Should You Tell Customers an AI Is Replying on WhatsApp?
The short answer
Yes, in most cases. A short line saying a reply is AI assisted costs almost nothing, and it is far cheaper than a customer working it out later and feeling deceived. Rules in some regions are also moving towards requiring it. The reply still has to be accurate and useful, because disclosure excuses nothing.
The WhatsAssist team ·
Yes, in almost every case, and the reason is not idealism. A short line saying a reply was written with AI help costs you very little. A customer discovering it for themselves, after they had assumed they were talking to you, costs a great deal more, because what they learn is not that you use software. It is that you let them believe something untrue.
The practical question is not whether to say it. It is how to say it in a way that does not undercut the reply it is attached to.
Does telling customers actually cost you anything?
Less than most owners expect, and the fear is usually about the wrong thing.
People do not object to a business using software. They object to being handled. The complaint behind “I was talking to a bot” is almost never about the technology; it is that the answer was useless, or circular, or that there was no way to reach a person. A disclosed reply that answers the question properly generates almost no reaction at all.
There is also an asymmetry worth taking seriously. Nobody has ever lost a customer by mentioning that a reply was AI assisted. Plenty of businesses have lost customers who felt they had been fooled. One of those risks is small and survivable, and the other damages the thing you cannot rebuild quickly.
The exception is where disclosure would be absurd rather than honest. A one word confirmation, an address, an opening time. Labelling every trivial message turns a courtesy into noise, and noise is what people learn to ignore.
What does a good disclosure line look like?
Short, factual, at the end, and never an apology. The reply is the message. The line is a footnote.
Illustration. Not a real customer.
“Yes, we have that in [size], and it is [price] including delivery.
Replies here are AI assisted. Ask for [your name] any time and I will pick it up myself.”
Three things make that work. It states the fact plainly. It gives an exit to a human, which is the reassurance people actually want. And it does not open with the disclosure, because a message that leads with its own caveat reads as lower-value before the reader reaches the answer.
What to avoid is the pretence in either direction. Do not give the assistant a human name and a personality, which is the version customers find genuinely offensive when they work it out. And do not over apologise for it either, since treating it as something shameful invites the reader to agree.
Where is disclosure becoming a legal duty rather than a courtesy?
The direction of travel is towards requiring it. The clearest example is the European Union’s AI Act, Regulation (EU) 2024/1689, which includes a transparency obligation for AI systems that interact directly with people: they should be built so that the person is made aware they are dealing with an AI, except where that is already obvious from the context.
Treat that as the shape of the rule and not as advice about your own situation. Whether it applies to you depends on where you and your customers are, what you are using the system for, and how the rules are being enforced where you trade, and those differ. Other jurisdictions are moving at different speeds and some consumer protection regulators are reaching similar conclusions under older rules about misleading practices. This is not legal advice. If you sell into Europe and the answer matters to your business, check it properly rather than relying on a blog post.
The useful takeaway is simpler than the law: a practice you would have to stop if the rules tightened is a practice worth reconsidering before they do.
When should you disclose, and when is it just noise?
| Situation | Say it | Why |
|---|---|---|
| A first substantial reply to a new customer | Yes | This is where the assumption gets set, and it is cheapest to be honest early |
| Anything about money, a promise or a commitment | Yes | If it later turns out to be wrong, how it was written becomes part of the argument |
| A complaint or anything emotionally charged | Yes, and offer a person | Someone who is upset needs a route to a human more than they need a fast answer |
| A one word confirmation, an address, opening hours | No | Labelling trivia trains people to skim past the label |
| Every single message in a long thread | No | Repetition turns disclosure into wallpaper and it stops being read |
| You are actually typing the message yourself | No | It would not be true |
The pattern is that disclosure earns its place where a reader’s assumption could change what they do. Everywhere else it is decoration.
What does WhatsAssist do about this by default?
It is worth being concrete, because the answer differs by plan and that is a real difference rather than a detail.
On Free and Basic, a short line identifying the reply as coming from an AI assistant is added automatically, in the customer’s language. You do not configure it and you cannot remove it. On Pro and Max there is no such line, so the decision moves to you, and this article is our answer to what you should do with it. What each plan includes is on the pricing page, and what the product does and does not do is on the features page.
Two related behaviours matter more than the label itself. It never starts a conversation, so nobody receives an AI message from you out of the blue. And when it would have to guess, it does not answer: that reply waits for you in the app with the reason attached, rather than producing something confident and wrong for you to disclose afterwards. That second one is the substance behind the label, and what happens when a reply gets something wrong goes through it properly.
Does disclosure make the reply itself matter less?
The opposite, and this is the part that gets missed. Once you have told someone a reply is AI assisted, you have removed your excuse. A wrong price is no longer a typo they might forgive; it is the software you chose, answering for you, in your name.
So the disclosure line is the easy half. The hard half is that everything the assistant is allowed to say has to be something you would have said: your real prices, your real policies, your real limits. If you would not be comfortable seeing a reply screenshotted next to your business name, the fix is not a better label. The wider version of this argument is in how to use AI to reply to customers without sounding like a bot.
Frequently asked questions
Will customers stop replying if I tell them it is AI?
Overwhelmingly no, provided the answer is actually useful and a person is reachable. What loses people is a reply that does not answer the question, or a thread with no way out to a human. Neither of those is caused by the disclosure.
Where should the line go, at the start or the end?
At the end. A message that opens with a caveat gets read as less valuable before the reader reaches the answer. Answer first, then note how it was written.
Do I have to disclose it by law?
It depends on where you and your customers are, and it is changing. The EU AI Act contains a transparency obligation for systems that interact with people directly, and other regulators are moving in the same direction. Check your own jurisdiction rather than relying on a general article, and treat this as a reason to disclose voluntarily rather than a reason to wait.
Can I turn the disclosure line off in WhatsAssist?
Not on Free or Basic, where it is added automatically. Pro and Max do not add a line at all, so on those plans the choice is yours to make.
Does it work on an iPhone?
No. WhatsAssist depends on Android notification access and requires Android 10 or newer. There is no iPhone version and no announced release date.