# What counts as a conversation in chatbot pricing?

> Conversation, chat, Active Contact, outcome, credit, Lyro conversation: how each of nine chatbot vendors defines its billing unit, when it resets, what happens at the limit, and what to ask in writing.

Published: 18 September 2026 · Updated: 18 September 2026 · Prices verified: 18 September 2026 · By the chatbots.reviews team

Researched and drafted with AI assistance from the sources listed on this page. We have not built or tested bots on these platforms yet. Method: [How we review](https://chatbots.reviews/en-gb/how-we-review)

There is no standard meaning. Each vendor defines its own billing unit, and the definitions differ on things that move the bill: whether silence counts, whether chats handled by a person count, how long one session lasts, and whether one person on two channels counts twice. Read the definition before you read the price.

**The short version**

- Nine vendors, at least six different units: conversation, chat, Active Contact, outcome or resolution, credit, and Lyro conversation.
- Some count every chat. Some count only chats the AI settles. One counts people, not chats.
- Most allowances reset monthly and do not roll over. A few free allowances are one-time only.
- At the limit, bots stop, charge overage, or buy more for you. Find out which before launch.

## Each vendor's unit

| Product | Unit | What counts | At the limit |
| --- | --- | --- | --- |
| [Botpress](https://chatbots.reviews/en-gb/pricing/botpress) | Conversation | Any exchange with at least two messages from the end user in the billing month. Human-handled chats and emulator tests count. On voice, 3 minutes count as one. | Paid plans add a pack of 100 automatically at 95%. Free has no overage. |
| [Landbot](https://chatbots.reviews/en-gb/pricing/landbot) | Chat (and AI chat) | A session started by a button, keyword or trigger; the pricing table labels it "24 h, all channels". Chats that use the AI blocks are a second meter. | Overage at €0.05 per chat, within fair-use ceilings. |
| [Typebot](https://chatbots.reviews/en-gb/pricing/typebot) | Chat | Counted when a user starts a discussion, whatever the number of messages. One chat equals one row in the Results table. | Paid plans move to the next tier automatically. Free is not documented. |
| [Manychat](https://chatbots.reviews/en-gb/pricing/manychat) | Active Contact | A person who sent or received a message in the billing month, counted once however many messages. | Automations keep running; each extra contact is billed at the plan's rate unless you set a threshold. |
| [Intercom Fin](https://chatbots.reviews/en-gb/pricing/intercom-fin) | Outcome | A resolution, a Procedure handoff or a sales disqualification at US$0.99, or a qualification at US$9.99. Charged at most once per conversation. | No cap by default; optional alerts and a hard limit. |
| [HubSpot](https://chatbots.reviews/en-gb/pricing/hubspot-chatbot) | Resolved conversation | The AI agent shared a source or took an action, and no visitor-started handoff followed within 72 hours. The rule-based bot is not metered. | The agent stops taking new chats unless pay-as-you-go is on. |
| [Tidio Lyro](https://chatbots.reviews/en-gb/pricing/tidio-lyro) | Lyro conversation | A conversation the Lyro AI agent handles. Tidio's pages disagree on returning visitors (see below). Human-answered conversations and Flows visitors are separate meters. | Stops, or upgrades automatically at 95% if automatic upgrades are on. |
| [Voiceflow](https://chatbots.reviews/en-gb/pricing/voiceflow) | Credit | 1 credit per user message, 10 per phone minute, plus model tokens at per-model rates. | No automatic overage is published; top-ups are bought in the app. |
| [Chatfuel](https://chatbots.reviews/en-gb/pricing/chatfuel) | Credit (dollar wallet) | AI replies and paid WhatsApp messages draw from a dollar balance. Automation without AI is not metered. | At zero the bot stops replying until the wallet is refilled; automatic top-ups are optional. |

_From each vendor's pricing page and help centre as recorded in our product database. Each product name links to our full pricing page._

## The details that change the count

**Silence.** Intercom Fin bills an assumed resolution when a customer says nothing for 24 hours after Fin's last answer. HubSpot bills a visitor who reads a sourced answer and leaves. On a per-resolution meter, a visitor who gave up can look like a success.

**Who answered.** Botpress counts a chat a person handles the same as one the AI handles. Tidio's Starter and Growth plans meter conversations answered by a person, and its Lyro tiers meter AI conversations. The same word, conversation, means different things across plans at one vendor.

**Session length and returning visitors.** Landbot's allowance label implies one chat can last up to 24 hours. Typebot does not count a returning visitor whose session is remembered. Tidio's pricing FAQ says a customer who leaves and returns is still one conversation, while its help article on the Lyro limit says a thread closes after 15 idle minutes and the next reply counts again. Botpress counts each anonymous session separately, and a conversation that spans two months counts in both.

**People versus chats.** Manychat counts a person once a month, however much you talk. But sending a broadcast makes every recipient an Active Contact, and the documentation says the same person on two platforms may be counted twice.

## Resets, rollovers and one-time allowances

- **Monthly, no rollover** is the norm: Botpress, Landbot (monthly even on a yearly plan), Manychat (monthly on annual plans too), HubSpot credits and Chatfuel credits all reset or expire each cycle.
- **Unused paid extras expire too.** Botpress conversation packs lapse at month end, and Fin on another helpdesk loses unused outcomes from its monthly minimum of 50.
- **One exception carries over:** Botpress's automatic AI credit grants last up to 12 months.
- **Some free allowances never renew.** Tidio's Free plan includes 50 Lyro conversations once, and Voiceflow's Free plan has a one-time grant of 100 credits.

## Questions to ask a vendor in writing

Where a vendor's own pages disagree or say nothing, get the answer in an email or the contract. Ask:

1. What exactly starts a billable unit, and what ends it? Is there a time window?
2. Does a returning visitor start a new unit? After how long?
3. Do chats handled by a person, test chats and chats with no reply count?
4. Is silence after an answer billed as a resolution?
5. Does one person on two channels count once or twice?
6. When does the allowance reset, and does anything roll over?
7. What happens at the limit: hard stop, overage, or automatic upgrade? Can I turn it off, and can I set a spending cap?
8. At what levels do usage alerts go out, and to whom?
9. Are AI use, WhatsApp messages or voice minutes billed on a separate meter?

Then price a month of your own traffic under each answer. For how these units fit into whole pricing models, read [chatbot pricing models](https://chatbots.reviews/en-gb/guides/chatbot-pricing-models). For the difference between a scripted chat and an AI one, which many vendors meter separately, read [rule-based vs AI chatbots](https://chatbots.reviews/en-gb/guides/rule-based-vs-ai-chatbots). Our [containment rate guide](https://chatbots.reviews/en-gb/guides/containment-rate-explained) explains why a resolution count is not the same as a happy customer.

## Frequently asked questions

### What counts as a conversation in chatbot pricing?

It depends on the vendor, because there is no standard definition. Botpress counts any exchange with at least two user messages in a month, Typebot counts each discussion a user starts, Manychat counts each person once a month, and Intercom Fin counts only outcomes such as resolutions. Always read the vendor's written definition before comparing prices.

### Do chats handled by a human count towards a chatbot allowance?

At some vendors, yes. Botpress's pricing page says a chat a person handles counts the same as one the AI handles, and Tidio's Starter and Growth plans meter conversations answered by a person. Intercom Fin does not charge when Fin escalates by default or when the customer asks for a person.

### Is a customer who stops replying billed as resolved?

With Intercom Fin and HubSpot's Breeze Customer Agent, it can be. Fin bills an assumed resolution after a set period of silence following its last answer, and HubSpot bills a resolved conversation when the agent shared a source and the visitor did not ask for a person within its window. Ask any per-resolution vendor how it treats silence.

### Do unused chatbot conversations roll over to the next month?

Usually not. Botpress, Landbot, Manychat, HubSpot and Chatfuel allowances reset or expire each billing cycle, and Botpress's extra conversation packs expire at month end. Botpress's AI credit grants are an exception and can carry over for months.

> **Scored from public documentation, on a rubric you can read.** Scores come from five published criteria: pricing clarity, access, builder depth, reach and handoff, and data and trust. Every price and limit is stored once, in one product database, with a link to its source, so no two pages on this site can disagree. Rubric v1.0 · The reference-bot build test has not been run yet. Until it has, no page here says a product was tested. [Read the full method](https://chatbots.reviews/en-gb/how-we-review) · [The reference-bot test](https://chatbots.reviews/en-gb/reference-bot-test)

## Keep reading

- [What is the difference between a chatbot and an AI agent?](https://chatbots.reviews/en-gb/guides/chatbot-vs-ai-agent)
- [Rule-based or AI chatbot: which should you build?](https://chatbots.reviews/en-gb/guides/rule-based-vs-ai-chatbots)
- [How is chatbot software priced?](https://chatbots.reviews/en-gb/guides/chatbot-pricing-models)
- [How do you write a knowledge base for a chatbot?](https://chatbots.reviews/en-gb/guides/how-to-write-a-bot-knowledge-base)
- [Best chatbot builders](https://chatbots.reviews/en-gb/best/chatbot-builders)

---

Canonical: https://chatbots.reviews/en-gb/guides/what-counts-as-a-conversation