GUIDE
Chatbot vs live chat, and the option in between
What a chatbot and live chat each cost you in money and hours, where each one lets customers down, and the setup where AI answers first and you step in.
Published
A chatbot answers your website visitors by itself, at any hour, in seconds. Live chat puts a person in the chat, but only while someone is there to watch it. So the choice between them is really a choice about hours. A chatbot covers the night and the weekend, and live chat gives a customer a person when a question needs judgement. If you answer support yourself, you don’t have to pick one. An AI chatbot can answer everything your docs cover, and you can step into the same chat when a question needs you.
Deacon is an AI customer support agent that works that way. It answers your customers in seconds from your own docs, says so when it doesn’t know, and lets you reply in the visitor’s chat from its dashboard. It also shows you what customers are stuck on and what they want next. There’s a free plan, and it needs no card.
Every rival price below was read on the vendor’s own pages on 23 September 2026, and each one links there.
Chatbot vs live chat at a glance
The two differ most in who is behind the chat, and so in when it works.
| Chatbot | Live chat | AI chatbot with a person behind it | |
|---|---|---|---|
| Who answers | Software, from rules or from your content | A person on your team | The AI first, then you when it can’t help |
| When it works | Every hour of every day | While someone is watching | Every hour, with a person during yours |
| First reply | Seconds | However long it takes you to notice | Seconds, and yours when you get to it |
| What you pay for | A plan, credits, resolutions or conversations | Seats, and the hours someone spends watching | One plan, or a chat plan plus an AI add-on |
| Where it fails | Questions your content doesn’t cover, and anything that needs a decision | Nights, weekends and busy days | When it hides the way to a person |
| What you learn | Every question, if the tool saves and groups them | Whatever you remember from the chats you took | Every question, and which ones needed you |
Who answers
- Chatbot
- Software, from rules or from your content
- Live chat
- A person on your team
- AI chatbot with a person behind it
- The AI first, then you when it can’t help
When it works
- Chatbot
- Every hour of every day
- Live chat
- While someone is watching
- AI chatbot with a person behind it
- Every hour, with a person during yours
First reply
- Chatbot
- Seconds
- Live chat
- However long it takes you to notice
- AI chatbot with a person behind it
- Seconds, and yours when you get to it
What you pay for
- Chatbot
- A plan, credits, resolutions or conversations
- Live chat
- Seats, and the hours someone spends watching
- AI chatbot with a person behind it
- One plan, or a chat plan plus an AI add-on
Where it fails
- Chatbot
- Questions your content doesn’t cover, and anything that needs a decision
- Live chat
- Nights, weekends and busy days
- AI chatbot with a person behind it
- When it hides the way to a person
What you learn
- Chatbot
- Every question, if the tool saves and groups them
- Live chat
- Whatever you remember from the chats you took
- AI chatbot with a person behind it
- Every question, and which ones needed you
Two kinds of chatbot
The word covers two very different things, and many complaints about chatbots are really about the first kind.
A rule-based chatbot follows a script someone built. It offers buttons, asks set questions and sends people down branches, and anything the script didn’t expect ends in “Sorry, I didn’t get that.” HubSpot’s free bots are this kind. Its product catalog limits its free and Starter bots to ticket creation, lead qualification and meeting booking, and its help centre says Starter and free users can’t add if/then branches.
An AI chatbot reads what the visitor typed and writes a reply in its own words, usually from content you give it, such as your help pages and docs. It copes with questions nobody scripted, because it answers from what’s written down rather than from a menu. The risk is the opposite one. A careless AI chatbot fills a gap in your docs with a confident guess, which is what people mean by AI hallucinations. Pick one that answers only from your content and says when it doesn’t know. Why support bots make things up explains how to tell them apart.
What live chat asks of you
Live chat is a person replying in a chat window while the visitor waits. The software is cheap, and sometimes free. tawk.to’s live chat costs nothing, with unlimited team members, and LiveChat’s Starter plan is $25 a month for one person, or $19 billed yearly. Free tawk.to alternatives and LiveChat alternatives compare more of them.
The real price is attention. A chat that pings while you’re writing code or on a sales call pulls you out of it, and a visitor who has waited five minutes may be gone by the time you type. Many live chat tools show whether anyone is online, and when nobody is, the chat turns into a form that takes a message. That’s email with extra steps.
Hours are the cost nobody prints
A week has 168 hours. A 40-hour working week covers 40 of them, less than a quarter. To keep live chat staffed every hour, you’d need 4.2 people working 40 hours each, before anyone takes a holiday or falls ill.
Your customers’ time zones make the gap wider. Here’s where a working day of 9am to 5pm in London falls for customers elsewhere in late September.
| Your customer is in | Your 9am to 5pm is their |
|---|---|
| Berlin | 10am to 6pm |
| New York | 4am to noon |
| San Francisco | 1am to 9am |
| Sydney | 6pm to 2am |
Berlin
- Your 9am to 5pm is their
- 10am to 6pm
New York
- Your 9am to 5pm is their
- 4am to noon
San Francisco
- Your 9am to 5pm is their
- 1am to 9am
Sydney
- Your 9am to 5pm is their
- 6pm to 2am
So a founder in London with American customers is online for almost none of San Francisco’s working day. A chatbot answers 24/7, including the question someone in California asks at 2pm their time. Live chat alone leaves them a form.
What each one is bad at
Each fails in its own way, and knowing how tells you which failure you can live with.
Where chatbots let customers down
- They can’t decide anything. Refunds, exceptions, discounts and a customer who is angry need a person, even when the docs cover the topic.
- They can’t see the customer’s account. A bot on your website doesn’t know what plan someone is on or why their card failed, unless you’ve built something to tell it.
- They only know what’s written down. A question your docs never cover gets a guess, a dead end or an honest “I don’t know”, and only the last of those is acceptable.
- They can trap people. A bot that won’t hand over, or makes someone type “human” three times, turns a question into a complaint.
Where live chat lets customers down
- It’s only live while you are. Nights, weekends, holidays and busy afternoons all leave the chat unwatched.
- It doesn’t get faster as you grow. Every chat takes a person, so twice the visitors means twice the chats.
- It spends your time on repeats. The twentieth person to ask how to export their data gets the same answer as the first, typed by you again.
- It forgets. Unless you keep notes, what customers asked last month lives only in your memory.
Try Deacon for free
Install DeaconThe option in between: AI answers, a person takes over
The middle option is an AI chatbot that answers first and hands the conversation to a person when it can’t help, in the same chat window. The AI takes the repeat questions at every hour, and you take the questions that need judgement. For a founder, it’s the one setup that covers both nights and hard questions without hiring anyone.
What matters is the handover. Before you pick a tool, check these five things.
- When it hands over. It should hand over when it can’t answer from your content, when the question is about money or an account, and the first time someone asks for a person.
- How you hear about it. An email or a notification you’ll actually see, so nobody has to sit watching a queue.
- Where your reply lands. The best case is your reply appearing in the chat the visitor still has open, and by email if they’ve gone.
- What the visitor sees. Your name on your reply, so they know a person is answering.
- What the handover costs. Some tools meter the AI and the people separately, so a handover can mean paying twice.
Most live chat tools now sell the AI as an add-on to the chat, with a separate meter, and Tidio vs tawk.to, Tidio vs Crisp and Tidio vs LiveChat compare four of them in pairs.
| Tool | Live chat | The AI on top | How the AI hands over |
|---|---|---|---|
| Tidio | Free for 50 conversations a month that your team replies in, with 10 seats | Lyro, from $39 a month for 50 Lyro conversations, or $32.50 billed yearly | Online, Lyro offers the visitor a person. Offline, it can create a ticket and ask for an email, Tidio’s help centre says |
| tawk.to | Free, with unlimited team members | AI Assist, free for 100 message credits a month, then from $29 a month for 1,000 | Handing over to your team is listed from the $29 Growth plan, not on the free tier, in tawk.to’s billing article |
| Crisp | Free for 2 seats, with no AI | Hugo, from Mini at $45 a month with $5 of AI credits, which Crisp puts at around 90 conversations | Hugo can pass a conversation to your team when it can’t answer confidently, and does when the credits run out |
| LiveChat | From $25 a person a month, or $19 billed yearly | Its sister product Text, from $19 a user a month billed yearly or $25 monthly, with 10 AI resolutions a month | ChatBot.com’s pricing FAQ, for the same Text plans, says every plan includes human takeover |
| Deacon | Reply in the visitor’s chat from the dashboard, on every plan | Free for 50 answers a month. Starter is $50 a month, or $40 billed yearly, for 1,000 answers a month | Says it doesn’t know and asks for an email. The owners get an email when a visitor leaves one or asks for a person. The chat stays open to your reply for twenty quiet minutes, and after that it goes by email only if the visitor left an address in that conversation |
Tidio
- Live chat
- Free for 50 conversations a month that your team replies in, with 10 seats
- The AI on top
- Lyro, from $39 a month for 50 Lyro conversations, or $32.50 billed yearly
- How the AI hands over
- Online, Lyro offers the visitor a person. Offline, it can create a ticket and ask for an email, Tidio’s help centre says
tawk.to
- Live chat
- Free, with unlimited team members
- The AI on top
- AI Assist, free for 100 message credits a month, then from $29 a month for 1,000
- How the AI hands over
- Handing over to your team is listed from the $29 Growth plan, not on the free tier, in tawk.to’s billing article
Crisp
- Live chat
- Free for 2 seats, with no AI
- The AI on top
- Hugo, from Mini at $45 a month with $5 of AI credits, which Crisp puts at around 90 conversations
- How the AI hands over
- Hugo can pass a conversation to your team when it can’t answer confidently, and does when the credits run out
LiveChat
- Live chat
- From $25 a person a month, or $19 billed yearly
- The AI on top
- Its sister product Text, from $19 a user a month billed yearly or $25 monthly, with 10 AI resolutions a month
- How the AI hands over
- ChatBot.com’s pricing FAQ, for the same Text plans, says every plan includes human takeover
Deacon
- Live chat
- Reply in the visitor’s chat from the dashboard, on every plan
- The AI on top
- Free for 50 answers a month. Starter is $50 a month, or $40 billed yearly, for 1,000 answers a month
- How the AI hands over
- Says it doesn’t know and asks for an email. The owners get an email when a visitor leaves one or asks for a person. The chat stays open to your reply for twenty quiet minutes, and after that it goes by email only if the visitor left an address in that conversation
How Deacon covers both halves
Deacon is built for the founder who answers support alone. It’s an AI chatbot that answers first, with you behind it, and no live chat seat you have to sit in.
It answers in seconds, at any hour. Deacon replies from your website, your files, text you paste in and the answers you write, in the language each visitor writes in. When your content doesn’t cover a question, it says it doesn’t know instead of guessing, and asks for the visitor’s email. It keeps answering their other questions meanwhile. It never answers a question about a visitor’s own account, invoice or data, so money problems come to you rather than getting a guess.
You hear about the ones that need you. The owners on your team get an email three minutes after a visitor leaves an address or asks for a person. If someone on your team already has the dashboard open, or has opened that conversation, no email goes out, because the dashboard has already told them.
You reply in their chat. Type your reply in the conversation in Deacon’s dashboard. Sending it takes the conversation from Deacon, which stops answering in it. A visitor with the chat open sees your reply there, under your first name and your company’s, such as “Alex from Acme”, and the box they type in changes from “Ask AI” to “Reply to the team”. The chat stays open to your reply for twenty quiet minutes. After that, your reply goes by email, and only if the visitor left an address in that conversation. When you’re done, press Hand back to Deacon, and it tells the visitor it’s answering again.
Nobody has to watch it. There’s no queue, routing or online status, so the chat never tells a visitor you’re away. Your replies, and the visitor’s messages to you, don’t count towards your plan’s answers.
It needs you less over time. Every question your content didn’t cover has an Answer this button. Write the answer once and press Save & check, and Deacon shows you the reply it now gives and how many other open questions your answer covers. Every answer you write becomes part of what Deacon knows, so the more you answer, the less it needs you.
You learn what customers want, the customer insights live chat rarely leaves behind. Every conversation is saved and searchable by what visitors asked. Topics group similar questions, count how many people asked and label most of them as a how-to, a bug report, a feature ask and so on, so you hear from everyone who writes in and not just the loudest few.
Every feature is on every plan, the free one included. Free gives you 50 answers a month with no card and no end date. Starter is $50 a month, or $40 a month billed yearly, for 1,000 answers a month and 3 seats. Nothing is billed past a plan’s answers, and no plan upgrades itself.
Which one to choose
Start from who answers and when they’re around.
If you answer support alone and your customers are spread across time zones, choose an AI chatbot with a way to reach you. It covers the hours you can’t, and hands you the questions that need you. The best AI chatbots for customer service ranks them by when a person can take over.
If you sell large contracts and can staff the chat during your customers’ working day, live chat earns its seats. Add a chatbot for the hours nobody is on, so the chat never goes quiet.
If your product has no docs yet, write the answers to your ten most common questions first. A chatbot can only answer what’s written down, and those pages help every visitor whether they chat or not. How to reduce support tickets shows how to find those ten.
If a team of people answer from a rota, you need live chat with routing and assignment, usually as part of a helpdesk. Deacon has neither, so it suits the founder more than the support floor. Helpdesk automation shows which plan unlocks routing, and AI help desks how each bills its AI.
Questions about chatbots and live chat
Is a chatbot the same as live chat?
No. A chatbot is software that replies by itself, and live chat is a person replying in real time. Many tools now offer both in one chat window, with the bot answering first and a person taking over.
Can a chatbot replace live chat?
For the questions your docs already answer, yes, and at hours live chat can’t cover. It can’t replace a person for refunds, exceptions or a sale that hinges on something your docs don’t say. The setup that works is a chatbot that answers what it can and hands the rest to you.
Do customers prefer talking to a person?
They prefer an answer. A correct reply in seconds beats waiting for a person, and a person beats a bot that guesses or loops. What annoys people is a bot that won’t hand over, so make the way to a person easy to find.
Does Deacon have live chat?
Yes, the kind a founder needs. Anyone on your team can reply in any conversation from the dashboard, and a visitor with the chat open sees the reply there. There’s no queue, routing or online status, so nobody has to sit waiting for chats.
What if I don’t reply to a handover straight away?
Deacon keeps answering the visitor’s other questions until you reply. The chat stays open to your reply for twenty quiet minutes. After that, your reply goes by email, and only if the visitor left an address in that conversation. That’s why Deacon asks for one when it can’t answer.
Does the visitor know when a person is answering?
Your replies appear under your first name and your company’s, without the Deacon mark, and the box they type in changes from “Ask AI” to “Reply to the team”. No other notice says a person has joined. When you hand back, a line in your name tells them the AI assistant is answering again.
Try it on your own documentation
Add your help pages, ask it the question you know your docs cannot answer, and watch it say so.
The free plan covers 50 answers a month on one seat, and the paid plans are on the pricing page.
Free plan, no card.
READ NEXT
Why do chatbots hallucinate, and how to stop yours
Chatbots guess because they were trained to. Why support bots invent policies, what it has cost two companies, and the rules that make one say it doesn’t know.
How much does an AI chatbot cost? Four ways to pay
What an AI chatbot costs when it’s billed by the seat, the credit, the resolution or the month, with one month of 1,000 questions priced on each.
Customer support tools for small teams, with prices
The five kinds of customer support tool, what each costs on prices read on 23 September 2026, and how the number of people answering tells you which you need.