How to build an AI chatbot for your website (2026)

Last updated October 2026

Adding an AI chatbot to your website used to require a development team, months of work, and a budget that ruled out most small businesses. That has changed. The current generation of chatbot platforms lets you train an AI on your own content and embed it on your site in an afternoon, often without writing a single line of code.

This guide walks through every step of the process: choosing the right platform, preparing your training content, building a knowledge base, integrating with your website, handling edge cases when the bot falls short, and measuring whether the whole thing is working. Whether you run an ecommerce shop, a SaaS product, or a services business, the principles are the same. The tools just vary depending on how much control you want.

If you are still deciding which platform to use, our best AI chatbot for websites comparison covers the full landscape. This guide assumes you are ready to build.

Why every website needs an AI chatbot in 2026

The shift from static support to conversational AI

Customer expectations have moved beyond FAQ pages. Visitors to your website expect instant answers, and they expect those answers to be specific to their question rather than a link to a help article they have to read themselves. AI chatbots trained on your own content can provide exactly this. They understand natural language questions, pull from your documentation and product information, and respond in seconds at any hour of the day.

The cost equation has flipped entirely. Running a support team around the clock is expensive. Even a modest chatbot deployment can handle 60 to 80 percent of incoming questions without human involvement, freeing your team to focus on complex issues that require judgement. For small businesses and solopreneurs, this means offering a level of support that previously required hiring staff. For larger teams, it means scaling support without scaling headcount at the same rate.

Search behaviour is changing too. Visitors increasingly prefer asking a question in a chat widget over browsing through navigation menus. A well-trained chatbot becomes an alternative interface to your entire website, surfacing product pages, pricing details, and policy information through conversation rather than clicks. This keeps visitors engaged longer and reduces bounce rates, both of which matter for SEO performance.

What a chatbot can do for different business types

Ecommerce stores benefit from product recommendations and order support. A chatbot trained on your product catalogue can help visitors find the right item, compare options, and answer questions about shipping, returns, and sizing. This reduces cart abandonment and takes pressure off your support inbox. For more on tools that complement this setup, see our guide to the best tools for ecommerce founders.

SaaS companies use chatbots to reduce churn and accelerate onboarding. When a new user gets stuck, a chatbot trained on your documentation can walk them through the solution immediately. This is faster than submitting a ticket and waiting for a reply, and it means fewer users abandon your product out of frustration during the critical first few days.

Service businesses and agencies can qualify leads automatically. A chatbot can ask the right questions, understand what the visitor needs, and either provide an answer or route them to a human with context already gathered. This saves time on both sides and ensures that when a team member does step in, they have the information they need to help.

Types of chatbots: rule-based, AI-powered, and hybrid

Rule-based chatbots

Rule-based chatbots follow predefined decision trees. You map out every possible conversation path in advance, and the bot follows the script. If a visitor asks something outside the tree, the bot either gives a generic fallback or gets stuck. These were the standard approach for years and still work well for very narrow use cases, such as booking appointments or routing enquiries to the right department.

The limitation is obvious: they cannot handle the unexpected. Real customer questions are messy, varied, and often phrased in ways you would never predict. A rule-based bot trained on 50 common questions will fail on the 51st. For most businesses in 2026, pure rule-based chatbots are no longer sufficient.

AI-powered chatbots

AI chatbots use large language models to understand and respond to natural language. Instead of following a script, they interpret the visitor's intent, search through your training content, and generate a relevant response. They can handle rephrased questions, follow-up queries, and topics you never explicitly programmed, as long as the information exists somewhere in their training data.

The quality depends entirely on the training content. An AI chatbot is only as good as the knowledge base behind it. Feed it comprehensive, well-organised documentation and it performs impressively. Give it scattered, outdated content and it will confidently provide wrong answers, which is worse than no answer at all. We will cover content preparation in detail later in this guide.

Hybrid approaches

Most modern chatbot platforms blend both methods. They use AI for understanding and generating responses but allow you to define specific flows for high-stakes interactions like purchases, cancellations, or sensitive support topics. This gives you the flexibility of AI with the predictability of rules where it matters most. Chatbase, for example, lets you train an AI on your content while also setting up specific response rules for particular topics.

Chatbase: the fastest path to a trained AI chatbot

Why Chatbase leads the no-code chatbot space

Chatbase has become the default recommendation for businesses that want a working AI chatbot without hiring developers. The platform lets you upload your website content, documents, and FAQs, then generates a chatbot that can answer questions based on that material. The entire process, from signup to a working chatbot embedded on your site, takes less than an hour for most businesses.

The training process is straightforward. You can point Chatbase at your website URL and it will crawl your pages automatically, or you can upload PDFs, Word documents, and text files directly. The platform processes this content and builds a knowledge base that the chatbot draws from when answering questions. You can also add specific Q&A pairs for questions where you want precise, controlled answers.

Setting up your first Chatbase chatbot

Start by gathering your core content. Before you create your chatbot, collect your most important pages: product descriptions, pricing information, FAQ answers, shipping and returns policies, and any documentation that customers regularly ask about. Having this ready before you begin saves time and produces better results from the first version.

Create your chatbot and add your data sources. Chatbase offers several ways to ingest content. The website crawler is the easiest starting point, as it pulls in your public pages automatically. For content that lives behind a login or in documents, use the file upload option. You can add multiple data sources to a single chatbot, so there is no need to choose just one method.

Customise the appearance and behaviour. Match the chat widget to your brand colours, set a welcome message, and configure how the bot should introduce itself. You can also set the tone of responses (professional, friendly, concise) and define topics the bot should avoid or redirect to a human agent. These settings make a meaningful difference in how visitors perceive the experience.

Test thoroughly before going live. Ask your chatbot every question you can think of, especially the awkward ones. Check that it handles edge cases gracefully, that it does not fabricate information, and that it knows when to suggest contacting a human. Chatbase provides a testing interface where you can iterate quickly before embedding the widget on your site.

When to upgrade beyond the free tier

The free tier is generous enough for testing and low-traffic sites. For businesses with higher volumes or those that need advanced features like lead capture, analytics, or custom integrations, the paid plans offer meaningful upgrades. If your chatbot is handling a significant portion of your support load, the cost is typically a fraction of what you would spend on equivalent human support hours.

For a detailed comparison of Chatbase against other options, see our Chatbase vs Intercom breakdown.

Other chatbot platforms worth considering

Intercom

Intercom has evolved from a live chat tool into a full AI-first customer service platform. Its Fin AI agent can resolve questions using your help centre content, and the platform includes robust routing, ticketing, and analytics. Intercom works best for established businesses that want chatbot functionality as part of a broader customer communication stack. The pricing reflects this positioning; it is considerably more expensive than standalone chatbot tools, which makes it a better fit for companies that will use the full suite.

The integration depth is a strong point. Intercom connects with CRM systems, analytics platforms, and internal tools, making it possible to build sophisticated support workflows that go beyond simple question-and-answer interactions. If your team already uses Intercom for live chat or email support, adding the AI chatbot layer is a natural extension.

Botpress

Botpress sits between no-code simplicity and developer-grade flexibility. It offers a visual flow builder for designing conversation paths, combined with AI capabilities for natural language understanding. Botpress is open-source at its core, which appeals to teams that want to self-host or heavily customise their chatbot's behaviour.

The learning curve is steeper than platforms like Chatbase. You will get more control over conversation logic, but you will also spend more time building and maintaining your bot. For teams with some technical capacity who want a middle ground between a fully managed platform and a custom build, Botpress is worth evaluating.

Voiceflow

Voiceflow focuses on conversation design with a collaborative visual canvas. Teams can map out complex dialogue flows, test them in real time, and deploy across web, mobile, and voice channels. The platform is particularly strong for chatbots that need sophisticated multi-step workflows, such as guided product selection or troubleshooting sequences.

The collaborative features make it suitable for larger teams. Designers, product managers, and developers can work on the same chatbot project, which helps maintain quality and consistency as the bot grows more complex over time.

Tidio

Tidio combines live chat, chatbot automation, and email marketing in a single platform. It is particularly popular with ecommerce businesses using Shopify, thanks to a native integration that connects the chatbot to order data, product catalogues, and customer information. The AI chatbot can answer product questions and check order status without custom development.

The pricing is accessible for small businesses. Tidio offers a free tier with basic chatbot functionality, making it a reasonable starting point for businesses that want to experiment before committing to a paid plan. The trade-off is that the AI capabilities are less sophisticated than dedicated chatbot platforms.

Watermelon

Watermelon is a European chatbot platform that emphasises GDPR compliance and data privacy. For businesses operating in the EU or handling sensitive customer data, this focus on compliance can simplify regulatory concerns. The platform offers GPT-powered chatbots that can be trained on your website content and documents, similar to Chatbase but with a stronger European data residency story.

Choosing the right platform for your business

Matching platform complexity to your needs

Start with the simplest tool that solves your problem. For most small to mid-sized businesses, a platform like Chatbase provides everything needed: AI-powered responses, easy training, simple embedding, and reasonable pricing. You can always migrate to something more complex later, but you cannot get back the weeks spent building an over-engineered solution.

Consider your team's technical capacity with care. If nobody on your team is comfortable with APIs or code, choose a fully managed no-code platform. If you have developers who want control over the conversation logic, something like Botpress might be more satisfying. If you have a large support team already using Intercom, adding their AI layer is the path of least resistance.

Budget considerations

No-code chatbot platforms typically cost between $19 and $99 per month for small business plans. This covers hosting, AI processing, and basic analytics. Enterprise plans with higher message limits, advanced integrations, and priority support can run from $200 to $500 per month. Custom-built solutions have higher upfront costs but lower ongoing expenses if you have the engineering team to maintain them.

Calculate the return on investment by looking at support ticket volume. If your chatbot deflects even 30 percent of incoming support requests, the time savings for your team usually pay for the platform cost several times over. Track this from day one so you can make informed decisions about scaling up or switching platforms.

Training your chatbot: content preparation

Why content quality determines chatbot quality

The single most important factor in chatbot performance is the quality of your training content. A well-trained chatbot with average AI capabilities will outperform a poorly trained chatbot on the most advanced model every time. Before you worry about which platform to use or how to customise the chat widget, invest time in preparing your content.

Audit your existing content for accuracy and completeness. Go through your website, help centre, and documentation. Remove outdated information, fill in gaps, and make sure every topic a customer might ask about is covered somewhere. If your pricing page has not been updated in six months, the chatbot will confidently quote old prices. If your returns policy changed last quarter but the help article was not updated, the bot will give wrong instructions.

Structuring content for AI consumption

Write content in clear, direct sentences. AI models parse straightforward prose more reliably than content buried in complex formatting, nested tables, or heavily stylised layouts. This does not mean your content needs to be plain or boring, just that the information should be easy to extract from the text.

Organise by topic, with one clear subject per page or document. A single page covering "Shipping, Returns, and Privacy Policy" is harder for a chatbot to parse than three separate, focused pages. When the AI searches your knowledge base to answer a question, topic-focused content produces more accurate results.

Include the questions themselves, not just the answers. If customers frequently ask "Do you ship to Canada?", include that exact phrasing in your FAQ documentation along with the answer. This helps the chatbot match incoming questions to the right information, especially when visitors phrase things differently from how you might expect.

Keeping your training content current

Set a schedule for reviewing and updating your chatbot's knowledge base. Monthly reviews work well for most businesses. Check for outdated pricing, discontinued products, changed policies, and new features that need documenting. A chatbot that gives outdated information erodes trust faster than having no chatbot at all.

Use tools that make content management easier. Maintaining a knowledge base across scattered documents and web pages becomes unmanageable quickly. Centralising your training content in a single system, where you can search, organise, and update everything in one place, makes the ongoing maintenance sustainable. This is where a tool like Fabric becomes valuable, which we cover in the next section.

Building your knowledge base with Fabric

Centralising your chatbot training content

Your chatbot's knowledge base is only as good as the system you use to maintain it. Scattered Google Docs, outdated PDFs, and fragmented help articles create gaps that the chatbot will stumble over. Fabric provides a single place to gather, organise, and maintain all the content your chatbot needs to draw from.

Use Connections to pull in content from your existing tools. Fabric connects to the services where your information already lives, including Google Drive, Notion, Slack, and other platforms. Rather than manually copying content into a separate system, Connections automatically sync relevant material into Fabric, keeping your chatbot's source content up to date as your business evolves.

Store chatbot training documents in Your Cloud. Fabric's cloud drive gives you a central repository for all the files your chatbot should know about: product documentation, policy PDFs, training manuals, and any other reference material. Having everything in one searchable location means you can quickly check whether a topic is covered and update it when things change.

Organising and finding content efficiently

AI Search lets you query your entire knowledge base in natural language. Before training your chatbot, you can use Fabric's AI search to test whether your content covers the questions customers are likely to ask. If a search returns nothing useful, that is a gap you need to fill before it becomes a gap in your chatbot's responses.

Write and maintain FAQ documentation using Notes. Rather than scattering FAQ answers across different platforms, create structured FAQ documents in Fabric that serve as the authoritative source for your chatbot. When a policy changes or a new product launches, update the note once and the information flows through to your chatbot's training data on the next sync.

Connecting Fabric directly to your chatbot

The MCP server enables direct integration between Fabric and your chatbot platform. For teams building custom chatbots or using platforms that support external knowledge sources, Fabric's MCP server provides a structured way to connect your centralised knowledge base to your chatbot's retrieval system. This means your chatbot always draws from the latest version of your content, without manual exports or re-uploads.

Fabric's pricing is accessible for businesses of any size. The free tier covers basic usage, with paid plans at $8 per month on yearly billing or $10 per month on monthly billing. Given that Fabric also serves as your team's general knowledge management tool, the cost is typically justified beyond just the chatbot use case. The platform also plants trees for subscribers through their environmental programme, which is a pleasant secondary benefit.

Integrating your chatbot with your website

Embedding on common website builders

Most chatbot platforms provide a simple embed code. You copy a snippet of JavaScript, paste it into your website's header or footer section, and the chat widget appears. This works on virtually every website builder, including WordPress, Squarespace, Webflow, Wix, and custom-built sites. Chatbase generates the embed code automatically, and the process takes under five minutes.

Framer sites require the embed code in the custom code section. If you are building your site on Framer, add the chatbot's JavaScript snippet in your project's custom code settings. This ensures the widget loads on every page without needing to add it to individual components. For more guidance on building websites without coding, see our how to build a website without coding guide.

WordPress users can often use a dedicated plugin. Several chatbot platforms offer WordPress plugins that handle the integration without touching code. These plugins also sometimes provide additional features like restricting the chatbot to certain pages or adjusting the widget position based on the page type.

Positioning and behaviour settings

Place the chat widget in the bottom-right corner unless you have a specific reason not to. This is where visitors expect to find it, and deviating from the convention creates friction. The widget should be visible but not intrusive; a small icon that expands into the full chat window on click is the standard pattern.

Configure the welcome message carefully. The first message your chatbot displays sets the tone for the entire interaction. Avoid generic greetings like "How can I help you?" in favour of something specific to your business: "Questions about our plans, features, or getting started? Ask away." This primes visitors to ask relevant questions and signals that the bot can provide useful answers.

Set up page-specific behaviour where it makes sense. On your pricing page, the chatbot could proactively offer to explain plan differences. On a product page, it could highlight related items or answer common questions about that specific product. Most platforms support conditional triggers based on the page URL, time on page, or scroll depth.

Using link tracking for chatbot interactions

Track which chatbot links drive meaningful engagement with Dub. When your chatbot shares links to product pages, help articles, or signup flows, using tracked links through Dub lets you measure which recommendations lead to conversions. This data helps you understand which topics your chatbot handles well and which need improvement.

Shortened, tracked links also look cleaner in chat responses. Long URLs in chatbot messages are visually noisy and can feel suspicious to visitors. Dub provides clean, branded short links that maintain trust while giving you detailed analytics on click-through rates and downstream conversions.

Fallback strategies: what happens when the bot cannot answer

Why fallbacks matter more than getting every answer right

No chatbot will answer every question correctly, and pretending otherwise is a mistake. The difference between a good chatbot and a frustrating one is often not in the quality of correct answers but in how gracefully it handles questions it cannot answer. A confident wrong answer damages trust. A clear acknowledgement that the bot is unsure, paired with a path to human help, maintains it.

Using forms as a safety net

Fillout provides a clean fallback when your chatbot reaches its limits. When the bot cannot answer a question confidently, it can direct the visitor to a short form that captures their question, contact details, and any relevant context. This ensures no enquiry falls through the cracks, even outside business hours when no human agent is available.

Design your fallback forms to be as short as possible. Three to five fields is the maximum before completion rates drop significantly. Name, email, and a free-text question field are usually sufficient. You can use Fillout's conditional logic to show additional fields only when relevant, keeping the form lean for simple enquiries while capturing more detail for complex ones.

Escalation to human agents

Define clear escalation triggers in your chatbot platform. Common triggers include: the visitor explicitly asks to speak to a human, the bot has low confidence in its response, the conversation involves a complaint or sensitive topic, or the visitor has asked the same question in multiple ways (suggesting the bot is not providing useful answers).

Pass context to the human agent. When a conversation escalates, the human should see the full chat history, not just the visitor's latest message. This prevents the visitor from having to repeat themselves and lets the agent pick up the conversation smoothly. Most chatbot platforms support this through integrations with help desk tools.

Measuring chatbot performance

Key metrics to track

Resolution rate is the most important metric. This measures the percentage of conversations where the chatbot successfully answers the visitor's question without escalation to a human. A good target for a well-trained chatbot is 60 to 80 percent, depending on the complexity of your product or service. Below 50 percent suggests the training content needs significant improvement.

Track customer satisfaction alongside resolution rates. A chatbot can technically resolve a query by providing an answer, but if the answer was unhelpful or the experience was frustrating, the resolution does not count for much. Simple post-chat surveys (a thumbs up or down, or a one-to-five star rating) provide a useful signal on quality.

Monitor the questions your chatbot cannot answer. These unanswered or poorly answered questions are a goldmine for improving your training content. Review them weekly and add the missing information to your knowledge base. Over time, this feedback loop steadily improves the chatbot's coverage.

Using analytics tools for deeper insights

Dub helps you track the downstream impact of chatbot interactions. By using tracked links in your chatbot's responses, you can measure not just whether visitors click through but whether those clicks lead to signups, purchases, or other conversions. This connects chatbot performance to actual business outcomes rather than just conversation metrics.

Combine chatbot analytics with your broader website analytics. Look at how chatbot interactions correlate with time on site, pages visited, and conversion rates. Visitors who engage with the chatbot often show different behaviour patterns than those who do not, and understanding these patterns helps you optimise both the chatbot and the broader site experience.

Content performance matters for search visibility too. If your chatbot is driving visitors to specific pages on your site, make sure those pages are optimised for search engines. Ranked AI can help ensure the content your chatbot references also performs well in organic search, creating a reinforcing cycle between your chatbot and your SEO strategy.

Common chatbot mistakes and how to avoid them

Training and content mistakes

Launching without enough training content is the most common error. A chatbot that cannot answer basic questions about your product, pricing, or policies will frustrate visitors and create a negative first impression. Invest the time to build a comprehensive knowledge base before going live. Use Fabric to organise and audit your content beforehand.

Failing to update training content is the second most common. Your chatbot's knowledge base is a living resource, not a one-time setup. Set a monthly calendar reminder to review and update your training data. Check for outdated pricing, changed policies, new products, and recently asked questions that the bot could not answer.

Over-training on marketing copy produces unhelpful responses. If your training content reads like sales material, your chatbot will respond to support questions with promotional language. Visitors who ask "How do I cancel my subscription?" want clear instructions, not a pitch about why they should stay. Include straightforward, factual content alongside your marketing material.

Design and behaviour mistakes

Making the chatbot pretend to be human is counterproductive. Visitors can tell they are talking to a bot, and attempting to disguise this creates distrust. Be transparent that the visitor is interacting with an AI assistant. Most people are comfortable with this, provided the bot is helpful.

Setting the chatbot to pop up immediately and aggressively annoys visitors. Give people a moment to orient themselves on the page before the chat widget announces itself. A subtle entrance animation after five to ten seconds, or simply showing a static chat icon without auto-opening, respects the visitor's attention and leads to higher-quality conversations when they do engage.

Ignoring mobile users is a significant oversight. More than half of web traffic comes from mobile devices, and chatbot widgets that work well on desktop can be intrusive or difficult to use on smaller screens. Test your chatbot on multiple screen sizes and ensure the widget does not obscure important page content on mobile.

Advanced: building a custom chatbot for developers

When a custom build makes sense

Custom chatbots are worth considering when your requirements go beyond what no-code platforms offer. This typically means you need deep integration with internal systems, custom AI behaviour that platforms do not support, strict data residency requirements, or you are processing such high volumes that platform pricing becomes prohibitive compared to self-hosting.

Be honest about the ongoing maintenance commitment. A custom chatbot is not a one-time project. You will need to update the underlying model as new versions become available, maintain the retrieval pipeline, monitor performance, fix edge cases, and scale infrastructure as usage grows. Budget for this ongoing work before deciding to build custom.

Architecture overview

A modern custom chatbot typically has three components: a retrieval system, a language model, and a conversation interface. The retrieval system searches your knowledge base to find relevant content for each question. The language model generates a response based on that content. The conversation interface manages the chat widget, message history, and user experience.

Retrieval-augmented generation (RAG) is the standard architecture. Your content is split into chunks, converted to vector embeddings, and stored in a vector database. When a visitor asks a question, the system finds the most relevant chunks and passes them to the language model along with the question. The model then generates a response grounded in your actual content, reducing hallucination compared to relying on the model's general knowledge alone.

Use Fabric's MCP server to connect your custom chatbot to a managed knowledge base. Rather than building and maintaining your own content ingestion pipeline, you can maintain your knowledge base in Fabric and connect your custom chatbot to it through the MCP protocol. This separates the concerns of content management and chatbot engineering, letting your team focus on the conversation experience rather than content infrastructure.

Tools that complement a custom build

CleanShot is useful for capturing screenshots and screen recordings during chatbot testing. When reporting bugs or documenting chatbot behaviour for your team, annotated screenshots provide clarity that text descriptions cannot match. This speeds up the iteration cycle during development and makes it easier to communicate issues to non-technical stakeholders.

Wispr Flow can accelerate the content creation process for your knowledge base. Writing comprehensive training documentation is time-intensive, and voice-to-text tools like Wispr Flow let you dictate content naturally, which is often faster than typing. This is particularly useful when domain experts need to contribute knowledge but find writing documentation tedious.

Video and media-rich content can enhance training data. Platforms like BytePlus provide tools for video processing and delivery that can complement your chatbot's text-based knowledge. For products where visual explanations are clearer than written ones, linking to relevant video content from chatbot responses improves the overall support experience.

SEO benefits of adding a chatbot to your site

How chatbots improve engagement signals

Chatbots increase time on site and reduce bounce rates, both of which matter for search rankings. When a visitor engages with a chatbot rather than bouncing after scanning a page, search engines interpret this as a signal that your site provides value. While the direct SEO impact of engagement metrics is debated, the indirect benefits through improved user experience are well documented.

Chatbot interactions can guide visitors deeper into your site. A well-configured chatbot links to relevant pages in its responses, encouraging visitors to explore content they might not have found through navigation alone. This increases page views per session and distributes link equity across your site, both of which support broader SEO goals.

Creating content from chatbot data

The questions your chatbot receives are a direct window into what your audience wants to know. Review these questions regularly and create dedicated content for the most common topics. A blog post, help article, or landing page that answers a frequently asked chatbot question can rank in search results and drive additional organic traffic. Ranked AI can help you identify which of these topics have the most search potential and optimise the resulting content.

Frequently asked questions

What is the easiest way to add an AI chatbot to my website?

The easiest route is a no-code platform like Chatbase. You point it at your website URL, it crawls your content, and you get an embeddable chat widget within the hour. No coding required. For a broader comparison of options, see our guide to the best AI chatbot for websites.

How much does an AI chatbot cost?

No-code platforms range from free tiers with limited messages to paid plans between $19 and $99 per month for small businesses. Enterprise plans with higher volumes and advanced features can run from $200 to $500 per month. Custom-built solutions have higher upfront development costs but potentially lower ongoing costs at scale.

Can I train a chatbot on my own content?

Yes. Most modern chatbot platforms support training on your own website content, documents, and FAQ databases. Chatbase lets you upload PDFs, paste text, or crawl your website automatically. The quality of responses depends directly on the quality and comprehensiveness of your training content.

How long does it take to set up an AI chatbot?

With a no-code platform, you can have a working chatbot embedded on your site within a few hours. The initial setup is fast; the ongoing work is in refining the training content and monitoring performance. Custom-built solutions typically take weeks to months depending on complexity.

What happens when the chatbot cannot answer a question?

A good chatbot acknowledges its limitations and offers alternatives. You can configure fallback behaviour such as directing visitors to a contact form built with Fillout, escalating to a live agent, or suggesting related help articles. The key is ensuring no visitor hits a dead end.

Do AI chatbots work on mobile?

Yes, though you should test the chat widget on mobile devices before going live. The widget should not obscure important page content, and the chat interface should be easy to use on smaller screens. Most reputable platforms handle mobile responsiveness automatically.

Is Fabric free?

Fabric offers a free tier that covers basic usage. Paid plans are $8 per month with yearly billing or $10 per month on monthly billing. The paid plans include expanded storage, more Connections, and full access to AI Search and the MCP server.

Can I use a chatbot with Shopify?

Yes. Several chatbot platforms integrate directly with Shopify, including Tidio, which has a native Shopify app. Chatbase can also be embedded on Shopify stores using the embed code in your theme's custom code section. For more ecommerce tools, see our best tools for ecommerce founders guide.

How do I measure whether my chatbot is working?

Track resolution rate (percentage of questions answered without human help), customer satisfaction ratings, and the volume of unanswered questions. Use Dub to track clicks on links shared by the chatbot and measure downstream conversions. Review unanswered questions weekly to improve training content.

Should I build a custom chatbot or use a platform?

Use a platform unless you have specific requirements that platforms cannot meet, such as deep integration with internal systems, custom AI behaviour, or strict data residency needs. Platforms like Chatbase cover the needs of the vast majority of businesses, and the development and maintenance costs of a custom solution are substantial.

Can a chatbot replace my entire support team?

No. A chatbot handles routine, repetitive questions well, typically covering 60 to 80 percent of incoming queries. Complex issues, complaints, and situations requiring empathy or judgement still need human agents. The goal is to free your team from repetitive work so they can focus on the interactions that benefit most from a human touch.

How do I keep my chatbot's answers up to date?

Set a monthly review schedule for your training content. Update pricing, policies, product information, and any other content that changes over time. Using a centralised knowledge management tool like Fabric makes this easier by keeping all your training content searchable and organised in one place.

What is the best chatbot for a small business or solopreneur?

Chatbase is the strongest option for most small businesses and solopreneurs. It offers a straightforward setup, reasonable pricing, and solid AI capabilities without requiring technical expertise. The free tier lets you test the concept before committing to a paid plan.

How do I connect my chatbot to my knowledge base?

If you are using a no-code platform, you typically upload documents or provide URLs during setup. For a more integrated approach, maintain your knowledge base in Fabric and use the MCP server to connect it to your chatbot platform. This keeps your content management separate from your chatbot configuration, making both easier to maintain.

Can I use a chatbot on a Framer website?

Yes. Add the chatbot's JavaScript embed code in your Framer project's custom code settings. The widget will appear on all pages of your site. For more on building websites with Framer and similar tools, see our guide on how to build a website without coding.

How do I track the ROI of my chatbot?

Calculate the number of support queries your chatbot handles per month and estimate the cost of handling those manually (agent time multiplied by hourly rate). Compare this to your chatbot platform cost. Use Dub to track conversions from chatbot-shared links for a fuller picture of revenue impact beyond just cost savings.