Chatbot vs Live Chat in 2026: Real Costs, and Which Wins Each Conversation
A chatbot answers instantly and costs about $0.50 to $1.50 per conversation. Live chat puts a real person on the other end and costs $10 to $25 for the same routine question, because 94% of that cost is wages, not software. But cheap does not mean better: chatbots resolve 85–95% of “where is my order” questions and only about 17% of billing disputes, which humans resolve 58% of the time. The right answer is not one or the other. It is a chatbot on the easy volume and a human on anything involving money, anger or apologies.
Most articles on this subject end with “use both” and leave you no wiser. This one gives you the actual numbers: what each costs per conversation, how fast each replies, which questions each one wins, and exactly when the bot should step aside.
Table of Contents
What is a chatbot, and what is live chat?
A chatbot is software that reads a customer’s message and answers it by itself, at any hour, without a person involved. Modern ones read your help pages and write real answers instead of offering menu buttons.
Live chat is the same little window on your website, except a real member of your team is typing the replies. It only works when someone is at their desk.
Think of a corner shop. The chatbot is the sign on the door listing your opening hours, your prices and where the milk is. It answers the same three questions for everybody, instantly, forever. Live chat is the shopkeeper. She can find the thing you cannot describe, notice you look upset, and refund you on the spot.
You would not run a shop with only a sign. You also would not pay someone to stand there reading out the opening hours all day.
That is the whole difference between chatbot and live chat, and everything below is detail.
People search for this both ways round — chatbot vs live chat and live chat vs chatbot — and it is the same question: which one should answer this particular customer, right now?
What are the main differences between chatbot and live chat customer support?
The main differences are who replies, how fast, what it costs and when it is available. A chatbot replies in about one second, works 24 hours a day, costs roughly $0.50 to $1.50 per conversation and handles unlimited chats at once. Live chat replies in about 2 minutes 40 seconds on average, works only during staffed hours, costs $10 to $25 per routine conversation and one agent can hold only a handful of chats at a time.
| Chatbot | Live chat | |
|---|---|---|
| Who replies | Software | A person on your team |
| Average reply time | About 1.1 seconds | 2 minutes 40 seconds, and 7–12 minutes at busy times |
| Available | 24 hours, every day | Only when someone is working |
| Cost per conversation | About $0.50–$1.50 | About $10–$25 for a routine query |
| Chats at the same time | Effectively unlimited | A few per agent |
| Biggest cost | Software | Wages, about 94% of the total |
| Handles brand new problems | Poorly | Well |
| Gets tired or grumpy | Never | Sometimes, and that is normal |
| Can say sorry and mean it | No | Yes |
The row people underestimate is the last one. An apology from software is a string of words. An apology from a person who can also fix the problem is what keeps a customer.
What are the key differences between AI chatbots and human customer service?
The key differences are judgement, authority, empathy and memory. An AI chatbot is excellent at recalling facts and following rules, and it never forgets your policy. A human can weigh up a situation that no policy covers, decide to bend a rule, hear that someone is upset before they say so, and take responsibility for a mistake. Research by Drift found people rated human replies about four times better than bot-only replies.
| Ability | AI chatbot | Human agent |
|---|---|---|
| Remembering facts and policy | Perfect, every time | Good, with slips |
| Following a set process | Perfect | Good |
| Judging an unusual case | Weak | Strong |
| Bending a rule to keep a customer | Not allowed to | Can, with authority |
| Hearing that someone is upset | Getting better, still unreliable | Natural |
| Taking responsibility | Cannot | Can |
| Working at 3am | Yes | Only if you pay for it |
| Handling 500 chats at once | Yes | No |
| Cost of the 500th chat | Almost nothing | Another wage |
The honest summary: AI is better at knowing things. People are better at deciding things. Most customer conversations only need knowing. The ones that need deciding are the expensive ones, in both senses.
Which conversations each one wins
This is the table that should decide your setup, because it is based on what each actually resolves.
| Conversation | Chatbot resolves | Human resolves | Who should take it |
|---|---|---|---|
| Where is my order? | 85–95% | — | Chatbot |
| What are your opening hours or prices? | Almost always | — | Chatbot |
| How do I reset my password? | Usually | — | Chatbot |
| Do you deliver to my area? | Usually | — | Chatbot |
| A billing dispute | About 17% | About 58% | Human |
| “You charged me twice and I am furious” | Rarely | Usually | Human, immediately |
| A complaint about a policy | Rarely | Usually | Human |
| Something nobody has asked before | Rarely | Usually | Human |
| A big customer deciding whether to renew | Rarely | Usually | Human, and call them |
Look at the billing row again: 17% against 58%. That one comparison is why “just get a chatbot” is bad advice, and why “humans only” is expensive advice. Send the easy 60% to software and put your people where they change the outcome.
Real costs: chatbot vs live chat
What a chatbot costs
| Platform | Cost per automated conversation |
|---|---|
| Freshdesk (Freddy) | $0.10 per bot session |
| Markleyo | Included in $14.50/month, no per-chat charge |
| Kustomer | $0.60 per engaged conversation |
| Gorgias | $0.90–$1.00 per AI interaction, plus the ticket charge |
| Intercom Fin | $0.99 per resolution, 50 a month minimum |
| Zendesk | $1.20–$1.50 per verified resolution |
Industry estimates put the all-in figure at roughly $0.50 to $0.70 per interaction.
What live chat costs
| Cost | Amount |
|---|---|
| LiveChat software | About €20 per agent per month |
| Crisp | About €25 per month, flat |
| Tidio | About €29 per month (3 seats included) |
| Zendesk | About €55 per agent per month |
| Intercom | About €74 per seat per month |
| Wages | The other 94% of the bill |
Three in-house agents covering 24 hours a day costs roughly $7,500 to $12,900 a month all in. A five-person team on LiveChat pays about €200–€590 a month for software — and many times that in salaries.
For comparison, a human-handled support ticket costs $6–$7 on average across industries, $2.70–$5.60 in retail, $18–$35 in software, and $30–$60 in business-to-business support. Phone costs three to five times more than chat for the same problem.
A worked example
A shop gets 1,000 chats a month.
| Setup | Monthly cost | What happens |
|---|---|---|
| Humans only | About $4,000–$10,000 | Everything answered well, in working hours, slowly at peak |
| Chatbot only | About $500–$1,500 | Fast answers, but the angry and complicated ones go unresolved |
| Chatbot for 60%, humans for 40% | About $2,000–$4,500 | Fast on the easy 600, proper care on the 400 that matter |
The hybrid is not a compromise. It is the only one where the expensive resource is spent on the expensive problems.
How does chatbot software improve customer service efficiency versus live chat?
Chatbot software improves efficiency in four ways: it replies in about one second instead of nearly three minutes, it handles unlimited conversations at the same time instead of a handful per agent, it works overnight and at weekends without overtime, and it removes the repetitive 40–60% of questions so your team can spend their day on the hard ones. Well-integrated bots keep 70–90% of conversations from reaching a human at all.
But there is a catch, and honest vendors will tell you about it.
Containment is not resolution
Three words get mixed up in sales calls, and they mean very different things.
| Word | What it means | Does the customer’s problem get fixed? |
|---|---|---|
| Deflection | They read a help article instead of starting a chat | Nobody checked |
| Containment | The chat ended without a human joining | Maybe. They may have given up |
| Resolution | The problem was fixed and they did not come back | Yes. This is the only one that counts |
Ada published data showing 72% containment but only 52% resolution on the same conversations — a 20-point gap. Intercom reports a 76% average resolution for Fin across 12,000 customers. Realistically, expect 40–60% in your first months, rising past 60% after six to twelve months of improving your help content. Teams that never update that content stall at 30–45%.
So ask any vendor for the resolution rate, not the containment rate, and ask how they check the customer did not come back.
Where live chat earns its money back
Live chat looks expensive when you judge it as support. Judge it as sales and the maths flips.
| Effect of live chat | Reported figure |
|---|---|
| Increase in conversion rate | About 20% on average |
| Visitors who chat versus those who do not | 2.8× more likely to buy |
| Spend per purchase | About 60% higher |
| Cart abandonment | Down 25–30% with proactive chat |
| Companies seeing positive return in year one | 57% |
A word of caution on these: most come from vendors who sell chat software, so treat them as a direction of travel rather than a promise. The underlying logic holds, though. A person who answers “will this fit my kitchen?” in the moment of doubt closes a sale that a help article would have lost.
Speed is the bit both sides agree on. Research from MIT and InsideSales found replying to a new lead within five minutes made it up to 21 times more likely to qualify than replying after 30 minutes. A chatbot buys you that speed. A human closes what the speed brings in.
Top platforms offering both chatbot and live chat functionalities
The top platforms that do both are Markleyo ($14.50/month flat, with chat and human handover across five channels), Tidio (from about €29/month with Lyro AI as an add-on), Crisp (about €25/month flat), Intercom (about €74/seat plus $0.99 per AI resolution), Zendesk (about €55/agent plus $1.20–$1.50 per resolution) and Gorgias (from $10/month plus $0.90–$1.00 per AI interaction). The cheapest ones bill flat; the most capable ones bill per seat and per resolution.
| Platform | Chatbot | Live chat | Billing model | Entry cost |
|---|---|---|---|---|
| Markleyo | Yes | Handover to your team | Flat | $14.50/month |
| Crisp | Yes | Yes | Flat | About €25/month |
| Tidio | Yes (Lyro add-on) | Yes | Per seat + AI add-on | About €29/month |
| Gorgias | Yes | Yes | Per AI interaction | From $10/month |
| Zendesk | Yes | Yes | Per agent + per resolution | About €55/agent |
| Intercom | Yes (Fin) | Yes | Per seat + per resolution | About €74/seat |
How to read that table without getting caught
- Flat versus metered. Flat pricing is predictable and suits small volumes. Per-resolution pricing punishes a busy month, and a viral day can double your bill.
- Seats multiply. Anything priced per seat costs five times as much for a team of five. Check before you hire.
- “Live chat” sometimes means an inbox, not a chat desk. Ask whether your team can reply in real time, whether they see the whole conversation, and what happens out of hours.
What we built, and what we did not
Our own plan is $14.50 a month, flat, and it covers the chatbot on your website, WhatsApp, Telegram, Instagram and Messenger, with handover to your team when a customer asks for a person, plus our AI calling agent, social suite and AI writer in the same price.
What we are not: a full help desk with ticket queues, SLAs and agent performance dashboards like Zendesk or Intercom. If you have a support department that lives in a ticketing system all day, buy one of those. Where we fit is the small business that wants fast answers on every channel, a human able to step in, and one bill instead of four.
Can I integrate chatbot platforms with existing live chat systems?
Yes, usually. There are three ways: a native integration built by one of the vendors, a direct API connection built by a developer, or a middleman app like Zapier. Native is best because the handover carries the whole conversation across. Before you commit, check that the chat history transfers, that the customer stays in the same chat window, and that your agents get a notification fast enough to matter.
| Way to connect | How it works | Good | Bad |
|---|---|---|---|
| Native integration | One vendor built support for the other | Minutes to set up, handover keeps the history | Only for popular pairings |
| API | A developer connects them directly | Exactly what you want | Needs a developer and upkeep |
| Middleman (Zapier, Make) | A third app passes messages between them | No coding | Monthly cost, delays, breaks quietly |
The six things to check before you sign
- Does the chat history transfer? If your agent starts blind, the customer has to repeat everything, which is the fastest way to annoy them.
- Does the customer stay in the same window? Being bounced into a different widget feels broken.
- How does the agent get told? Desktop alert, email, mobile push. An email that arrives in an hour is useless.
- What happens outside working hours? The bot should say plainly when a human will reply, and take an email address.
- Can the bot pass a summary, not just a transcript? A one-line summary saves your agent a minute on every chat.
- Who owns the conversation data? Ask where it is stored and for how long, especially if customers share order or payment details.
The simplest option many businesses miss: one platform that does both. You skip every question above, because there is nothing to join together.
The hybrid setup that actually works
Seven steps, and you can do it this week.
- Write your top 20 questions and real answers. This decides how well the bot performs, more than the software does.
- Let the bot take everything first. It answers in a second, which beats a human answering in three minutes.
- Write your handover rules. Be specific. Here is a starting set:
| Trigger | Action |
|---|---|
| Customer asks for a person | Hand over immediately, no “are you sure?” |
| Words like refund, cancel, complaint, charged twice | Hand over |
| Anything about legal or health matters | Hand over |
| The bot has failed twice on the same question | Hand over |
| A known high-value customer | Hand over, and flag it |
- Always show a way to reach a human. A hidden escape route is what makes people write angry reviews.
- Pass the history with the chat. Nobody should have to explain twice.
- Say what happens out of hours. “Our team replies by 9am, and I have your email” is a fine answer. Silence is not.
- Read the handed-over chats every week. Each one is a gap in your help content. Fix the top ten a month and your resolution rate climbs from 45% to 65%.
Five mistakes that cost you customers
- Hiding the human. If there is no obvious way through, people stop buying and start posting.
- Judging the bot by containment. See the 72% versus 52% gap above.
- Letting the bot argue with an angry customer. It will lose, publicly.
- Buying per-resolution pricing without checking your volume. A busy month becomes a nasty invoice.
- Never updating the answers. This is why most bots stall at 30–45% and get blamed for being “useless”.
How we compared the numbers
We took platform prices from vendors’ published plans and independent 2026 pricing breakdowns in September 2026, and the performance figures from sources that publish across many customers rather than single case studies. We have flagged where numbers come from companies that sell chat software, because the conversion figures in particular are vendor-published and should be read as directional. Where sources disagreed we give a range instead of the prettier number. We make one of the platforms listed here, and we have said plainly what it does not do.
Frequently asked questions
What is the difference between a chatbot and live chat?
A chatbot is software that answers on its own, instantly, at any hour. Live chat is the same chat window with a real person typing. The chatbot is cheaper and faster; the person is better at judgement, apologies and anything unusual.
Which is cheaper, a chatbot or live chat?
A chatbot, by a long way. Automated conversations cost roughly $0.50 to $1.50 each, while a routine live chat handled by a person costs $10 to $25, because wages are about 94% of live chat’s total cost. Three agents covering 24/7 run $7,500 to $12,900 a month.
Do customers prefer chatbots or humans?
They prefer whichever solves the problem. For order status and opening hours, instant bot answers win. For anything about money or mistakes, people want a person, and Drift’s research found human replies rated about four times better than bot-only replies.
Should I use both a chatbot and live chat?
Yes, for most businesses. Let the bot take every chat first and resolve the repetitive 40–60%, and hand over the moment a customer asks for a person or mentions refunds, cancellations or complaints. That keeps costs down without losing the conversations that matter.
When should a chatbot hand over to a human?
Immediately when the customer asks; when the message mentions refunds, billing, cancellation or a complaint; when the subject is legal or health related; when the bot has failed twice on the same question; and when the customer is a high-value account.
What percentage of chats can a chatbot handle?
Expect 40–60% resolved in the first months and over 60% after six to twelve months of improving your help content. Well-integrated systems keep 70–90% of chats from reaching a human, but containment is not the same as resolution — Ada reported 72% containment against 52% actual resolution on the same conversations.
Does live chat really increase sales?
Vendor-published figures suggest about a 20% conversion lift, visitors who chat being 2.8 times more likely to buy and spending around 60% more, and cart abandonment falling 25–30% with proactive chat. These come from companies selling chat software, so treat them as a direction rather than a guarantee.
Can one platform do both chatbot and live chat?
Yes. Markleyo, Crisp, Tidio, Gorgias, Zendesk and Intercom all offer automated chat with a route to a human. Using one platform for both avoids the biggest hybrid problem, which is the chat history failing to transfer at handover.
The bottom line
Stop asking which is better. Ask which conversation you are looking at. Order status, prices and password resets belong to the chatbot, which answers in a second and costs pennies. Money, anger and anything unusual belong to a person, who costs more and earns it back. Get the handover rules right, always leave an obvious door to a human, and read the handed-over chats every week — that habit, not the software you pick, is what turns a mediocre setup into a good one.
Want both without two subscriptions? We built Markleyo for small businesses that need instant answers and a real person on standby. One plan covers our AI chatbot on your website, WhatsApp, Telegram, Instagram and Messenger, with handover to your team when a customer asks, plus our AI calling agent, social suite and AI writer. Start your free Markleyo trial from $14.50/month and test it on your own 20 questions.
