You asked two or three people what an AI chatbot for your website would cost, and the answers did not line up. One quote is a small monthly subscription. Another is a setup fee plus a monthly charge plus “usage”. A third will not give a figure until after a call. It is tempting to pick the cheapest and move on.
Before you do, it helps to know what affects AI chatbot cost, because the quotes are often pricing different jobs. Once you can see the parts, you can compare like with like, and you can decide which parts you actually need.
Why chatbot quotes vary so much
A chatbot is not one product. It is a stack of separate things, and each person quoting may be including a different slice of that stack:
- the chat platform itself (the widget, inbox and settings),
- the AI model that writes the replies,
- the information the bot answers from,
- the conversation design: what it asks, when it hands over, what it refuses,
- connections to other tools, like your email, a spreadsheet, a CRM or WhatsApp,
- and the upkeep after launch.
A low quote usually covers the first item and leaves the rest to you. A higher one may include all six. Neither is wrong, but they are not the same purchase. Our page on an ai chatbot for business website setup shows how these pieces fit together as one job; the sections below take them one at a time.
What affects AI chatbot cost, part by part
1. How the platform charges
Chatbot platforms use a few different billing models, and the model matters more than the headline figure. Common ones are:
- Per seat: you pay for each team member who logs in to the inbox.
- Per conversation or per resolution: you pay by how many chats the bot handles.
- Usage credits: AI replies draw from a pool of credits or messages each month.
- Feature tiers: AI answers, removing the vendor’s branding, or extra channels may only appear on higher plans.
Check the vendor’s own pricing page for current figures, and read what happens when you go over a limit. Depending on the tool, the bot may stop replying, drop to a basic mode, or bill the extra.
2. Your traffic and question volume
Under any usage-based model, more chats means a bigger bill. A site with modest traffic and a site running paid ads to a landing page can see very different monthly totals on the same plan. Look at your analytics before choosing a plan, not after.
3. Building your own versus renting a ready tool
A ready-made chatbot service bundles the AI inside its subscription. A custom build calls an AI model’s API directly, which means you pay the model provider for usage, separately from whoever builds and hosts the bot. Custom gives more control over behaviour and data; it also adds hosting, monitoring and a developer who understands it. For most small sites, renting is the sensible starting point.
4. The information the bot answers from
This is the part people underestimate. A bot is only as good as what it reads. If your website has outdated prices on old pages, contradicting policies, or simply does not say things customers ask about, someone has to fix that before the bot goes live. That work is writing and checking, not technology, and it can take longer than the technical setup. We cover the process in a guide to preparing your website content for a chatbot.
5. Conversation design and guardrails
Deciding what the bot should ask, which topics it must never answer, how it says “I don’t know”, and when it hands a visitor to a person takes thought and testing. A quick setup skips this. A careful one spends real time on it. The AI agent setup service page explains why those boundaries matter so much.
6. Integrations
Each connection adds setup and testing time:
- sending leads to email or a Google Sheet (simpler),
- creating records in a CRM (moderate, depends on the CRM),
- reading live data such as order status or appointment slots (more involved),
- running the bot on WhatsApp as well as the website.
WhatsApp deserves its own note. Automating WhatsApp means the WhatsApp Business Platform (the API), which Meta charges for separately from any chatbot tool, under its own published pricing. The WhatsApp app versus API comparison explains when you need it.
7. Languages
English-only is the simplest case. Adding Hindi, Hinglish or a regional language means testing replies in each language with real phrasing, and sometimes writing separate answer content. Quality varies by language, so budget testing time.
8. Upkeep after launch
Prices change, services change, staff change. Somebody needs to read transcripts, correct wrong answers and update the information. You can do this yourself or pay someone, but it is never zero.
One-time costs and recurring costs
Splitting the parts this way makes quotes much easier to compare.
| Usually one-time | Usually recurring |
|---|---|
| Choosing and configuring the platform | Platform subscription |
| Cleaning up and writing answer content | AI usage, if billed separately |
| Conversation flow and handover design | WhatsApp API charges, if used |
| Integrations and testing | Transcript review and content updates |
| Installing the widget on your site | Changes when your business changes |
A useful habit is to estimate a full year: the one-time setup plus twelve months of recurring costs at your expected volume. That figure is the honest comparison, not the first month.
Chatbot cost factors you control
Several of the drivers above are choices, not fixed facts. You can keep the bill sensible by deciding them deliberately:
- Start with one channel. Website first, WhatsApp later if the website bot proves useful.
- Start with one job. “Answer the common questions and take a phone number” is a much smaller build than “answer everything, book appointments and check orders”.
- Write your own FAQ draft. Nobody knows your customers’ questions better than you. A plain document of questions and correct answers saves setup time.
- Limit seats. If the platform charges per seat, not every staff member needs a login.
- Set usage alerts. Where the platform allows it, get notified before you hit a limit.
Questions that make two quotes comparable
Send the same list to everyone you are considering:
- Which platform will you use, and whose account is it in? (It should be yours.)
- Is the platform subscription included in your quote, or paid by me directly?
- How is AI usage billed, and what happens if we exceed it?
- Who writes and checks the content the bot answers from?
- What will the bot refuse to answer, and how does it hand over to us?
- Which integrations are included, and which are extra?
- How many rounds of testing and correction are included?
- After launch, who updates the bot when our details change, and how is that charged?
- If we stop working together, can we keep the bot and its data?
The answers to questions 2, 3 and 8 explain most of the differences you saw at the start.
Signs a quote is missing something
- No mention of who prepares the answer content.
- No mention of a fallback when the bot does not know.
- The platform account will be in the provider’s name, not yours.
- No testing on mobile, where many of your visitors will use it.
- A promise of a fixed number of leads or sales. Nobody can honestly promise that.
When a cheaper option is the right one
Not every business needs an AI chatbot. If you get a handful of questions a week and answer them quickly on WhatsApp, a clear FAQ page and a WhatsApp button may serve you better than any bot. If your products suit it, a chatbot for an online store has different cost drivers again, mostly around order data. Spend where the volume of repeat questions justifies it.
Estimating your own first year
You can rough out a realistic budget without any vendor’s help, using your own numbers:
- Count how many chat-worthy questions you get in a typical month across WhatsApp, phone, email and forms. That is your starting volume estimate.
- Pick two or three platforms and read their pricing pages with that volume in mind. Note the plan you would actually need, not the cheapest one listed.
- List the integrations you truly need at launch, and mark the rest as “later”.
- Decide who will review transcripts and update the bot, and how many hours a month that will take them. Their time is a cost even if no invoice arrives.
- Add the setup quote, if you are using someone to build it.
Put all of it on one sheet for twelve months. If the total looks large next to the enquiries the bot might handle, narrow the first version until it does not.
How we price our own setup
We do not publish a fixed figure, because the parts above differ so much from one business to another. Our setup is quoted on request, confirmed after a short discovery call about your questions, channels and integrations. Whoever you talk to, ask for the quote split into one-time costs, recurring costs and what you pay the platform directly. If that sounds useful, send us your list of common customer questions and we can talk through whether a bot is worth it for you.
Next in this series: Chatbot for Ecommerce Website: What It Should Handle
