
How to Set a Chatbot Session Timeout and Handle Customers Who Come Back
Quick answer: Learn what your platform does by default, then choose the silence you will tolerate from how long your customers really pause. Add a gentle reminder before the close, write a close message that says how to restart, let returning customers pick up where they left off if you can, and test the expiry. The number matters less than the messages around it.
Step 1: Learn what your platform does by default
Before changing anything, find three facts in your platform's documentation: the default inactivity period, what resets it, and what expiry clears. Vendors differ a lot. Google's Dialogflow CX session page says a session stays active for 30 minutes after the last request and can be extended with a time-to-live setting up to 24 hours. Microsoft's Copilot Studio inactivity trigger page states no default; you choose the duration (up to seven days) and the conversation ends only if your topic ends it. Intercom's auto-close article describes a workflow measured in days.
If the documentation is silent on one of the three facts, test it (Step 6) rather than guess.
Step 2: Measure how long your customers actually pause
Do not pick a number from a blog post, including this one. Open a sample of recent conversations that continued after a gap and note how long the gap was. Ask two questions:
- How long do people take to find an order number or check something? That is a normal pause; do not time out inside it.
- After what gap do they rarely come back in the same sitting? That is where a reminder belongs.
A rough rule: the reminder goes after the normal pause, and the close goes a short while after the reminder. For a site chat, minutes; for a messaging app where people reply hours later, hours or days. Write the choice down with the reason.
Step 3: Choose the sequence: reminder, then close
A silent reset is the worst option. Prefer three beats:
- Nudge. One short message after the pause.
- Notice. A second message that says the chat will close, and how to continue.
- Close. End the session and keep a way back.
Copilot Studio's page shows one way to build this: a trigger runs a topic after silence, and multiple triggers at different durations are ordered shortest first. Intercom's template does a similar thing with a check-in, a snooze and a closing notice. If your platform has no such feature, a scheduled job that checks the last message time can do the same.
Step 4: Write the three messages
Short, plain, no guilt. Templates to adapt:
Nudge: Still there? I'm holding your place. Tell me when you're ready and we'll carry on from [last step].
Notice: I'll close this chat soon to keep things tidy. Reply any time before then to keep going.
Close: I've closed this chat. Start a new one whenever you like, or email [address] and a person will pick it up.
More wording ideas are in chatbot fallback copywriting and chatbot welcome message design. If a person was already involved, do not close on them without checking the handoff design guide.
Step 5: Let returning customers resume
A customer who returns after the close should not have to start over. Options, from simplest to hardest:
- Show the transcript and say what you remember.
- Offer a choice: "Continue with your return request, or start something new?"
- Restore state. Dialogflow CX's page describes saving the current page and parameters yourself and sending them back under a new session ID. It notes the handler call stack is not restored.
- Hand the saved answers to a person with a short recap; see summarizing chatbot conversations for handoff.
Long chats can also lose early messages because of the model's input limit, which is a separate problem; see managing the chatbot context window and the chatbot memory guide.
Step 6: Test the expiry
Write the cases down and run them after every setting or wording change.
| Test | Expected result |
|---|---|
| Reply just before the nudge | Chat continues, timer restarts |
| Say nothing past the nudge | Nudge arrives once, not repeatedly |
| Say nothing past the notice | Close message arrives, session ends |
| Return one minute after the close | Resume offer or clear new start |
| Return a day later | Same, with the old context stated or cleared as designed |
| Silence while a person is on the chat | No bot close message |
| Channel check (web, app, messaging) | Same behavior or documented difference |
Then check privacy: what does expiry clear, and what stays in storage? Timeout and retention are separate settings; see data retention policy, the chatbot data privacy guide and, if it applies to you, the GDPR chatbot compliance guide. This is not legal advice.
Check the arithmetic
An invented example, not a measurement. A small shop's chat handles 100 conversations a week, and 20 of them go quiet before they finish. After adding a nudge, 6 of those 20 come back and finish: 6 ÷ 20 = 30%. The other 14 receive the close message with an email address. Counted before the change, all 20 would have counted as silent drop-offs; counted after, 14 remain and 6 are recovered.
Suppose the nudge goes out after 10 minutes of silence, the notice follows 5 minutes later, and the session closes 5 minutes after that. The customer has 10 + 5 + 5 = 20 minutes in total before the session ends. If the platform's own default is 30 minutes, your sequence shortens it; if the platform's maximum is lower than your plan, the platform wins. These numbers are made up to show the method.
Common mistakes
- Changing the number without reading what expiry clears.
- Closing silently, so the next message looks like a bug.
- Nudging repeatedly every few minutes.
- Closing on a customer while a person is helping them.
- Mixing up timeout and retention, and deleting transcripts by accident or keeping them too long.
- Testing only on web chat when customers also use messaging apps.
FAQ
How do I set a chatbot session timeout? Find your platform's default and what expiry clears, measure how long your customers pause, choose a nudge-notice-close sequence, write the messages, offer a way to resume, and test every step.
How long should a chatbot wait before ending an inactive conversation? There is no standard number. Dialogflow CX documents 30 minutes by default; Copilot Studio documents no default for its inactivity trigger; Intercom's template works in days. Base yours on observed pauses.
What should the reminder message say? That you are holding their place and how to continue, in a sentence or two, without pressure.
Can customers resume a chat after it closes? Yes if you build it: show the transcript, offer a choice, or restore saved state. Some platforms do not provide this by default.
Does a timeout delete the conversation history? Not necessarily. Check what your platform clears at expiry; retention is a separate policy.
What if a person is already in the chat? Do not let a bot timeout close it. Check how your platform treats conversations that were handed to an agent.
Sources
- Google Cloud, Dialogflow CX: Session — docs.cloud.google.com/dialogflow/cx/docs/concept/session, read 6 November 2026. Source of the 30-minute default, the 24-hour maximum for the extended lifetime and the restore-state option. Vendor documentation; may change.
- Microsoft, Inactivity trigger guidance (Copilot Studio) — learn.microsoft.com/en-gb/microsoft-copilot-studio/guidance/inactivity-trigger-guidance, read 6 November 2026. Source of the configurable duration, seven-day maximum, ordered triggers and End conversation behavior. States no default.
- Intercom, Auto-close inactive conversations — intercom.com/help/en/articles/9636573-auto-close-inactive-conversations, read 6 November 2026. Source of the check-in, snooze and close sequence and default timings. Silent on bot-only conversations and plan requirements.
- Chatbotscape evaluation methodology. /methodology (continuously updated).