Chatbot Pricing: What a Chatbot Really Costs and How to Calculate Its ROI

Chatbot pricing works on three models: a flat monthly plan, a price per user seat, or a fee for each conversation the AI resolves, and most vendors now combine two of them. Entry plans start around $20 per month, while AI agents add usage fees such as $0.50 or $0.99 per resolution, so the real cost depends on your traffic and on what the chatbot must do.

This article is part of our guide to the AI agent for website visitors. Here we look at the money side: how vendors price chatbots, what they charge today, the costs that do not appear on pricing pages, and how to calculate whether a chatbot pays for itself on a B2B website.

How is chatbot pricing structured?

Pricing pages look different from one vendor to the next, but they are built from the same three blocks.

Flat monthly plans

You pay a fixed amount per month for a plan with limits: number of conversations, visitors reached, seats or features. This is the model of most small business chatbots. It is easy to budget, but the limits matter more than the headline price.

Per seat pricing

You pay for each person on your team who uses the inbox. This model comes from live chat and help desk software, where people answer the chats. It makes sense when your team replies to conversations, and much less when the bot does most of the work.

Usage pricing: per conversation or per resolution

AI agents increasingly charge for what the AI does. Some vendors count conversations, others count resolutions, which are conversations the AI handled without a person taking over. The definition changes from vendor to vendor, so read it before comparing numbers.

What do chatbot vendors charge in 2026?

Here are public prices from vendor pricing pages, checked in November 2026. They change often, so verify them before you buy.

• Tidio. A free plan with 50 billable conversations, then Starter at $24.17 per month with 100 billable conversations and Growth from $49.17 per month. The Lyro AI agent is an add-on starting at $32.50 per month for 50 Lyro AI conversations. The page offers monthly or annual billing.
• ChatBot.com. Essential costs $19 per user per month billed yearly, or $25 billed monthly, with 10 AI resolutions per month. Growth costs $79 per user per month billed yearly, or $99 monthly, with 200 resolutions. Extra packages of 50 resolutions cost $49.50.
• Intercom. Seats range from $29 to $132 per seat per month depending on the plan, and the Fin AI agent costs from $0.99 per outcome, charged once per conversation. Fin can also be bought alone, without seats, with a minimum monthly commitment.
• HubSpot. The rule-based chatbot builder is free. The AI Customer Agent requires a Professional or Enterprise edition and costs $0.50 per messaging resolution, paid with HubSpot Credits. We detail what it does for sales in our article on the HubSpot chatbot.

Two things stand out. First, the cheapest plans cap AI usage tightly: 10 resolutions per month is a test, not a deployment. Second, usage fees grow with your success. The more visitors the AI helps, the more you pay, which is fair, but it has to be modeled.

Is a chatbot free?

Some are, within limits. Tidio has a free plan and HubSpot's chatbot builder is free. Free plans are useful to test a scripted flow on a few pages: a welcome message, three qualifying questions, a meeting link.

They stop being enough when you need AI answers on open questions, branching by company size or use case, or more than a few dozen conversations a month. At that point, the question is not whether to pay but which pricing model fits your traffic.

What costs are not on the pricing page?

The subscription is the visible part. On a B2B website, these costs often weigh more.

Setup and content

A scripted chatbot needs someone to write every path. An AI chatbot needs good source content: product pages, documentation, pricing logic, answers to objections. If that content is outdated, the bot will repeat it.

Upkeep

Every change to your offer, integrations or packaging has to reach the bot. With a rule-based bot, that means editing flows. With an AI agent, it means keeping the pages it learns from current.

Review time

Someone should read transcripts every week, at least during the first months. That is time from marketing or sales, and it is the best way to find questions the bot handles badly.

Integrations and routing

Booking meetings with the right rep, passing context to sales and logging conversations take setup work, and sometimes a higher plan.

Custom builds

A chatbot built in house has no license fee, but you pay for development, hosting, model usage and maintenance. It makes sense for unusual integrations or strict data rules, if a team can maintain it after launch.

How do you calculate chatbot ROI?

Chatbot ROI compares what the chatbot adds with what it costs. On a B2B website, what it adds is usually meetings or qualified leads that would not have happened otherwise.

The calculation has four steps.

1. Start from your current baseline. Monthly visitors on the pages where the chatbot will run, and your current visitor-to-lead rate. For reference, First Page Sage puts the average traffic-to-lead conversion rate of B2B SaaS websites at 2.5%.
2. Estimate the extra leads or meetings. How many visitors start a conversation, and how many of those end with a meeting booked or contact details left.
3. Turn them into revenue. Multiply the extra meetings by your meeting-to-deal rate and your average deal value.
4. Subtract the full monthly cost. Subscription, usage fees at your expected volume, and the hours your team spends on setup and review.

Here is an example with assumed numbers. A site gets 4,000 visitors a month on its product and pricing pages. The chatbot leads to 8 extra meetings a month. One meeting in five becomes a deal, worth $12,000 on average. That is about $19,200 of new revenue per month, against a monthly cost you can now compare line by line. If the same chatbot produced 2 meetings and a deal every three months, the answer would be very different.

You can run your own numbers with our chatbot ROI calculator, which compares what a conversation adds to a form.

Measure, do not assume

Vendor case studies are not your traffic. Run the chatbot on a set of key pages for at least a month, keep comparable pages without it if you can, and count meetings booked, not conversations started. A chatbot that chats a lot but books nothing has a cost and no return.

Which pricing model fits a B2B website?

The right model depends on who does the work in the conversation.

• Your team answers most chats. Per seat pricing is logical: you pay for the people who reply.
• The bot follows a short script. A flat plan or a free rule-based builder is enough. We compare scripted bots and AI agents in AI agent vs chatbot.
• An AI answers open questions and books meetings. Usage pricing is common. Estimate your monthly conversations and check the cost at two or three times that volume, since good results bring more conversations.

Whatever the model, the cost per meeting booked is the number to watch. It ties chatbot pricing to the goal of a B2B website. If your site converts poorly overall, start with our guide to improve your B2B website conversion rate.

A concrete example: two visitors, two outcomes

A visitor on a SaaS pricing page types: "Do you charge per user or per company?"

A scripted bot with no path for that question replies with its default flow: "Thanks! What is your company size?" The visitor's question is ignored. The conversation counts in the plan's quota, but it produces nothing.

An AI agent such as Nudge answers from the pricing page, cites that page, and then offers a short call to see what fits the visitor's team. The same conversation now has a chance of producing a meeting, which is the only number that justifies the chatbot's cost.

That is the job of Nudge's Website agent: it answers during the visit and proposes a slot in the conversation.

How Nudge handles website conversations

Nudge's Website agent answers your visitors before they leave. It is trained on your website and your documents, and every answer cites the page it came from, so visitors can check it. It answers while the visitor is reading, including detailed questions that no script would have planned. When the visitor is ready, it books the meeting in the conversation, without a form. It installs on WordPress, Webflow, Framer or Shopify, or through Google Tag Manager, so you can test it on a few key pages and measure the meetings it books against your current setup. See how it works on the Website agent page, or ask it a question directly on our site.