UX Design for Chatbots: Usable, Natural, and Human-Centric

In an era when Artificial Intelligence (AI) is becoming an integral part of User Experiences (UX) on digital platforms, chatbots have emerged as a key tool that many organizations use to provide information, answer questions, and even conduct preliminary health assessments. A core principle in chatbot design is usability — the system’s ability to be effectively and efficiently used.

This article presents UX design concepts for chatbots, emphasizing usability factors and suggesting the use of sentiment detection technology to make interactions feel more human. It also references international frameworks like ISO 9241–11, Nielsen’s Heuristics, PARADISE, and SASSI, which can be applied to evaluate and design conversational systems.

Key Usability Factors to Consider

Designing UX for chatbots requires more than just applying traditional UI principles, since Conversational User Interfaces (CUI) have unique characteristics, especially as users expect interactions to feel like conversations with real people.

To design effective chatbots, usability should be assessed using internationally recognized standards, such as:

• ISO 9241–11 (1998): Focuses on Effectiveness, Efficiency, and Satisfaction
• Nielsen’s Heuristics: Emphasizes Learnability, Error Prevention, and Satisfaction
• PARADISE Framework (Walker et al., 1997): For dialog systems — Naturalness, Clarity, Willingness to Use
• SASSI (Hone & Graham, 2000): For voice/chat interfaces — Likability, Humanness, Cognitive Demand

From these frameworks, five core usability factors relevant to chatbot design can be summarized:

1. Effectiveness — Users can accurately obtain the information or complete the task they want.
2. Learnability — The system is easy to use without requiring much learning.
3. Reliability — Users trust the answers provided.
4. Humanness — Communication feels natural, not mechanical.
5. Likability — Users have a positive emotional experience.

Industry Challenge: Chatbots That Don’t Feel Human

In many sectors — particularly in finance and customer service — chatbots are frequently reported to lack human-like communication, which results in user dissatisfaction, even when the responses are technically correct.

“As conversational AI becomes more advanced, designing for trust — including emotional sensitivity, transparency, and fairness — is crucial to create meaningful user experiences.”
— Accenture, Building Trust into Conversational AI Solutions

This trend aligns with research by Brave & Nass (2003), which found that chatbot personality and emotional expression significantly impact perceived reliability, intimacy, and likability.

As a response, many organizations now adopt a sentiment-aware design approach by integrating sentiment detection systems, enabling the chatbot to adjust its tone based on the user’s emotions.

Example: User types: “How long do I have to wait!?”
→ Chatbot replies with empathy and politeness: “I sincerely apologize. I understand your frustration — I’ll check on this for you right away.”

This approach draws on the SASSI Framework, which stresses the importance of Likability and Humanness in influencing Willingness to Use and Overall Satisfaction.

Experiment Results: Emotionally Intelligent Chatbots Feel More Human

Our UX/UI design team was tasked by a financial organization to design and enhance a chatbot used as the first point of contact for customer inquiries.

Using A/B testing, users interacted with two chatbot versions — one with sentiment-aware capabilities and one without. During the sessions, users were asked to “Think-Aloud (A method where users verbalize their thoughts while working to reveal behaviors and expectations)” and afterward complete a Usability Questionnaire.

Findings:

• Users interacting with the sentiment-aware chatbot gave higher satisfaction ratings (4–5 stars).
• Users felt the chatbot “understood their emotions” and resembled a “real assistant” rather than an automated system.
• Positive feedback highlighted the importance of a polite tone, empathy, concise responses, and emotional expression.

Factors that improved usability scores included:

• Reducing overly formal language in certain contexts
• Using emojis and tone shifts to match the user’s mood
• Offering helpful, personalized replies instead of generic system templates

Example user insights:

• “It felt like the chatbot acknowledged my frustration and took responsibility.”
• “I didn’t feel like I was being ignored — like someone was actually listening.”

Even though sentiment-based responses didn’t drastically improve satisfaction in every case, they had a notable positive impact on users in negative emotional states, making them critical moments for maintaining brand-user relationships.

LLMs and the New Role of Chatbots: UX Opportunities and Challenges

In recent years, Large Language Model (LLM)-based chatbots like GPT, Claude, or Gemini have gained rapid popularity, especially in large organizations that require flexible, multi-purpose conversational systems.

However, while LLMs greatly enhance chatbot capabilities, they also introduce new UX concerns designers should consider:

• Latency (Response Delays): LLMs process responses in real-time, which can be slower than rule-based systems. Use feedback elements like “Typing…” indicators to manage user expectations.
• Uncertainty (AI Hallucination): Some users expect flawless answers. Include disclaimers such as “This is an AI-generated response and may not be 100% accurate” or offer reference links.
• Tone & Personality Consistency: LLMs may vary in tone. Use UX layers like prompt tuning, style guides, or sentiment-aware rewriters to maintain brand-aligned personas.
• Context Coherence: While LLMs can remember conversations, managing memory (e.g., showing users what the chatbot “remembers” or allowing them to reset context) enhances transparency.

In sectors like finance, combining LLMs with sentiment-aware design is key. Since LLMs are not inherently emotional-aware, adding a sentiment detection layer before LLM interaction can help tailor responses more appropriately.

Recommendations for Chatbot Designers

• Listen more than respond: Avoid replying in a one-size-fits-all formal tone. Adapt responses based on user emotion.
• Design multiple response variations: Prevent robotic patterns and make the bot feel more human.
• Use emojis contextually: In informal or friendly settings, emojis can help express the bot’s “emotions.”
• Set tone & politeness levels: Adjust language based on question type and user mood — opt for a more polite and less formal tone in negative or sensitive situations.
• Let the bot acknowledge its limitations: e.g., “Sorry, I may not be able to answer that question at the moment, but I can forward it to a staff member right away.” This helps build trust.
• Regularly test with real users: Gather feedback on emotional experience — something task success rates alone can’t measure.

Conclusion

Designing UX for chatbots requires balancing accuracy and effectiveness with the emotional experience of the user. A good system isn’t just a quick-response assistant — it should reflect a brand that “understands and cares.”

Applying international usability frameworks like ISO 9241–11, PARADISE, and SASSI in real-world contexts helps transform chatbots into valuable user touchpoints and builds long-term trust between users and organizations.