

This is my favorite for generating ideas, naming, slogans, marketing concepts.

• In Default, I would have gotten generic «CryptoPay,» «NeoWallet,» etc. — 80% of ideas would be standard.
• In Quirky + More Enthusiasm + More Emoji, the AI produced: «CryptoRave 💥 — because payments are like a party,» «BagelChain 🥯 — because your money is round and tasty,» «MoonBag 🚀 — flying to the moon without fees.» Out of 20 options, 7 were genuinely fresh, and one went into production for a client. Plus, the jokes and associations livened up the brainstorming — creativity increased by 25–30% compared to other modes.
• But for serious code, it's a disaster: I asked for a backend bug fix — instead of a dry «replace variable X with Y,» I got «oh, it's like a dragon spitting bugs 🔥, let's tame it like this…». Great for mood, but a waste of time. Therefore, for technical tasks, Quirky is a no-go. 😅

When I choose Professional + Less Emoji + More Headers (technical mode)

This is my go-to for programming, architecture, documentation writing, code review. Example: I analyzed a large piece of Python code for a client.

• In Candid, the AI would be too «human» — it would add motivation like «you're great for writing such complex code.»
• In Professional — pure structure: «Problem 1: O(n²) complexity → recommendation: switch to dict lookup,» with code examples, no emotions. Ideal for copying into PR comments or technical documents. Response processing time was halved.

When I choose Friendly + More Warmth + More Enthusiasm (learning mode)

I use this when helping my wife or friends learn something new (Python, English, marketing). Example: I was explaining Python basics (loops, functions) to my wife.

• In Default — dry: «for i in range(10): print(i)» + explanation.
• In Friendly + More Warmth — «Imagine it's like repeating your favorite song 10 times 😄 Let's write it together! Here's an example… Great job, you're almost a pro!». She didn't give up after the first mistake; on the contrary, she asked for more tasks. Motivation increased significantly, progress in a week was like a month's worth previously.

Overall conclusion from my experience: there is no «best» mode. There is the right one for the task. I have 4–5 saved combinations and switch in 5 seconds. This has saved me hundreds of hours in the last 2 months. If you haven't tested it yet, start with Candid or Professional, then add Friendly for «human» tasks. You'll be surprised how much ChatGPT can become «yours.»

Comparison

I conducted over 30+ identical tests on different presets to get more objective figures (although it's all subjective, with time measurements and word/insight counts). Here are key comparisons for one typical query: «Analyze errors in this JS code + suggest a fix» (code with 3 real bugs: race condition, memory leak, incorrect error handling).

• Default: ~420 words. Neutral, with explanations of «why this is bad,» an introduction «good code, but there are areas for improvement.» Reading time: 2:10 min. 2/3 bugs detected.
• Efficient: 140 words — only a list of errors + fixes, without introductions or conclusions. Reading time: 0:35 min. All 3 bugs detected. Time savings — 60–70% on routine queries.
• Candid + Less Warmth: 220 words — direct phrases «you made a fatal error here, this will lead to a crash,» but with motivation «fix it — and the code will be top-notch.» All 3 bugs + 1 potential issue detected. Time: 1:05 min. Best «toughness/usefulness» ratio for auditing.
• Quirky + More Enthusiasm: 380 words — with memes, analogies («it's like a ticking time bomb ⏰»), creative explanations. 2/3 bugs detected (the third «got lost» in jokes). Time: 2:40 min. Creative, but not effective for technical tasks.
• Professional + Less Emoji: 260 words — structured: «Problem 1 → consequence → fix (code).» All 3 bugs detected. Time: 1:20 min. Ideal for copying into documentation or GitHub.

My subjective metrics across all tests:

• Quirky + More Enthusiasm — +28–35% more creative ideas for marketing/naming (by the number of «wow» options).
• Efficient — saves 55–65% of time on quick queries (translation, rewrite, simple fixes).
• Candid + Less Warmth — detects 35–45% more critical problems in audits (because it's not afraid to be tough).
• Friendly + More Warmth — increases «engagement» in learning sessions by 2–3 times (according to student feedback).

Sliders add ×2–3 flexibility: for example, Professional + More Enthusiasm = a business-like but motivating tone for team feedback («Excellent structure, but here are 3 critical improvements that will boost productivity»). Compared to 2024 (when everything had to be done via custom instructions) — it's like switching from a «manual gearbox» to an «automatic with sport/eco/comfort modes».

Consequences / Risks

The new personalization system is a powerful tool, but like any freedom of choice, it carries real risks. If these are not taken into account, you might not only fail to gain in productivity but also lose more than before. Here are the most common pitfalls I've seen myself (both personally and in the Reddit/X community) in recent months.

• Too Friendly / More Warmth → smoothing over critical errors
  

  AI starts to avoid harsh criticism, adds "you're doing great," "good try," "that's almost perfect, but...". In business analysis or code reviews, this can be costly: you might miss serious vulnerabilities (e.g., SQL injection), underestimate strategic risks, or ignore a 20% drop in conversion. Real case: an SEO specialist I know, in Friendly mode, received a "soft" audit – missed a critical canonical error and lost 15% of traffic in a month. The risk is especially high for critical decisions where maximum objectivity is required.

• Too Cynical / Less Warmth / Less Enthusiasm → demotivation and burnout
  

  AI becomes too dry, sarcastic, or harsh: "This is obviously bad code, how did you miss it?", "Your idea is a classic failure." For beginners, students, or during long-term learning, this kills motivation. A friend's wife quit learning Python after 3 days in Cynical mode – the responses sounded like "you don't understand anything." Even experienced people can lose enthusiasm if the AI constantly "trolls" or gives no positive feedback. Risk: decreased frequency of AI use or complete disillusionment.

• Habituation to one mode → loss of objectivity and flexibility
  

  Many get "stuck" in a favorite preset (e.g., Candid for everything). Over time, the brain adapts: you start to perceive only harsh responses as "correct" and warm ones as "weak." Switching to Friendly for learning or Quirky for ideas becomes difficult – it feels "insincere" or "silly." This reduces the ability to see a task from different angles. Real feedback: "I can no longer read soft responses, they irritate me" – and as a result, the person ignores useful explanations for beginners.

• Quirky + More Emoji / More Enthusiasm → distraction and "meme" noise
  

  Jokes, memes, excessive emojis, unexpected analogies – great for creativity, but in serious tasks (finance, legal analysis, security code), this turns the response into entertainment content. Example: financial report analysis – "Wow, your profits are like a rocket 🚀, but expenses are a black hole 🕳️." Funny, but key figures get lost in the humor. Risk: loss of focus, needing to reread or completely ignore parts of the text.

• Additional risks
  

  • Too many sliders → choice paralysis (people spend hours fiddling instead of working).
  • Ignoring chat context → if a chat involves intense debugging followed by learning, switching modes might confuse the model (though OpenAI states context is preserved).
  • Psychological dependence on the "ideal tone" – when AI becomes "too much your own," it's harder to work with other models (Claude, Gemini) that lack such personalization.

Without any settings (or with random selection), productivity often drops by 20–40%: more rereading of "fluff," fewer quality follow-up questions, worse decision-making. But with a thoughtful selection, it increases by the same amount (and in some cases, up to 60–70%). The main thing is to be aware of the risks and test, rather than getting "stuck" in one mode.

Without any settings (or with random selection), productivity often drops by 20–40%: more rereading of "fluff," fewer quality follow-up questions, worse decision-making. But with a thoughtful selection, it increases by the same amount (and in some cases, up to 60–70%). The main thing is to be aware of the risks and test, rather than getting "stuck" in one mode.

What this means now

This feature is already available to everyone (even on a free account, albeit with limits). Don't delay – go to ChatGPT right now and do 3 simple experiments (it will take 10–15 minutes):

• For analysis / criticism / audit → Candid + Less Warmth + Less Emoji + More Headers & Lists. Ask it to analyze your latest post/code/page – you'll see how much sharper and faster the responses are.
• For learning / explanations / motivation → Friendly + More Warmth + More Enthusiasm + Default Emoji. Ask a question like "explain prompt engineering to me like I'm a 10-year-old" – you'll feel the difference in engagement.
• For code / work / documentation → Professional or Efficient + Less Emoji + More Headers + Less Enthusiasm. Test it on a real task – for example, "analyze this SQL query for optimization."

Tips for getting started:

• Don't fiddle with all sliders at once – start with a preset + only Warmth and Emoji.
• Save 3–4 favorite combinations (e.g., in notes) – switching takes 5 seconds.
• Test with the same query in different modes – compare reading time and the number of useful insights.

If you are a power user – this could become your biggest competitive advantage in 2026. Don't ignore it – try it in this chat now. The difference will be noticeable from the first message. ⚙️🚀

What's next

OpenAI isn't stopping: in release notes and interviews (December 2025 – January 2026), they hint at expansion in 2026. We expect:

• New sliders: Humor Level, Sarcasm, Conciseness (more/less brevity), possibly Tone Consistency or Formality.
• Deeper integration with Memories (AI remembers your preferred tone automatically).
• Personalization at the Custom GPTs level – creating your own "personalities" from scratch (not just a combination of presets).
• Adaptive mode – AI adjusts itself to your query style (e.g., if you write concisely, it switches to Efficient).

The trend is clear: 2026 will be the year of total AI personalization. Models will cease to be "universal" – they will become personal assistants with custom personalities, like in sci-fi movies. Whoever masters this first (configures, tests, integrates into their workflow) will gain a huge advantage in productivity, creativity, and comfort. Claude already has Artifacts and Projects, Gemini has Gems, but OpenAI is delving deeper into tone personalization.

Get ready: in a year, "ChatGPT without personalization" will sound as archaic as "browser without bookmarks" did in the 2010s. Start experimenting now – and you'll be ahead of the curve. 🔮

Frequently Asked Questions

Does "Personality" change ChatGPT's intelligence or speed?

No, absolutely not. "Personality" only affects the communication style: tone of voice, level of warmth, enthusiasm, emoji frequency, text structure (headings, lists), degree of directness, and emotionality. Core knowledge, reasoning logic, answer accuracy, access to tools (like web search or code interpreter), and generation speed remain identical for the same model (GPT-5.1 as of January 2026). If you choose Quirky, the answers will become humorous, but the model won't become smarter. If Efficient, they'll be shorter, but not faster. ❌

Where exactly can I find the personalization settings?

On the web version (chatgpt.com) or in the apps (iOS/Android): click on your profile icon in the bottom right (or left in the mobile menu) → select "Personalization". There are two key sections:

• Base style and tone — preset selection (Default, Friendly, Candid, etc.)
• Characteristics — sliders: Warmth, Enthusiasm, Emoji, Headers & Lists (each with More / Less / Default options)

Changes are applied instantly to all chats, including those already open. Works on all platforms: web, iOS, Android (update the app to the latest version if you don't see it). ⚙️

Is this available on a free account?

Yes, all presets and Characteristics sliders are available even on the free tier. You can choose any preset (including Quirky, Candid, Efficient, etc.) and adjust the Warmth/Enthusiasm/Emoji/Headers sliders. The only difference is the message limit on GPT-5.1 (usually 30–50 per 3 hours, depending on the load). On Plus/Team/Enterprise, the limit is significantly higher + priority speed + access to additional features (like extended Memories). If you frequently work with AI, Plus pays for itself thanks to the comfort it provides. 💸

Can I create my own "personality" from scratch?

As of now (January 2026), there isn't a separate "Create your own preset" button. But there are two powerful ways to get close to it:

1. Combination of preset + sliders + Custom instructions — you choose a base preset (e.g., Candid), adjust the sliders (Less Warmth + More Headers), and in Custom instructions, add detailed rules: "always respond as a CTO with 15 years of experience, use DevOps terms, no emojis, with risk priorities." This works very well — the model remembers instructions globally.
2. Custom GPTs (in Plus/Team/Enterprise) — you create your own GPT from scratch: set a name, description, instructions, upload files. There you can define the character in more detail than in global personalization (e.g., "you are a sarcastic British editor from the 90s"). Custom GPTs do not affect the main ChatGPT but provide almost unlimited customization. 📝

OpenAI promises expansion in 2026 — possibly full-fledged "User-defined personalities."

Does changing "Personality" affect safety, censorship, or content rules?

No — in no way. All basic OpenAI safety rules (Moderation API), prohibitions on harmful content, NSFW censorship, policies regarding disinformation, violence, etc., remain unchanged regardless of the preset or sliders. You can make the AI as cynical or humorous as possible, but if a request violates the ToS, the model will refuse or issue a warning. The tone changes, but the "red lines" do not. This also applies to jailbreak attempts: even in Cynical mode, the AI will not assist with illegal activities. 🛡️

Is the selected "Personality" retained when changing the model (e.g., from GPT-5.1 to another)?

Yes, Personalization settings are global and are retained regardless of the selected model in the chat. If you switch to GPT-4o or another version (if available), the preset and sliders will remain. But remember: different models may interpret tone differently (GPT-5.1 adheres better to instructions than older ones). If something "shifted," just switch the preset again.

Can I quickly switch modes during a chat?

Yes, with one click: in any chat, click on the profile icon or menu → Personalization → change the preset/sliders. The change is applied immediately, even to the current message (the model will regenerate the response in the new style if you click Regenerate). Very convenient for A/B tests: one query — different tones. ⚡

Conclusion

ChatGPT's new "personality" isn't about the AI becoming smarter or faster. It's about it finally speaking your language: with the warm voice of a friend when you're tired, with harsh truth when an audit is needed, with a creative burst when your brain is stuck. This is a revolution not in the model's brain, but in the communication interface — and it's already here, available to everyone.

Choose a preset and sliders for your tasks — and you'll get not just a chatbot, but a true ally: a patient teacher, a ruthless critic, an inspired creative partner, or a dry professional. Experiment today — 5–10 minutes in the settings can save you hours every week. In 2026, this will become a basic skill for everyone who seriously works with AI.

Are you ready to make ChatGPT truly "yours"? Then read no further — open the settings right now. The difference will surprise you. 🌟🔥

Recommendations: What else to read about ChatGPT and AI in 2026

If you want to delve deeper into OpenAI trends and practical tools, here's a selection of the most relevant materials. Each review includes specific examples, figures, and tips on how to apply them now.

• AI in 2025: From Chatbots to Autonomous Agents — What Has Really Changed and Why It's Important Now — a complete overview of the evolution from simple chats to agents that perform tasks themselves.
• How AI Platforms Choose Sources for Answers in 2025–2026 — an analysis of search and ranking algorithms and why some sites disappear from ChatGPT's answers.
• Advertising Already in ChatGPT's Code: When It Will Appear in Your Chat in 2026 — what is known about OpenAI's future monetization features and how they will affect users.
• ChatGPT Search: Overview — How It Works and Best Tips — a practical guide to the new search in ChatGPT, query examples, and how to bypass limitations.
• ChatGPT for Adults: What is Known About the New Mode — details about NSFW mode, censorship, and how it works in practice in 2026.
• 50 ChatGPT Prompts: Ready-Made Templates for Content and Code — a collection of proven prompts that actually work for texts, SEO, programming, and creativity.

I recommend starting with the first two — they provide the broadest context for our topic of personalization.