Skip to main content
← Back to guides

How to Write a Chatbot Welcome Message That Actually Gets a Reply

AIChatbotGuides
Karan Gosrani
Team Converzoy|
How to Write a Chatbot Welcome Message That Actually Gets a Reply

Here's something nobody tells you when you set up a chatbot: the first message it sends is the only one most people will ever see.

Not because the rest of your setup is bad. Because visitors make a decision in the first three seconds about whether to engage, and if that opening message doesn't give them a reason to type something, they won't. They'll close the widget and keep scrolling. You'll have a chatbot that technically works and practically does nothing.

The frustrating part is that almost everyone starts with the same opening message. "Hi! I'm [BotName]. How can I help you today?" It's what the platform suggests. It's what most examples show. And it's why most chatbots have engagement rates that make you question why you bothered.

What's Wrong With "How Can I Help?"

Put yourself on the other side for a second. You've landed on a website, you're still orienting yourself, and a chat widget pops up asking how it can help. Help with what, exactly? You haven't decided what you're looking for yet. You don't know if the bot can handle whatever you're about to ask. The question puts all the work on you before you've had a chance to decide you want help at all.

Most people don't answer. Not because they don't have questions, but because the question is too open-ended and the effort isn't worth it when they can just keep reading.

The other problem is invisibility. Years of chatbot popups have trained people to register and dismiss them in the same half-second. Your widget fires, their eyes flick to it, they categorise it as "chatbot," and they move on without consciously reading a word. A generic greeting confirms that categorisation instantly. A specific, relevant statement breaks it.

That's the real job of a welcome message. Not to announce the chatbot. Not to be friendly. To say something specific enough that the right visitor stops and thinks, "actually, yes."

What Specific Looks Like in Practice

Specificity doesn't mean long. It means relevant. Two sentences that speak directly to what someone is probably thinking when they land on that page will outperform four sentences of cheerful generality every time.

On a pricing page, the person is trying to figure out if the cost makes sense for what they're getting. They might be comparing plans, hunting for something that's not listed, or trying to understand what "per seat" actually means in practice. A message like "Comparing plans or trying to figure out what's included? I can walk you through it" works because it names the mental task they're already doing. It doesn't ask them to start from scratch.

On a product features page, the visitor is trying to understand what something does. They might be sceptical, they might be sold and just verifying, they might be a technical person who needs specifics that the marketing copy doesn't give them. "Questions about how [feature] works? Happy to explain it in plain terms" gives them an easy on-ramp without assuming they're either a beginner or an expert.

On a homepage or blog post — where intent is much harder to read — the message needs to be broader. But broader doesn't mean vague. "Not sure if this is what you're looking for? Tell me what you're trying to solve and I'll point you in the right direction" is still doing something a generic greeting doesn't: it acknowledges that the visitor might be uncertain, which is true for a lot of people who land on a homepage without knowing exactly what they want.

Telling Them What You Can Do

One thing that consistently improves engagement: being explicit about scope before they ask.

Most visitors have no idea what your chatbot can handle. They don't know if it's a glorified FAQ widget or something that can actually help them. And rather than risk the frustration of asking something it can't answer, they don't ask anything.

"I can help with questions about pricing, features, or getting started with a trial" removes that uncertainty. It's not a guarantee that the bot will handle everything — it's a signal that it handles something specific, which is enough to lower the barrier to starting.

Be accurate here. If your bot is great at product questions but struggles with anything billing-related, say so. "I can answer most product questions right away — for billing stuff, I'll connect you with the team." That kind of honesty builds more trust than a bot that confidently fumbles through questions it's not equipped to handle. We've written about [what actually separates good chatbot experiences from frustrating ones](https://converzoy.com/insights/ai-chatbot-customer-experience-done-right) — managing expectations is near the top of that list.

The Language Thing Nobody Pays Attention To

Read your last 30 support tickets before you write a single word of your welcome message.

Not to copy the questions. To copy the vocabulary. The words customers use when they're confused, frustrated, or trying to describe what they need. Those are the words your chatbot should use. Not because it makes the bot sound more human — but because when people read language that matches how they think about a problem, they feel understood before they've typed anything.

If your customers say "set up" not "onboard," say set up. If they say "connect" not "integrate," say connect. If they refer to your main feature by a nickname rather than its official product name, use the nickname. It sounds like a small thing. The difference in how it feels to the reader is not small.

The same logic applies to formality. A B2B enterprise product and a consumer app don't talk the same way, and their chatbots shouldn't either. Read your sales calls. Notice how prospects talk when they're relaxed. That's the register you're aiming for.

Timing Is Half the Battle

You can write the best opening message in the world and still get ignored if it fires at the wrong moment.

Three seconds after someone lands on your homepage? They're still reading the headline. They haven't formed a question yet. A chatbot popup at that moment doesn't feel like help — it feels like the digital equivalent of a sales rep who corners you the moment you walk into a shop.

The visitors most likely to engage are the ones who've spent enough time on a page to actually have a question. 30 to 45 seconds on a pricing page is usually when it lands well. After 60% scroll depth on a long feature page. When someone clicks back from a specific section and seems to be comparing information. These are the moments when a well-timed message feels useful rather than interruptive.

Most platforms let you set triggers per page and per condition. If you're using one universal trigger across the whole site, that's probably contributing to your engagement problems as much as the copy is. Test changing the timing before you change the message — you might find the message is fine and the problem is purely when it appears.

Checking Whether It Works

Give any new welcome message two full weeks before judging it. Traffic is variable, and a one-week sample can mislead you in either direction.

What to look at: engagement rate (the percentage of visitors who open the chat after seeing the message) and conversation completion rate (the percentage of started conversations that reach something useful — a question answered, a lead captured, a next step taken). If engagement goes up but completion stays flat, the opening is working but there are gaps further into the flow. Our guide on [chatbot scripts](https://converzoy.com/guides/how-to-write-chatbot-scripts-that-convert) covers how to think about those gaps.

If engagement stays flat despite a better message, look at timing before you look at copy again. And if you're seeing high engagement but visitors drop off on the second or third reply, the welcome message isn't the issue — something later in the conversation is.

Most teams write their welcome message once and never touch it again. That's usually the difference between a chatbot that quietly improves over time and one that never quite delivers what you hoped. Thirty minutes every month looking at recent conversations and asking "what should the opening have said?" compounds faster than you'd expect.

[Try Converzoy](https://app.converzoy.com/signin) to test different welcome messages across different pages — changes go live instantly, no code needed.

Ready to convert more visitors?

Try Converzoy free. No credit card required.

Get started for free