Answer delay: why the bot waits before replying
People rarely write a question in one message. They type a phrase, then a clarification, then a question mark. If the bot replied to every message instantly, the customer would get three answers to one thought. Answer delay is the waiting window that turns those messages into a single question. This guide explains how the window works, where double replies come from and which value to choose.
What the setting does
Answer delay is in the bot settings, in the behaviour section. It sets the pause between the customer's message and the bot's reply: from 0 seconds to 15 minutes, the default is 10 seconds. Two things happen during the pause.
- Messages are merged. Everything the customer sends while the window is open is collected into one request. The bot reads the whole thought, not the first fragment of it
- The bot looks less like a bot. An instant reply gives away automation; a pause of a few seconds feels like a person reading the message
How the waiting window works
The first message from the customer starts a timer equal to the delay. Every next message that arrives before the timer runs out is added to the same request and the bot keeps waiting. When the timer runs out, the bot generates one reply to everything it collected. A message that arrives after the window has closed starts a new timer and a new reply.
| Time | Customer | Bot (delay 15 seconds) |
|---|---|---|
| 00:01 | "hi, do you deliver" | Timer started |
| 00:10 | "to the north of the city" | Added to the same request, still waiting |
| 00:16 | "?" | The window closed at 00:16. The first reply is already being prepared; "?" opens a new window and gets a second reply |
Why the bot sometimes replies twice
This is the situation from the table: the customer sends the last fragment a second after the window closed. The bot has already started answering the first part and treats the tail as a new question. The result is two replies in a row, often saying almost the same thing. It is not a bug in the script: the window was simply shorter than the customer's typing.
Double replies are more likely in Telegram and WhatsApp, where people write in short bursts, and less likely in the website widget, where the message is usually composed in one go.
How to choose the delay
| Value | When it fits | Trade-off |
|---|---|---|
| 0 to 5 seconds | Website widget, quick FAQ bots, customers who expect an instant answer | Occasional double replies when someone types in fragments |
| 10 to 15 seconds | The default. Good for most Telegram and WhatsApp conversations | A short pause that still feels natural |
| 30 to 60 seconds | Personal chats where the bot replies on your behalf, customers who type slowly or dictate voice messages | The customer waits longer, but every thought is collected in one request |
| 2 to 15 minutes | Cold or low-priority conversations where a human-like pace matters more than speed | Customers may think nobody is there; combine with a greeting in the script |
The choice comes down to what matters more for you: the bot's instant reaction or a clean conversation without repeats. If speed is critical and customers rarely write "in pursuit", a few duplicates are an acceptable cost. If you see two replies in a row regularly, raise the delay to 30 to 60 seconds: the bot stays quiet a bit longer but collects even slowly typed messages into one thought.
Related settings
- If you don't answer… in the bot schedule is a different timer: it decides whether the bot steps in at all when your team stays silent for 1 to 30 minutes. Answer delay applies after that decision
- Maximum number of responses caps how many replies the bot gives in one conversation. Double replies count towards this limit
- Reactivation of frozen conversations sends follow-ups after hours or days of silence and is unrelated to the delay
Does the delay spend tokens?
No. Tokens are spent on the generated reply. Merging several messages into one request usually saves tokens because the bot answers once instead of several times.
Can the delay differ per channel?
The delay belongs to the bot. To have a fast bot for the website and a slower one for Telegram, duplicate the bot, change the delay and assign each copy to its account.
Does the delay apply when an operator replies?
No. The delay only affects the bot's own replies. After a manual reply the bot stays quiet in that conversation until you return it to the bot.
See also
Try BotB2B for free
Sign-up takes a minute and starter tokens are on us. Set everything up with this guide.