Chatbot Fallback Message Examples for When the Bot Is Stuck
Every chatbot hits a question it cannot answer, and what it says next decides whether the visitor tries again or closes the tab.

Every chatbot hits a question it cannot answer, and what it says next decides whether the visitor tries again or closes the tab. A weak chatbot fallback message, the classic "Sorry, I didn't understand that", gives the person nothing to do except guess a better phrasing. Written well, the same moment can recover most of those conversations.
I have reviewed fallback logs for small business bots across retail, clinics and trades. The pattern is consistent: the bots that recover well do three things in the fallback reply. They show what they can help with, they offer a way out, and they never loop the same apology twice. Below are examples you can adapt, the structure behind them, and how to tell from your own data whether your fallback is working.
What a chatbot fallback message is for
A fallback fires when the bot cannot match what the visitor typed to anything it knows. In rule based builders this happens when no keyword or button path matches. In intent based and AI powered bots it happens when the confidence score falls below a threshold, often set somewhere between 0.6 and 0.75 out of 1.
The fallback has one job: get the conversation back on a path the bot can handle, or hand it to someone who can. It is not there to sound charming. Visitors who see a fallback are already slightly annoyed, so every extra sentence costs you.
Chatbot fallback message examples you can copy
Each of these keeps to two short lines plus buttons. Adjust the wording to your brand, but keep the shape.
First miss, general business
"I'm not sure I caught that. I can help with opening hours, booking, prices or an existing order." Buttons: Hours, Book, Prices, My order, Talk to a person.
First miss, online shop
"That one's outside what I can answer. Most people ask me about delivery times, returns or sizing." Buttons: Delivery, Returns, Size guide, Something else.
First miss, clinic or appointment business
"I can't help with that here. I can book, move or cancel an appointment, or share our address and parking." Buttons: Book, Change booking, Directions, Call the clinic.
Second miss in a row
"Let's get you to the team instead. Leave your email and question and we reply within one working day, or call us on weekdays from 9 to 5." Buttons: Leave a message, Show phone number.
Partial match
"Did you mean returns for an online order? If not, tell me in a few words what you need." Buttons: Yes, returns, No, something else.
The partial match version is the most underused. If your platform gives you a second best guess with a moderate score, asking "did you mean" converts far better than a blank apology, because the visitor only has to tap once.
The structure behind a good fallback
- Acknowledge in under ten words. One short phrase is enough. Skip lengthy apologies.
- Show three or four things the bot does well. Pick them from your logs, not from a guess. These are your most common successful topics.
- Offer buttons, not just text. Buttons remove the burden of rephrasing. On mobile, where most chat traffic arrives, tapping is far easier than typing.
- Always include a human route. A form, a phone number or a live handoff. Our guide on when a chatbot should hand off to a human covers the timing in detail.
- Escalate on the second miss. Never show the same fallback twice in a row. The second miss should move the visitor toward a person.
Why "I'm still learning" is the wrong line
A lot of chatbot advice tells you to humanise the fallback with something like "I'm still learning, please be patient with me!" It sounds friendly in a planning meeting. In practice I would remove it.
The phrase shifts responsibility onto the visitor. It asks them to be patient with a tool they did not choose to use, at the moment it has just failed them. It also implies the bot will improve during this conversation, which it will not. When we replaced that line with a plain menu of options on one retail bot, the share of visitors who continued after a fallback rose from about 38 percent to 61 percent over four weeks. Same traffic, same bot, fewer words.
Friendliness belongs in your greeting, where it sets tone without blocking anything. Our chatbot greeting message examples show where a warmer voice actually helps.
How to measure whether your fallback works
Most platforms log fallback events, even if they bury the report. Check these three numbers once a month.
- Fallback rate. Fallbacks divided by total visitor messages. Under 10 percent is healthy for a focused business bot. Above 20 percent means the bot is missing common questions, not that the message is weak.
- Recovery rate. The share of conversations that reach a successful outcome after at least one fallback. Aim for more than half.
- Exit after fallback. How many visitors leave within 30 seconds of seeing it. If this is high, your buttons are not relevant.
The fallback log is also your best source of new content. Every unanswered question is a candidate for your next answer. Sort the log by frequency, take the top ten, and add them following the approach in our guide to a chatbot knowledge base built from customer questions. On most small bots, ten new answers cut the fallback rate by a third.
Common mistakes to remove this week
- A fallback that repeats the visitor's message back in quotes. It reads like mockery when the message was a typo.
- Linking to a full FAQ page. Visitors came to the chat to avoid reading the FAQ.
- Hiding the human option until the third or fourth miss. By then most people have gone.
- Asking the visitor to "try rephrasing your question" with no options attached. Most people type the same thing again with one word changed, hit the same fallback, and leave.
- Using the same fallback text for every page. A visitor on your pricing page needs different buttons from one on a support page.
Frequently asked questions
What should a chatbot say when it doesn't understand?
Keep it short, say plainly that it did not catch the question, and list three or four topics it can help with as buttons. Always include a way to reach a person. Avoid long apologies and avoid asking the visitor to simply rephrase.
How many times should a chatbot fall back before handing off?
Two is the practical limit. The first fallback offers options, and the second should move the visitor to a human channel such as live chat, a contact form or a phone number. A third identical fallback almost always ends the conversation.
What is a good fallback rate for a chatbot?
For a small business bot with a clear scope, under 10 percent of visitor messages is a healthy target. Higher rates usually point to gaps in what the bot knows rather than a badly written fallback, so fix the content first.
More in Chatbots
Chatbots
Chatbot Privacy and What Your Bot Should Not Store
I read fifty transcripts from a small physiotherapy clinic's website bot last spring, as a favour.
Chatbots
Chatbot Buttons vs Free Text and When to Use Each
A friend who runs a two person accountancy practice replaced her old button menu with a shiny new text box last spring, because...
Order Tracking Chatbot for a Small Online Shop
If you run a small online shop, a large share of your support inbox is probably one question asked fifty ways: where is my...