Two chatbots can both cost "$50 a month" and behave very differently once real traffic arrives. The reason is the pricing unit. One vendor bills by message, another by conversation, another by outcome and another by seat. Comparing prices without comparing the unit is how people end up with a bill they did not expect.
This guide explains the main models, how each shifts risk between you and the vendor, and a method for comparing them using your own traffic. Vendor figures below were read from public pricing pages on 3 November 2026 and may have changed, so confirm on their sites.
What Pricing Units Do Chatbot Vendors Use?
| Unit | What you pay for | Who carries the risk |
|---|---|---|
| Message or credit | Each AI reply or exchange consumes credits from a monthly allowance | You, if traffic grows; credits can run out |
| Conversation | Each chat session counts once, however many messages it contains | You, if many visitors start chats |
| Outcome or resolution | You pay when the AI resolves a conversation by the vendor's definition | Shared, but depends on how "resolved" is defined |
| Seat | You pay per team member who uses the inbox or dashboard | You, as your team grows |
| Plan limits | A fixed plan with caps on knowledge size, messages and seats | The vendor, until you hit a cap and must upgrade |
Many vendors combine two or more, such as a per-seat plan plus a per-outcome AI fee. Always list every unit in the bill, not just the headline price.
How Does Message or Credit Pricing Work?
A set allowance of messages or credits resets each month, and extra credits are sold on top. It is easy to understand, but cost depends on how many messages your visitors send, and some vendors weight credits by model or feature. Chatbase, for example, lists plans with 700, 4,000 and 15,000 monthly message credits, with extra credits sold separately, on its pricing page as of 3 November 2026.
How Does Conversation Pricing Work?
You pay per chat session. The definition matters enormously. Botpress counts any exchange with at least two messages from the end user in the billing month as a conversation, and it adds packs automatically once you reach 95 percent of your quota on paid plans, according to its pricing page. Tidio prices around billable conversations and counts its Lyro AI conversations separately.
Ask how a conversation is counted, whether human-handled conversations count, how repeat visits are treated and what happens when you reach the limit: a hard stop, or automatic extra packs.
How Does Outcome or Resolution Pricing Work?
You pay when the AI is judged to have resolved a conversation. Intercom's pricing page lists Fin from $0.99 per outcome, on top of seat plans if you use its helpdesk. This aligns cost with success, but the definition of an outcome decides your bill, and you may not control how often it is met.
Botpress has argued against this model, saying resolution pricing creates disputes over what "resolved" means. Whether you agree or not, ask any vendor using it exactly how an outcome is defined and measured.
How Does Seat Pricing Work?
You pay for each person with access to the inbox. It is predictable for a fixed team, but it can penalise growth. Check how many seats each plan includes and what extra seats cost. ConvoRail's plans include a set number of seats (Free 1, Starter 2, Growth 3, Scale 4), with extra seats available, and the real limits are Q&A pairs and monthly messages.
How Do You Compare Vendors With Your Own Numbers?
- 1Estimate monthly visitors who would open the chat, using a realistic share of your traffic.
- 2Estimate messages per conversation, from a trial or a small pilot.
- 3Count the people who would need inbox access.
- 4For each vendor, convert all three into that vendor's pricing unit.
- 5Add every add-on you would need: branding removal, extra storage, extra agents, voice and integrations.
- 6Price a busy month, not an average one, and see what happens above the limit.
Then ask the key questions: is usage capped or auto-billed, what is the contract length and can you cancel, can you export your data and what is the price after any introductory discount?
What Costs Hide in the Fine Print?
- Branding removal: some vendors charge extra to remove their badge.
- Overage packs: automatic extras can raise a bill without a new decision.
- Feature gating: integrations, voice or API access may sit on higher plans.
- Per-agent or storage add-ons: extra agents or knowledge storage billed separately.
- Introductory discounts: a first-year price that later rises.
For side-by-side facts on specific vendors, see our comparisons: ConvoRail vs Chatbase, ConvoRail vs Tidio, ConvoRail vs Intercom Fin and ConvoRail vs Botpress. They are written by ConvoRail, so check the cited pricing pages yourself.
Whatever you choose, calculate the cost of a bad answer too. A cheap chatbot that guesses can cost more than an expensive one that does not, as covered in the real cost of a bad chatbot.
Pro Tip: Before buying, ask each vendor for the cost of your busiest month in writing. The answer tells you far more about the real price than the headline number.
What Does a Worked Example Look Like? Hypothetical Numbers
Suppose 1,200 visitors open the chat in a month, and each conversation has about four visitor messages. That is 1,200 conversations and roughly 4,800 visitor messages, plus the replies. The same traffic produces very different units depending on how a vendor counts.
| Counting method | Result for this traffic | What to ask |
|---|---|---|
| Per conversation | 1,200 conversations | Does a conversation need two or more visitor messages to count? |
| Per message or credit | About 4,800 or more credits | Do replies, longer answers or advanced models use extra credits? |
| Per outcome | Depends on how many are resolved | Who decides what counts as resolved? |
| Per seat | Depends only on team size | What do extra seats and observers cost? |
The example shows why you must convert your own traffic into each vendor's unit. Do the same with a busy month and a quiet one, and ask what happens when you go over.
What Should You Ask Before You Sign?
- Is there a free trial or plan, and what does it not include?
- Is the price shown monthly or billed annually, and what is the cancellation rule?
- Can I export my knowledge base and conversations if I leave?
- What happens if I reach my limit: stop, or bill automatically?
Related reading: the real cost of a bad chatbot, build vs buy and security questions to ask vendors.
Conclusion
Chatbot pricing units differ, so headline prices are not comparable on their own. Identify the unit (message, conversation, outcome or seat), check how it is defined, convert your own traffic into each vendor's unit, add every add-on and price a busy month. Confirm any figure on the vendor's current pricing page.
Key takeaways
- Chatbot vendors bill per message, conversation, outcome, seat or plan limit.
- How a conversation or outcome is defined decides your bill.
- Find out whether usage is capped or auto-billed above the limit.
- Convert your own traffic into each vendor's unit before comparing.
- Vendor prices change, so verify on their pricing pages.
Quick Summary
AI chatbot pricing models include per message or credit, per conversation, per resolved outcome, per seat and fixed plan limits, and many vendors combine them. The unit and its definition decide the real cost, so convert your own traffic into each vendor's unit, add every add-on, check whether usage is capped or auto-billed and price a busy month. Vendor figures cited here were read from public pricing pages on 3 November 2026 and may have changed.
Want to see grounded answers on your own site?
Start free, add your Q&A pairs and paste one script tag. No credit card required.

Founder & CEO
Writes about grounded AI chatbots for websites: how they work, how to measure them and what to check before you buy.
Topics
Get new posts by email
Guides on AI chatbots for websites, straight to your inbox.
