How Much Does an AI Chatbot Cost

What Determines the Cost Per Message

The biggest factor is which AI model powers your chatbot. Each model has different pricing based on how much it costs to run. Here is what typical chatbot messages cost on common models:

• GPT-5-nano: 1 to 2 credits per response. The cheapest option, good for simple FAQ answers and short replies.
• GPT-4.1-mini: 2 to 4 credits per response. The best balance of quality and cost for most customer support chatbots.
• Claude Sonnet: 3 to 6 credits per response. Strong at nuanced conversations and following complex instructions.
• GPT-5.2 (reasoning): 10 to 15 credits per response. Best for complex analysis and multi-step reasoning, usually overkill for support chat.
• Claude Opus: 8 to 15 credits per response. Most capable for complex tasks, rarely needed for standard chatbot use.

The actual credit count per message depends on the length of the conversation (longer histories use more tokens) and how much knowledge base content the chatbot retrieves to formulate its answer. Short FAQ answers cost less than detailed technical explanations.

Knowledge Base Costs

Training your chatbot on your own documents is a one-time cost per document. The platform creates vector embeddings from your content at 3 credits per chunk. A chunk is roughly 200 to 500 words of text.

For example, uploading a 20-page product manual that produces 40 chunks costs 120 credits ($0.12). You pay this once when uploading. After that, the chatbot searches these embeddings for free, you only pay for the AI model response that uses the retrieved information.

Monthly Cost Examples

Small Business (100 conversations/month)

A local service business with a chatbot answering questions about hours, pricing, and services. Using GPT-4.1-mini, averaging 5 messages per conversation: roughly 500 to 2,000 credits per month, which is $0.50 to $2.00.

Medium Business (500 conversations/month)

An e-commerce store with a chatbot handling product questions, order status, and return policy inquiries. Using GPT-4.1-mini with a larger knowledge base: roughly 2,500 to 10,000 credits per month, which is $2.50 to $10.00.

High Volume (2,000+ conversations/month)

A SaaS company with a chatbot as the first line of customer support. Using GPT-4.1-mini for most queries with Claude Sonnet for complex technical questions: roughly 10,000 to 60,000 credits per month, which is $10 to $60. Compare this to even a single part-time support agent at $1,500+/month.

No Monthly Subscription

Unlike most chatbot platforms that charge $29 to $499 per month regardless of usage, this platform uses pure pay-per-use pricing. If your chatbot handles zero conversations in a given month, you pay nothing. If you get a spike in traffic, you pay only for what was used. Credits never expire and there are no per-chatbot fees, so you can create as many chatbots as you need.

Comparing to Alternatives

Traditional chatbot platforms like Intercom, Drift, or Zendesk charge $50 to $500+ per month plus per-resolution fees. A dedicated customer support agent costs $2,000 to $4,000 per month in salary. An outsourced support service charges $8 to $25 per conversation. At 500 conversations per month, an AI chatbot on this platform costs roughly 1% of what a human agent costs and 5 to 10% of what most chatbot subscriptions charge.

The biggest cost difference appears at scale. A human agent's cost is mostly fixed (salary), while a chatbot's cost scales linearly with usage. But even at high volumes, the per-conversation cost of AI chat remains a fraction of any human alternative.

How to Keep Costs Low

• Start with GPT-4.1-mini or GPT-5-nano for straightforward FAQ chatbots
• Write a focused system prompt that keeps responses concise
• Upload well-organized training data so the chatbot finds answers quickly
• Set a maximum conversation length to prevent runaway token usage
• Use a cheaper model for simple questions and reserve expensive models for complex tasks