AI Chatbots
How to Annoy Visitors With a Chatbot
Last updated:
The two worst
- No visible route to a person. This is the source of nearly every complaint about chatbots.
- Demanding an email before helping. It reads as a gate and the visitors closest to buying leave.
Fix those two and you have addressed most of what people dislike about chat widgets, before touching the answer quality.
Three about behaviour
- Popping up unprompted before the visitor has done anything
- Following them around with repeated prompts on every page
- Blocking the content on mobile, where the screen is small
Two about honesty
- Pretending to be human when asked directly
- Inventing answers rather than saying it does not know
Both are recoverable once. Neither survives being discovered twice.
What good behaviour looks like
| Instead of | Do |
|---|---|
| Popping up on arrival | Wait to be opened |
| Email before answering | Answer, then offer to follow up |
| “Let me connect you” that goes nowhere | A real route with a real timescale |
| Guessing | “I do not know — shall I get someone?” |
| A persona | A clear answer |
Test it on your phone
Most of these problems are worse on mobile, where a widget covering the content is genuinely obstructive and a route to a person is harder to find.
Open your own site on a phone and try to get help. It takes two minutes and it finds most of the seven.
Frequently asked questions
Should a chatbot ever pop up automatically?
Rarely, and if so, after real engagement and only once. Never on arrival.
What if visitors ask if it is a bot?
Say yes. It costs nothing and denying it costs everything.
Is it acceptable to require an email for a detailed answer?
After answering the immediate question, yes. Before, no.
How do we know if ours is annoying?
Read the conversations that ended abruptly. People tell you, usually quite directly.
Have a chat widget people complain about?
It is usually two of these seven. Happy to take a look at what you have.