An AI chatbot for your website costs $0 to around $1,000 a month on a ready-made SaaS or no-code plan. A custom chatbot built on an LLM API such as Claude is quoted from about $5,000 into six figures, and after that the token bill runs from a few dollars to around $100 a month. Market roundups put chatbot subscriptions at $9 to $999 a month, with custom builds from $5,000 up.
For most small and mid-sized sites you've got three real choices: a SaaS support widget, a no-code bot builder such as Chatling, or a custom build on a large language model (LLM) API that answers from your own content. They differ on setup cost, the monthly bill, how much upkeep they need and where your leads end up.
I've built the custom route myself. The chat on the home page of mohitgoyal.in answers questions as me, and it runs on the Anthropic Claude API. So I have a bias, but I'll keep this fair: all three options are below, with every price linked to where I found it.
What are the three ways to add an AI chatbot to a website?
You can use a SaaS support widget such as Intercom or Tidio, a no-code bot builder such as Chatling or Chatbase, or a custom build on an LLM API such as Claude. The big differences are who builds it and how you pay for it.
1. A SaaS support widget. Intercom (with its Fin AI agent) and Tidio (with Lyro) are full customer service platforms. The AI bot is one piece of a helpdesk that also has a shared inbox, tickets and handover to a human. You pay for seats, AI usage or both.
2. A no-code bot builder. Chatling and Chatbase let you point the bot at your website or upload documents, pick a model, style the widget and paste one script tag into your site. You pay a monthly plan with a fixed number of AI credits or messages.
3. A custom LLM build. A developer builds the chat interface into your site, connects it to a model API such as Claude, feeds it your content and writes the rules for what it can say and do. You pay once for the build, then for API usage and hosting.
How much does an AI chatbot for a website cost?
Ready-made AI chatbots cost $0 to about $1,000 a month depending on the vendor and your volume. Agencies quote custom builds at $15,000 to $150,000, and broader market roundups start from about $5,000. The figures below are list prices as published or reported in late 2026. Vendors change their plans all the time, so check the linked page before you commit.
| Option | Example | Setup | Monthly running cost |
|---|---|---|---|
| SaaS support widget | Intercom Fin | Configure it yourself | Seats from $29 per seat, plus $0.99 per AI outcome |
| SaaS support widget | Tidio + Lyro | Configure it yourself | Starter $29; Lyro from $39 for 50 AI conversations, up to $700 for 1,000 |
| No-code bot builder | Chatling | Configure it yourself | Free, then $25, $75 or $295 a month |
| No-code bot builder | Chatbase | Configure it yourself | $40, $150 or $500 a month on monthly billing |
| Custom LLM build | Claude or OpenAI API | Agency quotes of $15,000 to $150,000 | API tokens plus hosting |
The headline prices leave things out. Removing the "Powered by" badge is a $35 a month add-on on Chatling on every plan except the top one, and $99 a month on Chatbase. On a starter plan, that alone can double your bill.
Credits are not conversations
No-code builders bill in AI credits, and the credits each reply uses depend on the model you pick. On Chatling, a reply costs anywhere from 0.5 to 25 credits depending on the model, so the same plan can last a month or a day.
That credit spread comes from Chatling's own pricing FAQ, as summarised by eesel. This is how a cheap plan quietly turns into an expensive one.
What does 1,000 chatbot conversations a month actually cost?
At 1,000 conversations a month with about six AI replies each, my estimates run from about $6 for a custom build on Claude Haiku 5.5 to about $990 for Intercom Fin. That's before seats, hosting and build fees. Plan prices are hard to compare side by side because every vendor meters something different, so here's the worked example.
- Intercom Fin: if every conversation counts as an outcome, you pay about $990 in Fin fees at $0.99 each, before any seats.
- Tidio Lyro: the 1,000-conversation tier is $700 a month, on top of the Tidio plan.
- Chatling: 6,000 replies at 0.5 credits each is 3,000 credits. That fits inside the Growth plan's 6,000 credits at $75. Switch to a premium model, though, and you'd run out well before the month ends.
- Custom build on Claude: current API rates are $2 per million input tokens and $10 per million output tokens for Sonnet 5.5, and $0.10 and $0.50 for Haiku 5.5. Say each reply sends 8,000 tokens (instructions, your content and the chat history) and gets 300 back. That's roughly $114 a month on Sonnet and about $6 on Haiku, plus hosting.
- Intercom Fin990$
- Tidio Lyro700$
- Custom, Claude Sonnet114$
- Chatling Growth75$
- Custom, Claude Haiku6$
These are my estimates from published rates, not quotes, and they leave out seats, hosting and the build fee. Look at how each bill grows. Per-conversation pricing rises in step with your traffic. Token pricing also rises with traffic, but at a far lower cost per conversation. If your volume is low, a $25 to $75 no-code plan will beat a custom build for years. If it's high or climbing, the one-off cost of a custom build can pay for itself.
What drives the cost of a custom AI chatbot?
Mostly it's how much content the bot has to know and what it's allowed to do, then guardrails, cost controls, design and analytics. Agency quotes run from $15,000 to $150,000, with most SaaS and e-commerce builds at $15,000 to $80,000, and the same agency puts a focused MVP at 6 to 8 weeks. The range is wide because "chatbot" covers very different jobs. Here's what moves the price:
- How much content it needs to know. A 30-page marketing site can fit in the prompt. Thousands of help articles or product records need a search layer (retrieval) so the bot pulls in only the relevant parts.
- What it's allowed to do. Answering questions is cheap. Booking a call, checking an order or opening a support ticket means real integrations, and each custom API connection can cost $5,000 to $25,000 to build.
- Guardrails. A bot that speaks for your business needs limits. It has to know what it won't discuss and when to pass someone to a human, and it needs protection against people trying to trick it into saying something it shouldn't.
- Cost controls. Rate limits and per-visitor caps stop a bot, or a scraper, from running up your API bill.
- Design and analytics. A chat that matches your brand, logs conversations and shows which questions turn into leads is more work than a default widget.
Upkeep isn't free either. Industry estimates for yearly maintenance of a custom chatbot run from $1,000 at the low end to $20,000 or more, depending on how often your content changes and how many systems the bot touches.
What does running a custom Claude chatbot involve?
I built the chat on mohitgoyal.in on TanStack Start and React, with the Anthropic Claude API answering visitors' questions as me and Firestore storing the data. Visitors on mohitgoyal.in can ask about my projects, services or availability and get a straight answer instead of clicking through pages. I've written up the full approach in how to build an AI chat portfolio site.
What I learned applies to any business site:
- The content is the real work. The model is only as accurate as the facts you give it about the business. The ongoing job is keeping those facts complete and up to date. Tuning the model barely comes into it.
- Rules matter as much as the model. Tell the bot what it must never claim, like prices, numbers or promises that aren't in the content. That stops the most embarrassing answers before they happen.
- Cap usage from day one. My free AI tools on the same site are limited to 10 runs per tool per day. Limits like that keep the API bill predictable.
- You own all of it. No branding badge, no credit meter, and I can switch models when prices or quality change.
Which type of AI chatbot captures leads best?
All three can capture leads. SaaS and no-code chatbots offer CRM and Zapier integrations, often on higher tiers or for an extra fee. A custom chatbot gives you full control over which fields you collect, where each lead goes and how it's tracked, with no tier gating. On a business site, the bot's job is to turn questions into enquiries, so this matters. Here's how each option handles it:
| Lead capture | SaaS widget | No-code builder | Custom build |
|---|---|---|---|
| Ask for name and email in chat | Yes | Yes, usually on paid plans | Yes, designed to fit your flow |
| Where leads land | The vendor's inbox, plus native CRM integrations on most plans | The vendor's dashboard, plus integrations | Your CRM, a Google Sheet or email, directly |
| Hand over to a human | Built in | Varies by plan | Whatever you build (WhatsApp, email, booking link) |
| Tracking in GA4 | Partial | Partial | Full control of events |
If your sales team already lives in HubSpot, a custom chatbot can drop each lead straight in. It's the same kind of wiring I do for HubSpot forms on WordPress or when I send form entries to Google Sheets. With a SaaS widget, check which plan includes the CRM sync you need before you sign up.
That said, Chatling is one of the tools I've integrated, and for a lot of small sites it's a sensible first step. It's quick to set up and cheap at low volume. You can move to a custom build later, once you know what visitors actually ask.
Which AI chatbot option should you choose?
Go with a SaaS widget if you already run a helpdesk team, a no-code builder if you get fewer than a few hundred chats a month, and a custom build if your volume is high or the bot needs to do things rather than just answer.
- Pick a SaaS widget such as Intercom or Tidio if a support team already answers chats and you want AI to take the first pass inside the same inbox.
- Pick a no-code builder such as Chatling or Chatbase if you want an answer bot this week, your traffic is modest and your content is mostly public web pages.
- Pick a custom build if your volume is high, the bot needs to check data or trigger actions, you want leads in your own systems, or you'd rather not have a third-party badge and credit meter on your site.
Not sure a chatbot is even the best use of AI for your business? My free AI Use-Case Finder looks at how your business works and ranks the places where AI or automation would save the most time. Sometimes the honest answer is a better contact form, not a bot.
What should you ask before paying for an AI chatbot?
Ask how the vendor bills, which model answers, what happens at the plan limit, where conversations are stored and who owns the code and data. In practice that's six questions:
- What exactly counts as a billable conversation, resolution or credit?
- Which model will answer, and what does each reply cost on it?
- What happens when I hit the plan limit? Does the bot stop, or do I get billed for overage?
- Where are conversations stored, and can I export them? This matters for GDPR if you have UK or EU visitors.
- How does the bot refuse questions it can't answer from my content?
- For a custom build: who owns the code, the prompts and the API account, and what does monthly upkeep include?
Ready to put a chatbot on your site?
You probably have a rough idea of your chat volume and what you want the bot to do by now. Send me that, plus your site URL, through the contact page. I'll tell you straight whether a $25 plan will do the job or a custom build is worth it. If it is, I build custom chat and AI features as part of my web and mobile apps work, and I wire leads into HubSpot and GA4 through my landing pages with GA4, GTM and HubSpot service.



