How Much Does an AI Chatbot Cost in the UK?
Most UK businesses pay between £0 and £15,000 to build a chatbot, then £0 to £800 a month to run it. A simple rule-based bot can cost nothing on a free tier. A GPT-powered bot trained on your content typically costs £2,000–£8,000 to build and £50–£500 a month. A fully custom bot runs £8,000–£30,000 or more.
The spread is huge because "chatbot" covers three very different things. The cost depends on how the bot thinks, what it connects to, and how much traffic it handles. This guide breaks the price down by type, separates the build cost from the monthly running cost, and explains the API token charges most articles skip. If you want a partner to scope it, our UK AI automation service handles the build and the GDPR side together.
Why Does Chatbot Pricing Vary So Much?
A chatbot has no fixed unit cost. A pizza shop answering ten "are you open?" messages a night needs almost nothing. A national retailer handling thousands of order queries, refunds and stock checks needs a system that reads its database, follows business rules and stays compliant. Those two jobs share a name but nothing else.
Three things move the number most. First, the brain: a scripted menu of buttons costs a fraction of a large language model that understands free text. Second, the integrations: connecting to your booking system, CRM or Shopify adds development time. Third, the volume: every conversation a large language model handles costs money in API tokens, so high traffic means a higher monthly bill. Price the bot against the job, not a headline figure.
Cost by Chatbot Type
UK chatbots fall into three tiers. Knowing which tier you actually need is the single biggest factor in your budget, so it is worth understanding the difference before you ask anyone for a quote.
What Is a Rule-Based Chatbot and What Does It Cost?
A rule-based bot follows a fixed script. The visitor clicks buttons or picks from a menu, and the bot replies with a pre-written answer. It cannot understand free text or anything outside its script. Think of an FAQ with branching paths rather than a conversation.
This is the cheapest tier by far. Tools such as Tidio, Chatfuel or a basic WordPress plugin offer free tiers, with paid plans from £15–£40 a month. If you want an agency to map the conversation flows and connect them to your booking or contact forms, the build is usually £500–£2,500. Rule-based bots suit simple jobs: opening hours, directions, lead capture, "track my order". They never go off-script, which keeps them safe and predictable, but they frustrate anyone who asks a question the script does not cover.
What Does a GPT-Powered (LLM) Chatbot Cost?
An LLM bot uses a large language model such as OpenAI's GPT, Anthropic's Claude or Google's Gemini. You feed it your content (web pages, PDFs, a knowledge base), and it answers free-text questions in natural language. This is what most people mean in 2026 when they say "AI chatbot".
Build cost typically runs £2,000–£8,000 with a UK agency, covering content ingestion, prompt design, testing and the website widget. Off-the-shelf platforms such as Intercom Fin, Tidio Lyro or a custom build on the OpenAI API sit in this tier too. The monthly cost is £50–£500, made up of a platform fee plus API token charges that scale with how much people use it. This tier handles real customer support, qualifies leads and answers detailed product questions. It needs guardrails so it does not invent answers, which good prompt design and content grounding provide.
When Do You Need a Fully Custom Chatbot?
A fully custom bot is bespoke software. It connects to your live systems, your CRM, your inventory, your booking engine or your internal database, and it takes actions, not just answers questions. It can check a real order status, book an appointment, process a return or pull a customer record, all inside one conversation.
This tier costs £8,000–£30,000+ to build and £200–£800+ a month to run, depending on integrations and traffic. The cost reflects custom development, secure API connections, testing against your real data, and ongoing maintenance as your systems change. Fully custom bots suit larger UK SMBs, e-commerce operations and service businesses where the bot replaces measurable support hours. If a bot only answers questions, you do not need this tier. You need it when the bot does work.
Build Cost vs Monthly Running Cost
The biggest mistake UK buyers make is treating a chatbot as a one-off purchase. It has two costs: the build (one-time) and the running cost (every month, forever). For LLM bots, the running cost is the part most quotes hide.
What Goes Into the One-Time Build Cost?
The build is the setup. For a rule-based bot, it covers mapping the conversation flows, writing the answers and connecting the widget to your site. For an LLM bot, it adds ingesting your content, designing the prompt so the bot stays on-topic, testing against real questions, and styling the widget to match your brand. For a custom bot, it adds secure API connections to your systems and rounds of testing against live data.
Build is where you pay for expertise, not software. A cheap bot that hallucinates wrong prices or leaks data costs far more to fix than it saved. A proper build includes guardrails, fallback to a human, and a tested handoff. Expect a UK agency to spend more time on testing an LLM bot than on building it, because that is where the risk lives.
How Do API and Token Costs Work?
This is the part most articles skip. An LLM bot does not run for free after you build it. Every message costs money in "tokens", the unit the API bills by. Roughly, a token is three-quarters of a word, and you pay for both the question and the answer, plus the background content the bot reads to reply.
In 2026, mid-range models cost a few pence per conversation. A bot handling 1,000 conversations a month might cost £20–£80 in API tokens, plus a platform fee. A busy support bot handling 10,000 conversations could cost £200–£600. The exact figure depends on the model you choose and how much content it reads per reply. The lesson: a low-traffic bot is cheap to run, but token costs scale with success. Budget for them, and check whether your quote includes them or passes them to you separately.
A Cardiff dental practice we worked with wanted an AI bot to handle out-of-hours enquiries. They had been quoted £18,000 for a "fully custom" build. When we looked at the actual job, the bot only needed to answer treatment questions, give prices and capture a callback request. That is a grounded LLM bot, not custom software. We built it for a quarter of that, with running costs under £60 a month. The lesson: match the tier to the job. Most UK SMBs are sold the custom tier when the LLM tier does everything they need.
Pro tip: Before you get quotes, write down the ten questions your chatbot must answer and whether it needs to take any action (book, check an order, process a refund). If the answer to "take action" is no, you almost certainly need the LLM tier, not the custom tier, and you can halve your budget.
What Drives the Price
Two bots in the same tier can cost very different amounts. Five factors explain the gap, and understanding them helps you read a quote and spot padding.
Which Factors Add the Most Cost?
Integrations come first. A bot that only answers questions is cheap. A bot that connects to your booking system, CRM, Shopify or an internal database needs custom development for each connection, and each one adds testing time. Traffic comes next, because it drives your monthly token bill on any LLM bot. The more popular the bot, the higher the running cost.
Content quality matters more than people expect. A bot is only as good as the material it learns from. If your website content is thin or out of date, an agency spends time cleaning it before the bot can use it. Languages add cost, as does compliance work if you handle sensitive data. Finally, ongoing maintenance: an LLM bot needs monitoring so it does not drift, give wrong answers or break when you change your prices. A quote with no maintenance line is usually an incomplete quote.
Should You Build, Buy a Platform, or Use an Agency?
You have three routes. A DIY platform such as Tidio, Intercom or Chatbase lets you set up a bot yourself for a monthly fee, which suits simple needs and tight budgets. Building directly on the OpenAI or Anthropic API gives full control but needs developer time, so it only makes sense if you have the skills in-house. An agency build costs more upfront but handles the prompt design, integrations, GDPR and testing, which is where bots usually go wrong.
The right route depends on stakes. If the bot just deflects FAQs, a platform is fine. If it represents your brand to customers, touches personal data or connects to live systems, the agency route protects you. We often start UK clients on a platform to prove the use case, then move them to a proper build once the volume justifies it. There is no shame in starting small.
| Chatbot Type | Build Cost | Monthly Running Cost | Best For |
|---|---|---|---|
| Rule-based | £0 – £2,500 | £0 – £40 | FAQs, hours, lead capture |
| AI / LLM (GPT-powered) | £2,000 – £8,000 | £50 – £500 | Real support, product Q&A |
| Fully custom | £8,000 – £30,000+ | £200 – £800+ | System integrations, actions |
| DIY platform | £0 – £300 setup | £15 – £150 | Testing a use case first |
GDPR and Data Rules for UK Chatbots
A chatbot processes personal data the moment someone types their name, email or order number. That puts it squarely under UK GDPR and the Data Protection Act 2018, enforced by the Information Commissioner's Office. Skipping this is the most expensive corner you can cut.
What Are the Key UK Compliance Rules?
You must tell users what the bot does with their data, which means a clear privacy notice linked from the chat window. The ICO's UK GDPR guidance sets out the principles: a lawful basis for processing, data minimisation, and transparency. If your bot uses a US-based model provider, you also handle an international data transfer, which needs the right safeguards under the ICO's international transfers rules.
The ICO can fine up to £17.5 million or 4% of global annual turnover for serious breaches, the higher of the two. The ICO's own guidance on AI and data protection is the document to read before you build anything that uses customer data. For broader context on how the UK is approaching the technology, the government's pro-innovation approach to AI regulation on GOV.UK is worth a look.
How Do You Keep a Chatbot Compliant in Practice?
Start with a chat that does not store more than it needs. Do not log full conversations forever, and never feed customer data into a model that trains on your inputs without a contract that forbids it. Add a clear notice that the user is talking to a bot, not a person, and offer an easy route to a human. Set a retention period and delete transcripts after it.
If the bot makes decisions that significantly affect people, UK GDPR gives them rights around automated decision-making, so build a human review step. Keep a record of what data the bot collects and why. A good UK build bakes these in from the start rather than bolting them on later. Getting this right is not optional, and it is cheaper than a reprimand or a fine.
Common Mistakes to Avoid
- Buying the custom tier for an FAQ job — most UK SMBs need a grounded LLM bot, not £20,000 of bespoke software.
- Ignoring monthly token costs — an LLM bot is never a one-off purchase; running costs scale with traffic.
- No human fallback — a bot that traps frustrated customers in a loop damages the brand it was meant to help.
- Letting the bot invent answers — without content grounding and guardrails, an LLM will confidently give wrong prices.
- Skipping the privacy notice — the bot processes personal data the moment someone types an email, so UK GDPR applies.
- Feeding customer data to a training model — check the provider's terms; some use your inputs to train unless you opt out.
- No maintenance budget — bots drift, prices change, and an unmonitored bot slowly turns into a liability.
6 Frequently Asked Questions
Yes, for simple needs. Platforms such as Tidio, Chatfuel and several WordPress plugins offer free tiers that cover a rule-based bot answering hours, directions and basic FAQs. You can also build a small LLM bot on a free trial of an AI platform. The catch is limits: free tiers cap conversations, restrict branding and rarely include the integrations or GDPR controls a real business needs. A free bot is a fine way to test whether customers will use one. Once it handles real volume or personal data, you outgrow the free tier and move to a paid plan or a proper build. Treat free as a trial, not a finish line.
You pay OpenAI per token through its API, not a flat monthly fee, when you embed a model in your own bot. In 2026, mid-range models cost a few pence per conversation, so a bot handling 1,000 conversations a month might cost £20–£80 in tokens. A busy support bot handling 10,000 could cost £200–£600. The price depends on the model you pick and how much content it reads to answer each question. This is separate from a ChatGPT Plus subscription, which is for personal use, not powering a website bot. When you compare quotes, check whether the agency includes API costs in your monthly fee or bills them on top, because the difference adds up fast.
A rule-based bot follows a fixed script. The visitor clicks buttons or picks menu options, and the bot replies with pre-written answers. It cannot understand anything outside its script, so it is predictable but limited. An AI or LLM bot uses a large language model to understand free text and answer in natural language, drawing on content you give it. It handles questions you never scripted. The trade-off is cost and control: rule-based bots are cheap and never go off-script, while AI bots are more capable but need guardrails so they stay accurate. Most UK businesses start with rule-based for simple jobs and move to an LLM bot when customers expect real conversation.
A rule-based bot on a platform can go live in a day or two. An agency-built LLM bot grounded on your content usually takes two to four weeks, covering content ingestion, prompt design, testing and the website widget. A fully custom bot with system integrations takes six to twelve weeks or more, because each connection to your CRM, booking system or database needs building and testing against live data. The testing phase matters most on AI bots, since that is where you catch wrong answers before customers do. Rushing it is the false economy. A bot that goes live in three days but gives wrong prices costs more in lost trust than the time you saved.
It is worth it when the bot saves measurable hours or captures leads you would otherwise lose. A bot that answers out-of-hours enquiries, qualifies leads or deflects repetitive support tickets pays for itself quickly. A team spending several hours a day on the same questions has a clear case. It is not worth it when traffic is low, queries are varied and complex, or a quick contact form would do the same job. The honest test is volume: count how many repetitive enquiries you handle each week. If the number is high and the questions are similar, a grounded LLM bot at £50–£500 a month usually returns more than it costs within a quarter.
Yes. The moment a chatbot processes personal data, such as a name, email or order number, UK GDPR and the Data Protection Act 2018 apply, enforced by the ICO. You need a lawful basis, a clear privacy notice linked from the chat window, data minimisation and a sensible retention period. If your model provider is based outside the UK, you also handle an international data transfer with the right safeguards. Tell users they are talking to a bot, offer a route to a human, and never feed customer data into a model that trains on your inputs without a contract that forbids it. The ICO can fine up to £17.5 million or 4% of global turnover, so build compliance in from the start.
Thinking about adding an AI chatbot to your site? At Cambria Digital we have delivered 100+ UK projects, and we scope chatbots to the job, not the hype, so you do not overpay for a tier you do not need. Book a free discovery call and we will review your use case in 30 minutes, or read more about our UK AI automation and chatbot service to see how we build them with GDPR baked in. No obligation, reply within 1 business day.