Chatbot Dialogue Examples for the Hard Moments: 12 Annotated Exchanges With Angry, Confused, Off-Topic and Silent Visitors (Free Reply Drill)

Chatbot dialogue examples usually show the easy chats: a visitor asks about delivery, the bot answers, everyone is happy. The chats that decide whether customers trust your bot are the other ones: the angry repeat question, the one-word message, the insult, the question your bot can’t answer, the visitor who goes quiet. This guide gives you 12 short dialogues for those moments, each with a weak version, a better version and a note on what every bot turn is doing. A free drill lets you write your own reply for eight of them and checks it.

πŸ‘€ What the visitor seesHard moment, handled
This is the THIRD time I’m asking. Where is my refund?? Useless.
πŸ€– Shop assistantSorry, three times is too many to ask. I’m passing this to our team now so a person can check your refund. What’s your order number?
Chat handed to your team

Acknowledge, hand over, ask one thing.

⚑ In short

  • Hard moments are rare per chat but common per month. Plan the words for them before launch.
  • Most strong replies use the same four moves: acknowledge, answer, ask one thing, hand over.
  • Never guess, never pretend to be human, never lecture an upset visitor.
  • Write each reply once, save it, and test it with the drill below.
Chatbot dialogue examples in LiveChatBot: AI Agent knowledge sources and a sample chat answered from them
How it looks in LiveChatBot: AI Agent knowledge sources and a sample chat answered from them.

Why these moments

Happy-path dialogues don’t prepare your bot

Look at any vendor’s demo and the conversation goes perfectly. The visitor types a tidy question, the bot gives a tidy answer, the visitor says thanks. Real chat logs look different. A small share of visitors arrive annoyed, type one word, ask two things at once, wander off topic, or vanish halfway through. A bot that only knows the happy path handles those chats badly, and those are the chats people remember and talk about.

There’s a simple reason to plan for them. If your website gets, say, 400 chats a month (an illustrative figure), and only 5% of them hit a hard moment, that’s 20 tricky conversations every month. Each one is a customer deciding whether your business is easy to deal with.

The good news is that hard moments repeat. The angry “third time asking” chat looks much the same at a clothing shop, a clinic and a coaching institute. So does the one-word “price” message, and the “are you a bot?” question. You can write the reply once, test it, save it, and reuse it for years.

The chatbot dialogue examples below are short on purpose. Each one shows the visitor’s message, a weak reply that many bots really send, and a stronger reply. Under each one is a line on what the bot turn is doing. All business names and figures are illustrative.

If you want to plan who speaks when across a whole chat, our guide to turn-taking in bot dialogue covers that. This guide zooms in on the single reply that rescues or loses the conversation.

The four moves

What every strong reply is made of

Read the examples with these four labels in mind. Most good replies use two or three of them, in this order.

πŸ’¬

Acknowledge

Name what just happened: they asked twice, the bot got it wrong, they want a person. One short clause, not a paragraph of apology.

βœ…

Answer

Give the fact they need, key detail first. Only if it’s in your bot’s knowledge. No fact, no answer.

❓

Ask one thing

If you need something to move forward, ask for exactly one item: an order number, a product name, a city.

πŸ™‹

Hand over

Pass to a person when the bot has failed, the visitor asks, or the issue is about one specific order, payment or complaint.

Examples 1–3

Chatbot dialogue examples for frustrated visitors

1. The repeat question

Visitor: This is the THIRD time I’m asking. Where is my refund?? Useless.

Weak bot: As per our refund policy, refunds are processed within 7 working days. You should have read the policy page.

Better bot: Sorry, three times is too many to ask. I’m passing this to our team now so a person can check your refund. What’s your order number?

*What it does:* acknowledge, hand over, ask one thing. The weak reply quotes policy and blames the customer. By the third ask, the bot has failed; a person must take it.

2. The insult

Visitor: you’re a stupid useless bot

Weak bot: Please mind your language. Abusive messages are not tolerated.

Better bot: Sorry I haven’t been helpful. Let me get a person from our team to take over. What were you trying to sort out?

*What it does:* the visitor is frustrated, not dangerous. Owning the failure calms most people down. Keep the block button for real abuse; LiveChatBot lets your team block a visitor when it’s truly needed.

3. “Are you a bot?”

Visitor: Am I talking to a real person or a bot?

Weak bot: I’m here to help! What can I do for you today?

Better bot: I’m the shop’s automated assistant, not a person. I can answer delivery, returns and size questions, or pass you to our team. Which would you like?

*What it does:* answers honestly, says what it can do, offers a human. A dodge feels like a trick, and pretending to be a person is a line no business bot should cross. Our handover guide has more wording for the pass to your team.

Examples 4–6

Messages that don’t make sense yet

These visitors aren’t upset. Their message just doesn’t give the bot enough, or gives it too much.

πŸ‘€ What the visitor seesNarrow it down
price
πŸ€– Shop assistantHappy to help with prices. Which one are you looking at: kurtas, sarees or kids’ wear?
Kurtas
Sarees
Kids’ wear

Weak version: β€œOur prices are very competitive.” Better: one question with three likely options.

πŸ‘€ What the visitor seesAnswer both halves
do you deliver to kochi and is cod available
πŸ€– Shop assistantYes to both. We deliver to Kochi in 4 to 6 working days, and cash on delivery is available on orders up to β‚Ή5,000.

Weak version: β€œYes, we deliver to Kochi!” and silence on COD. Figures are illustrative.

πŸ‘€ What the visitor seesDecline and redirect
can you write my college essay on climate change?
πŸ€– Bookshop assistantThat’s not something I can help with, sorry. I’m the assistant for this bookshop, so I can help with orders, stock and delivery. Looking for a book on climate change?

Weak version: starts writing the essay. A business bot that does random tasks may say things you never approved.

Free tool

Write your reply, then check it

Pick one of eight hard moments, type the reply your bot would send, and press Check. The drill scores it on the moves that matter for that moment and shows a model reply. Load the weak reply first to see what it catches.

1. Pick a hard moment

2. The visitor says

3. Type the reply your bot would send

Runs only in your browser; nothing is sent or saved. The checks look for simple word patterns, so treat the score as a prompt to reread your reply, not a verdict. Figures in the model replies are illustrative.

Examples 7–9

When the bot doesn’t have the answer

7. The knowledge gap

Visitor: is there parking near your Andheri shop?

Weak bot: Yes, there is plenty of free parking available.

Better bot: I don’t have parking details, sorry. I can ask our Andheri team to check and reply here, or you can call the shop. Which works better?

*What it does:* admits the gap, offers two ways forward. The weak reply is the most dangerous kind: a confident guess. If the customer drives there and finds no parking, the bot has made a promise for you. An AI bot should only answer from the knowledge you give it; when the fact isn’t there, “I don’t know” is the right answer. See how to train your AI chatbot for filling those gaps.

8. The quiet visitor

*The bot asked for an order number. Three minutes pass.*

Weak bot: Hello?? Are you still there? Please reply or this chat will be closed.

Better bot: No rush. When you have your order number handy, type it here and I’ll pick up where we left off.

*What it does:* people go quiet because they’re searching their email for the order number, not because they’ve left. One calm nudge, then wait. LiveChatBot’s Flows include a chat rescuer that can send a follow-up message when a visitor goes quiet; write it in this tone. Our guide to chat silences covers the human side.

9. After hours, wants a person

Visitor (11:40 pm): I need to talk to someone about my booking

Weak bot: An agent will be with you shortly.

Better bot: Our team is offline now and back at 10 am. Leave your phone number or email and we’ll reply first thing tomorrow.

*What it does:* tells the truth about time, collects a contact. “Shortly” at midnight is a broken promise. In LiveChatBot, set your opening hours and the away form collects the email or phone number for you; offline messages can become tickets so nothing is lost.

Examples 10–12

Turns in the middle of a conversation

10. The topic change

Bot: To start a return, which item would you like to send back?

Visitor: wait, how long does delivery take for the replacement?

Weak bot: Please select the item you want to return.

Better bot: Replacements arrive in 3 to 5 working days after we receive the return. Shall we carry on with the return? Which item is it?

*What it does:* answers the side question, then returns to the task. A bot that repeats its last prompt and ignores the new question feels like a wall.

11. The correction

Bot: The black kurta in size M is in stock.

Visitor: no, I meant the blue one

Weak bot: The black kurta in size M is in stock.

Better bot: Got it, the blue one. The blue kurta in size M is in stock too, and there are two left in L.

*What it does:* accepts the correction in two words and answers the new item. Never argue with a correction, and never repeat the old answer.

12. Another language

Visitor: size M available hai kya?

Weak bot: I’m sorry, I didn’t understand your question.

Better bot: Haan, size M available hai. Kaunsa colour chahiye: blue ya black?

*What it does:* mixed Hindi and English is normal in Indian chats. Reply in the visitor’s style if your bot can; if not, reply in simple English and answer the question anyway. LiveChatBot detects the visitor’s language and can translate your welcome text; see multilingual live chat for more.

These three show why a bot reply can’t be judged on its own. Read the turn before it. A strong reply always fits the conversation so far, not just the latest message.

Quick reference

Weak lines and what to write instead

Upset visitor
🚩 Instead ofβ€œAs per our policy…”
βœ… Writeβ€œSorry, that’s taken too long. I’m passing this to our team now.”
Bot doesn’t know
🚩 Instead ofA confident guess
βœ… Writeβ€œI don’t have that detail. I can ask our team to check.”
After hours
🚩 Instead ofβ€œAn agent will be with you shortly.”
βœ… Writeβ€œWe’re back at 10 am. Leave your email and we’ll reply first thing.”
Quiet visitor
🚩 Instead ofβ€œHello?? Are you still there?”
βœ… Writeβ€œNo rush. Type it here when you’re ready.”
Asked if human
🚩 Instead ofβ€œI’m here to help!”
βœ… Writeβ€œI’m an automated assistant. Want a person?”

Put them to work

Turn these chatbot dialogue examples into your own replies

Reading examples helps; saving your own versions is what changes your chats. Here’s a one-hour routine.

  1. Pick your six most likely hard moments. For most small businesses: repeat question, bot doesn’t know, after hours, one-word message, two questions, and “are you a bot?”.
  2. Rewrite each better reply in your own words. Use your real hours, your real product names, your real delivery times. Keep each under about 40 words.
  3. Run each through the drill above. Fix anything marked with a cross.
  4. Save them where your bot and team can use them. In LiveChatBot, put the general ones into the AI Agent’s knowledge (for example, “If you don’t know a fact, say so and offer our team”) and save the exact wording as saved replies, so the free keyword bot and your team use the same words. Saved replies support {name}, so a person can personalise them in one click.
  5. Set the handover pieces. Opening hours with the away form, the chat rescuer follow-up in Flows, and tickets for anything that needs a follow-up.
  6. Review once a week. Read the chats where a hard moment happened, and see whether the reply worked. Adjust the words, not the whole bot.

Keep a short list of rules in your head while you write: one fact first, one question at most, a person when the bot has failed, and never a guess. For a ready bank of shorter lines for each stage of a chat, see our bot message examples.

Checklist

Hard-moment readiness

Hard moments covered

Getting started

Fix your weakest reply today

Open your bot or your chat history and find the reply it sends when it doesn’t know something. That single message decides more chats than any other. Rewrite it with the drill, save it, and test it with a question your bot can’t answer.

LiveChatBot’s free plan includes the AI Agent with a small free monthly allowance, a free keyword bot that answers from saved replies, Flows with a chat rescuer, opening hours with an away form, tickets, the team inbox and the Android app, for two websites with 100 visitors per website each month. See pricing and start at livechatbot.in.

Frequently asked questions

Are these chatbot dialogue examples for AI bots or rule-based bots?

Both. An AI bot can be told to follow these patterns in its knowledge and instructions; a rule-based or keyword bot needs each reply saved and matched to trigger words. The words themselves are the same.

Should my bot apologise in every reply?

No. Apologise when something has gone wrong: a repeat question, a wrong answer, a long wait. A bot that says sorry in every message sounds nervous and wastes space.

When should the bot hand over to a person?

When the visitor asks for one, when they repeat a question, when they’re upset, and when the issue is about one specific order, payment or complaint. Outside opening hours, collect a contact instead.

Is it okay for a bot to say it doesn’t know?

Yes. It’s far better than a confident wrong answer. Pair it with a next step, such as asking the team or leaving a contact.

Does the drill send my reply anywhere?

No. It runs in your browser and nothing is sent or saved. It looks for simple word patterns, so use it as a prompt to reread your reply.

Keep reading

Related guides

πŸ“˜

Chatbot Dialogue Turn-Taking: Who Talks, Who Asks and How Many Turns to an Answer (Free Turn Analyser)

Read the guide β†’

🧭

Chatbot Development, Stage by Stage: How Long Each Step Takes and Where the Hours Really Go (Free Effort Estimator)

Read the guide β†’

πŸ› οΈ

Chatbot Design Blueprint: Sketch Your Bot’s Menu Tree on Paper and Catch Problems Before You Build

Read the guide β†’

Plan the hard moments before they happen

Save your best replies, set the handover and let the AI Agent answer the rest. Start free on livechatbot.in.