AI Chatbot Pricing Models Explained: Flat, Per Conversation and Per Resolution

Three models dominate this category, and they behave so differently under load that comparing headline prices between them is close to meaningless.

1. Flat monthly

You pay for a plan. It includes an allowance of messages or conversations, and the price does not move with traffic inside that allowance.

Good for: anyone who needs to budget. Small businesses, service businesses, agencies reselling.

Watch for: what happens past the allowance, and whether "message" means one side of the exchange or both.

2. Per conversation

You pay for each conversation the AI handles, sometimes with a bundle included.

Good for: low, stable volume, or piloting before committing.

Watch for: what counts as a conversation. Does a visitor returning two hours later start a new one? Does a single "hi" count?

3. Per resolution

You pay when the AI resolves a query without escalating to a human.

Good for: established support teams that can model deflection accurately and compare it directly against agent cost.

Watch for: the incentive shape. You pay more precisely when the product works better, which is coherent but uncomfortable, and definitions of "resolved" vary between vendors.

The question that exposes the difference

"What does my bill look like in my busiest month?"

On flat pricing, the same. On metered pricing, materially higher, and your busiest month is often the one you least want a surprise invoice in: a seasonal surge, a campaign that worked, a storm week.

Neither is dishonest. But if you are a small business, the ability to forecast is worth real money, and metered pricing takes it away.

How to actually compare

Take your realistic monthly conversation volume, then your peak month, not your average.
Price both scenarios on each model.
Add anything charged separately: extra bots, seats, training, branding removal.
Compare the peak figures, because that is the number that will hurt.

For reference

Published pricing in this category ranges widely: per-resolution models around a dollar per resolved query plus seat costs, per-conversation models under a dollar, and flat plans from around $20 to several hundred a month.

SpideyChat is flat: $9.99 and $29.99 tiers with a monthly allowance of AI tokens (1M and 5M, about 750 and 3,750 replies), plus a free plan. We publish it that way deliberately, because "what will this cost me in a good month" should have a boring answer.

Frequently asked questions

What is per-resolution pricing?

You pay each time the AI resolves a query without a human. It aligns cost with value for a support team, but it makes your bill unpredictable and it rises as the bot gets better at its job.

Which model is best for a small business?

Flat monthly, almost always, because you can budget against it. Metered models suit larger support operations that model volume carefully.

Do message limits count both sides of a conversation?

It varies by vendor and it matters a lot. Ask whether a message means a visitor message, a bot reply, or both, because the difference can double your effective usage.

Are there hidden costs?

Watch for training or ingestion charges, per-seat fees on top of usage, charges for extra bots, and overage rates that step up sharply past the bundle.