聊天机器人消息设计 - 最佳响应长度与用户体验
Chatbot response length directly affects user experience. Messages that are too long overwhelm users, while overly brief responses feel unhelpful. This article covers platform-specific limits and design principles for effective chatbot communication. For foundational knowledge, books on chatbot UX design offer practical frameworks.
Beyond simple platform limits, this article dives into the cognitive science behind optimal message bubble length, the correlation between turn count and user drop-off rates, information density differences between Japanese and English in multilingual chatbots, and practical design guidelines for error messages and fallback responses.
Platform Message Limits
| Platform | Max Message Length | Recommended Length |
|---|---|---|
| Facebook Messenger | 2,000 characters | 100–300 characters |
| Slack Bot | 40,000 characters | 200–500 characters |
| LINE Bot | 5,000 characters | 100–300 characters |
| WhatsApp Business | 4,096 characters | 100–300 characters |
| Web Chat Widget | Varies | 150–400 characters |
Design Principles
- One message, one idea: Break complex responses into multiple short messages rather than one long block
- Front-load the answer: Put the most important information first, then add details
- Use quick replies and buttons: Reduce the need for users to type by offering selectable options
- Keep responses under 300 characters: Studies suggest chatbot messages over 300 characters see significantly lower engagement
- Minimize steps to resolution: Design conversation flows so users reach their goal within 3 steps. Too many branches cause users to feel lost and abandon the conversation
Turn Count and User Drop-Off Rates
An often-overlooked aspect of chatbot design is the relationship between turn count (the number of back-and-forth exchanges between user and bot) and drop-off rates. Analysis of customer support chatbot operational data reveals the following trends:
| Turn Count | Estimated Continuation Rate | Primary Drop-Off Cause |
|---|---|---|
| 1–3 turns | 85–95% | Minimal drop-off (goal often achieved) |
| 4–6 turns | 60–80% | "Runaround" feeling causes abandonment |
| 7–10 turns | 30–50% | User concludes bot cannot resolve issue |
| 11+ turns | Under 20% | Frustration-driven abandonment |
The critical finding is that drop-off rates spike sharply at the 4th turn. Users tend to conclude that "if this bot can't solve my problem in 3 exchanges, it never will." This pattern leads to the following design guidelines:
- Complete FAQ-type inquiries within 3 turns
- Offer a human escalation option by the 4th turn
- For complex procedures (address changes, returns, etc.), show progress indicators like "Step 2 of 4"
- When turn count increases, keep each message shorter (under 150 characters) to maintain conversational tempo
Optimal Response Length and Cognitive Load
Beyond technical limits, there is an optimal response length from a user experience perspective. Research suggests that chatbot responses of 100–300 characters are the most readable and engaging.
This "100–300 character" guideline has a basis in cognitive science. Human working memory can hold approximately 4±1 chunks (meaningful units) at a time. In English, one chunk corresponds to roughly 5–10 words, meaning a single message bubble can comfortably convey about 20–60 words (100–300 characters). Beyond 300 characters, users need to re-read portions, and cognitive load increases sharply.
Here is a breakdown of message length ranges and their characteristics:
| Character Range | Cognitive Load | User Reaction | Best Use |
|---|---|---|---|
| Under 50 | Very low | Feels insufficient, lacks trust | Confirmations ("Yes", "Got it") |
| 50–150 | Low | Quick comprehension, immediate action | Simple answers, option prompts |
| 150–300 | Optimal | Comfortable reading, good retention | Explanatory answers, single steps |
| 300–500 | Somewhat high | Skimming begins | Detailed explanations (split recommended) |
| Over 500 | High | Scrolling required, drop-off risk | Avoid (split into multiple bubbles) |
Responses that are too short (under 50 characters) feel insufficient, while overly long responses (over 500 characters) require scrolling and are often skimmed or ignored. When conveying complex information, splitting the content across multiple messages delivered in sequence is more effective than cramming everything into a single bubble.
Platform-Specific Considerations
Each platform renders message bubbles at different widths, so the same character count can look very different across platforms. Here is a comparison of message bubble display characteristics:
| Platform | Max Bubble Width (approx.) | English Words per Line | Line-Wrap Behavior |
|---|---|---|---|
| LINE | ~70% of screen width | 5–8 words | Character-level wrapping |
| Facebook Messenger | ~65% of screen width | 5–7 words | Word-level (English), character-level (CJK) |
| Slack | ~80% of channel width | 12–20 words | Word-level, code block support |
| ~65% of screen width | 5–7 words | Character-level wrapping | |
| Web Chat (typical) | 300–400px | 6–10 words | Implementation-dependent |
As this table shows, a 300-character message on LINE takes up 10+ lines and nearly fills a smartphone screen, while the same message on Slack's desktop client fits in just 4–5 lines. The same character count can feel "long" or "short" depending on the platform.
LINE bots can leverage rich messages and carousel templates to convey information visually, but each template type has its own character limits that must be checked in advance. Carousel titles are limited to 40 characters and descriptions to 60 characters, requiring concise writing skills to convey key points within these constraints.
Facebook Messenger enforces a 2,000-character limit per message, with quick reply labels capped at 20 characters and button titles also limited to 20 characters. In English, 20 characters equals roughly 3–4 words, which significantly limits expression compared to Japanese where 20 characters can convey a complete thought.
Slack bots enjoy a generous 40,000-character limit, but business chat contexts demand brevity. Use Block Kit for structured, rich message layouts that organize information clearly. A Slack-specific consideration is that markdown syntax (*bold*, `code`) works within messages, so character counts include markdown symbols.
WhatsApp Business API distinguishes between template messages (pre-approved) and session messages. Template messages have separate limits for header (60 characters), body (1,024 characters), and footer (60 characters). Since approval takes 24–48 hours, design templates with margin and avoid frequent revisions.
Web chat widgets often have no hard character limit, but the chat window's display width must be considered. Typical widgets are 300–400px wide, so aim for natural line breaks at 40–60 characters per line, and split long responses into separate messages. Since widgets expand to full screen width on mobile, always test message appearance on both desktop and mobile.
Multilingual Chatbot Character Count Design
In multilingual chatbots, the same content can vary dramatically in character count across languages. Japanese and English have an information density difference of roughly 1.5–2x, and ignoring this difference leads to degraded UX.
Here is a comparison of the same messages in Japanese and English:
| Message Content | Japanese | English | Character Ratio |
|---|---|---|---|
| Greeting + feature intro | "サポートボットです。ご質問をどうぞ。" (18 chars) | "I'm a support bot. How can I help you?" (39 chars) | 1:2.2 |
| Error notification | "入力内容を確認してください。" (14 chars) | "Please check your input and try again." (39 chars) | 1:2.8 |
| Option prompt | "以下から選んでください。" (12 chars) | "Please select one of the following options." (44 chars) | 1:3.7 |
This information density gap directly affects message bubble appearance. A 100-character Japanese message often becomes 200–280 characters when translated to English, causing the bubble height to double or triple. When designing multilingual chatbots, consider the following:
- Set language-specific optimal character targets (Japanese: 100–200 chars, English: 150–300 chars)
- Design bubble widths based on the language that produces the longest messages
- If option labels are 10 characters in Japanese, budget 20 characters for English
- Chinese (Traditional) has similar information density to Japanese, while Korean tends to require about 1.2x the character count of Japanese
Error Messages and Fallback Response Design
Chatbot quality is determined not just by normal responses, but by how the bot handles errors and unrecognized user intent (fallback responses). Fallback responses should be designed in escalating tiers:
| Level | Trigger Condition | Response Pattern | Character Target |
|---|---|---|---|
| Level 1: Mild | Low confidence on first attempt | "Did you mean [topic]?" — confirm intent | 50–100 chars |
| Level 2: Moderate | 2 consecutive unrecognized inputs | Present options to narrow down intent | 100–200 chars |
| Level 3: Severe | 3 consecutive unrecognized inputs | Offer human agent handoff | 150–250 chars |
A common mistake is returning the same "Sorry about that" message for every error. Users want to know "what went wrong" and "what to do next" — an apology alone does not enable action. Every error message should include two elements: a hint about the cause and a suggested next action.
Here is a concrete improvement example:
- Before: "Sorry, please try again." (24 characters)
- After: "I couldn't quite understand your question. Please select a category below or try rephrasing your request." (105 characters)
The improved message is longer, but by giving users a concrete action path, it significantly improves conversation continuation rates.
Optimizing Response Speed and Typing Indicators
Research on chatbot response speed has found that users actually prefer a brief 1–3 second "typing" indicator before the bot replies, rather than an instant response. Mimicking the natural rhythm of human conversation makes the interaction feel more authentic and increases overall satisfaction.
However, the optimal delay should scale with message length. A short 50-character response warrants only a 1-second delay, while a 300-character response feels natural with a 2–3 second pause. A long message that appears instantly feels like a canned response, while a short message with a 3-second delay feels sluggish. Here are recommended delay guidelines by message length:
| Response Length | Recommended Delay | Rationale |
|---|---|---|
| Under 100 chars | 0.5–1 second | Long waits feel unnatural for short replies |
| 100–250 chars | 1–2 seconds | Mimics natural human typing rhythm |
| 250–500 chars | 2–3 seconds | Creates impression of "thinking" |
| Over 500 chars | 2–3 sec + split delivery | Excessive wait times are counterproductive |
Common Design Mistakes
- Too many options in the welcome message: Presenting 6 or more choices triggers the "paradox of choice," increasing the likelihood that users select nothing and abandon the conversation. Limit options to 3–4. If you need 5 or more, add an "Other" option and use a two-tier structure to prevent drop-off
- Generic error responses: Replying with only "Sorry, please try again" leaves users unsure what went wrong. They repeat the same action and grow frustrated. Always suggest a specific alternative or next step
- Cramming multiple questions into one bubble: Asking "Please provide your name, phone number, and preferred date" requests three pieces of information at once, leaving users unsure where to start. Follow the one-message-one-question rule and collect information step by step
- Context-blind resets: When a user says "I want to go back to the first option" on the 5th turn, forcing them to restart from scratch is a major UX degradation. Implement a "back" function that returns to the previous decision point
Professional Design Techniques
- Insert confirmation questions: Adding "I'll look into [topic] for you. Does that sound right?" before executing a task reduces misunderstandings and user dissatisfaction significantly. Keep confirmation questions under 100 characters and make them answerable with "yes/no"
- Suggest concrete next actions at conversation end: Closing with "Here are related FAQs" or "Contact a human agent here" gives users a clear path forward and improves satisfaction scores
- Equalize option label lengths: Keeping option labels within a similar character range (±5 characters), such as "About products," "About shipping," "About returns," creates a visually organized impression and makes selection easier
- Visualize conversation progress: Showing "Step 2 of 4: Confirm delivery address" lets users predict remaining steps, reducing mid-process abandonment
Conclusion
Effective chatbot messages are concise, actionable, and platform-appropriate. Aim for 100–300 characters per message, and from a cognitive load perspective, keep content within working memory capacity (4±1 chunks). Follow the core principles of front-loading answers, one message per topic, and offering clear options so users can access information without friction. For deeper exploration of these principles, conversational AI design resources are worth consulting.
A particularly overlooked factor is the relationship between turn count and drop-off rates. Drop-off spikes sharply at the 4th turn, so FAQ-type inquiries should be resolved within 3 turns, with human escalation available beyond that point. For multilingual bots, account for the information density gap between languages (roughly 1.5–2x between Japanese and English). Use Character Counter to verify your message templates stay within optimal ranges.