How to Use the ChatAI Block in UniLink (Add an AI Chatbot to Your Bio Page)
How to Use the ChatAI Block in UniLink (Add an AI Chatbot to Your Bio Page)
A practical guide to setting up, training, and tuning the ChatAI block so your AI assistant actually knows what to say — and stops making things up.
- The ChatAI block adds a customizable AI chatbot to your UniLink page that answers visitor questions based on knowledge you provide — no coding required.
- The system prompt and knowledge base are the two things that determine whether your bot is useful or useless; invest real time in writing both.
- Enable lead capture to collect name and email before the chat starts — this turns casual curiosity into a qualified lead even if the visitor never clicks a link.
- Always set a fallback message so visitors who ask something the bot can't answer know exactly how to reach you next.
If you get the same five questions over and over in your DMs — "What's your pricing?", "Do you ship internationally?", "What's included in the course?" — you are spending real time and attention on questions that a well-configured bot could answer instantly at 2am for someone who will never send that DM at all. The ChatAI block is not a gimmick. It is a direct substitute for the fraction of your DM volume that is pure information delivery. Set it up correctly and visitors get instant answers to their specific questions without you being online; set it up lazily and it gives generic, sometimes wrong answers that erode trust instead of building it. This guide is about the difference between those two outcomes.
What the ChatAI block does
The ChatAI block places an interactive chat widget on your UniLink page. A visitor types a question, and the bot responds in real time based on the instructions and knowledge you've provided. Under the hood, UniLink passes each visitor message to an AI language model along with your system prompt and knowledge base, then streams the response back to the chat window. The visitor experience looks and feels like a normal chat interface — a greeting message, a text input, a response that appears within a few seconds.
The use cases split into a few clear patterns. FAQ automation covers the most common scenario: you write down your pricing, your product details, your policies, your turnaround times, and the bot answers those questions without you being available. Lead capture adds a layer on top: before the conversation starts, the bot asks the visitor for their name and email, which gets saved to your UniLink account even if the visitor never clicks any of your links. This is particularly powerful for coaches, consultants, and anyone with a longer sales cycle — the bot warms up the lead and collects contact info simultaneously. A third pattern is what some creators call the "AI twin" — a bot configured with your tone, your backstory, and your personality to give fans a way to interact with a version of you at scale. The personality-driven assistant use case is less about information delivery and more about engagement and connection.
One honest limitation to set expectations: the bot only knows what you tell it. It does not browse the web, it does not look at your social media, and it does not update itself as your prices or products change. If you change your pricing and don't update the knowledge base, the bot will quote the old price with full confidence. The knowledge base is a static document you maintain — treat it like a FAQ page that needs periodic updates, not a live data feed.
Before you start
- Identify the five questions you get most often: Before you open the Dashboard, write them down. These become the core of your knowledge base. If you don't know which questions come up most, check your DMs or email inbox from the last month and look for patterns.
- Draft your knowledge base content offline first: Open a text document and write out your pricing, your offer details, your turnaround or delivery times, your refund policy, and your background. This does not need to be formatted — plain prose works fine. You'll paste this into the knowledge base field in the Dashboard. Writing it offline first is much easier than writing it directly inside the editor.
- Decide on a bot name and tone: Is this "Sarah from [Your Brand] Support," your own name, or something like "Ask Maya"? Decide before setup so you write your system prompt consistently. The name and tone affect whether the bot feels personal or robotic.
- Log in to your UniLink Dashboard: Go to unilink.us and open the editor for the page where you want the bot to appear.
How to add the ChatAI block to your page
- Click "Add block" in the page editor: Search for "ChatAI" or scroll to the AI section of the block picker. Click the ChatAI block to insert it.
- Set the bot name: Replace the default "AI Assistant" with something that fits your brand. Your own name, a character name, or a simple "[Brand] Support" all work. Avoid leaving the default — it makes the bot feel like a placeholder.
- Upload an avatar image: Use a headshot, a brand logo, or an illustrated character. A face or recognizable image significantly increases the bot's perceived warmth. Leave the default avatar blank only if none of your assets feel appropriate — a blank circle is worse than either alternative.
- Write the greeting message: This is the first thing visitors see when they open the chat. It should introduce who the bot is, what it knows, and invite the first question. Example: "Hey, I'm Maya — I can answer questions about coaching packages, pricing, and what to expect from our program. What would you like to know?"
- Write the system prompt: This is the most important field. It defines who the bot is, what it knows, and how it should behave. See the optimization section below for how to write a good one. Paste your draft here.
- Paste your knowledge base: Copy the product info, pricing, FAQs, and background content you prepared offline. Paste it into the knowledge base field. The more specific and accurate this content is, the better the bot performs.
- Choose a response style: Select Friendly, Professional, or Custom. Friendly is conversational and warm. Professional is formal and concise. Custom lets you define the tone in the system prompt — choose this if your brand voice doesn't fit either preset.
- Configure lead capture: If you want to collect visitor contact info, enable Lead Capture. Choose whether to ask for name only, email only, or both. You can make it required (the visitor must enter before chatting) or optional (they can skip it). Required gives you more leads; optional gives a lower-friction first interaction.
- Set the fallback message: Write the message the bot sends when it cannot confidently answer a question. Something like: "I'm not sure about that one — DM me on Instagram @handle and I'll get you a proper answer." Without this, the bot either makes something up or sends a generic unhelpful response.
- Set max message length: This limits how long each visitor message can be. 500 characters is a reasonable default — it allows detailed questions without allowing essays that may confuse the bot.
- Save and test: Click Save, then open your published page in a separate browser window or incognito tab. Ask the bot the five questions you use most often. Read the answers critically. If any answer is wrong, vague, or off-brand, update the system prompt or knowledge base and retest.
Key settings explained
| Setting | What it controls | Best practice |
|---|---|---|
| Bot name | The display name shown in the chat header and in responses | Use your own name, a character name, or "[Brand] Support" — never leave the default "AI Assistant" |
| Avatar image | The profile picture shown next to bot messages | A headshot or recognizable brand icon increases perceived warmth; a face outperforms a logo for DM-replacement use cases |
| Greeting message | The opening message visitors see when they start a chat | State who the bot is, what it knows, and invite a question — one or two sentences maximum |
| System prompt | The core instruction set that defines the bot's identity, knowledge, behavior, and boundaries | Write at least 150–300 words covering: who the bot is, what it knows, what topics to avoid, the tone to use, and when to trigger the fallback |
| Knowledge base | Static text the bot references to answer questions — pricing, product details, policies, FAQs, bio | Be exhaustive and specific; vague or empty knowledge base is the single most common cause of a useless bot |
| Response style | The general tone preset: Friendly, Professional, or Custom | Use Custom + define tone in the system prompt if your brand voice is distinctive; use Friendly for most creator use cases |
| Lead capture | Collects visitor name and/or email before or during the conversation | Enable with Required for high-intent pages (coaching, consulting); Optional for general info bots where friction reduction matters more |
| Fallback message | What the bot says when it cannot answer confidently | Always set this — include a specific next step like a DM handle, email, or booking link; never leave it blank |
How to train your ChatAI bot effectively
The system prompt and knowledge base work together, and understanding the difference helps you use both correctly. The system prompt is behavioral instruction — it tells the bot who it is, what tone to use, what to do when it's uncertain, and what topics are off-limits. The knowledge base is factual content — your pricing, your product descriptions, your FAQ answers, your personal background. If you put pricing information in the system prompt instead of the knowledge base, it works, but it makes the system prompt unwieldy. If you put behavioral rules in the knowledge base, the bot may not follow them reliably. Keep them separated: system prompt for behavior, knowledge base for facts.
Specificity is the variable that most determines quality. Compare "My coaching program is 12 weeks and costs $1,200" with "My 12-week 1:1 coaching program is $1,200 paid upfront or three monthly payments of $450. It includes one 60-minute video call per week, a private Slack channel with 24-hour response time, and access to the client resource library which has 40+ templates and workbooks. The program is for early-stage founders in B2B SaaS who want to build a sales process. Prerequisites: you must have at least one paying customer already." The second version gives the bot something to work with. The first gives it almost nothing. Visitors asking "what's included?" will get a much better answer from the second.
Test the bot as a stranger would. After setup, open your published page in an incognito window and ask questions the way someone who has never heard of you would ask them — not "what's in the 12-week program?" but "do you do coaching?" or "how much is it?" or "what if I'm not in B2B?" These broader, less precise questions reveal where the knowledge base has gaps. Write down any answer that is wrong, vague, or made up, and add the correct information to the knowledge base. Run through this testing cycle at least twice before you rely on the bot for lead qualification.
Update the knowledge base whenever your offers change. This is the maintenance cost of using a ChatAI block, and it is easy to overlook. If you raise your prices, change your program structure, add a new offer, or remove a product, the bot will continue quoting the old information until you edit the knowledge base. A good habit is to update the ChatAI block knowledge base on the same day you update any pricing or offer page. If a visitor asks the bot about pricing and gets a different number than what's on your links or sales page, the mismatch erodes credibility fast.
Troubleshooting common issues
| Problem | Likely cause | Fix |
|---|---|---|
| Bot says it doesn't know about my products or services | Knowledge base is empty or contains only vague descriptions without specific details | Add detailed product or service descriptions to the knowledge base: name, price, what's included, who it's for. Paste it as plain prose — no special formatting needed |
| Bot makes up information that isn't true | The question is within the bot's topic area but the knowledge base doesn't have a specific answer, so the AI fills in plausibly-sounding content | Add the correct information to the knowledge base. Also instruct the bot in the system prompt: "If you are not certain of an answer, say so and direct the user to [fallback contact]" — this tells the bot to admit uncertainty rather than fabricate |
| Bot sounds generic and robotic despite friendly style selected | System prompt is too short or too vague — the bot has no personality or voice to work from | Extend the system prompt: add specific tone guidance ("casual and warm, like texting a knowledgeable friend"), give the bot a name and a brief backstory, and include example phrases that match your brand voice |
| Bot answers questions correctly but in the wrong language | Visitor is writing in a different language and the bot is responding in that language without restriction | Add to your system prompt: "Always respond in [your language], regardless of the language the visitor writes in." The bot will follow this instruction |
| Lead capture form is not collecting emails in my UniLink account | Lead capture may be enabled on the block but the form fields are set to Optional and visitors are skipping them | Change lead capture fields to Required so visitors must enter before chatting. Alternatively, write the greeting message to specifically invite them to share their email: "Drop your email below and I'll send you more info directly." |
| Bot quotes outdated pricing | Knowledge base was not updated after a price change | Open the ChatAI block settings in the Dashboard and edit the knowledge base to reflect current pricing. Save and retest by asking the bot about pricing directly |
| Bot responds too slowly and visitors leave before reading the answer | The knowledge base is very long or the system prompt is very complex, adding latency to the AI response | Trim the knowledge base to the most frequently needed information — remove content that covers edge cases rarely asked. Keep the system prompt focused. Leaner prompts produce faster responses |
Best fit for
- Creators, coaches, and consultants who get high DM volume with repetitive questions about pricing, services, or availability
- Anyone selling multiple products or services with complex options — the bot explains differences and guides visitors to the right offer
- Creators running paid traffic to their link-in-bio page who want to capture leads even when visitors don't click through to a sales page
- Anyone who wants their page to feel interactive and responsive around the clock, including time zones where they're asleep
Not the right tool if
- You have a simple, single-offer page where a well-written link and description communicate everything needed — the chatbot adds complexity without value
- You are not willing to maintain the knowledge base as your offers change — an outdated bot actively damages credibility
- Your audience expects real-time human responses for sensitive or high-stakes conversations (e.g., mental health support, legal advice)
- Your page has very low traffic — the setup investment only pays off when enough visitors are asking questions to make the automation worthwhile
Frequently asked questions
Does the bot learn from conversations over time?
No. Each conversation starts fresh from your system prompt and knowledge base. The bot does not retain memory between sessions or update its knowledge based on previous chats. If a visitor asks a question that reveals a gap in your knowledge base, you need to manually add that information to the knowledge base in the Dashboard. Think of the bot as reading the same document every time a new conversation starts — it gets no smarter on its own, but it gets smarter whenever you add more to that document.
Can the bot book appointments or process payments?
The ChatAI block is a conversational information layer — it answers questions and collects contact information, but it cannot execute actions like booking calendar slots or processing transactions. What it can do is tell visitors how and where to book, then direct them to your booking link. A typical setup: the bot answers qualifying questions, collects the visitor's email via lead capture, and then says "Based on what you've shared, here's my booking link: [link]." The bot handles the qualification; the booking tool handles the scheduling.
Where do the leads collected via lead capture go?
Leads collected through the ChatAI block's lead capture form are saved to your UniLink account's contacts section, where you can view, export, or integrate them with other tools. Each entry includes the visitor's name and/or email along with the timestamp of the conversation. You can export the contacts list as a CSV from the UniLink Dashboard for import into your email marketing platform or CRM.
How long should my knowledge base be?
Long enough to cover your most common questions in detail, but not so long that it slows down response time. For most creators and small business owners, 300–800 words of well-organized knowledge base content is the sweet spot. This typically covers: your main offer(s) with pricing and what's included, a few sentences about your background and who you work with, your delivery or turnaround policy, your refund or cancellation policy, and how to reach you. If you find yourself writing thousands of words, consider whether some content belongs in the knowledge base or whether it would be better handled by a dedicated FAQ block or a longer page.
Can I have more than one ChatAI block on the same page?
Technically yes, but it's rarely the right choice. Two chat widgets on one page creates confusion about which one to use and splits your visitor's attention. If you need different bots for different purposes — for example, one for product questions and one for general inquiries — consider creating separate UniLink pages with separate bots, and linking between them. On a single page, one well-configured bot is far more effective than two mediocre ones.
- The system prompt is the most important field — write at least 150–300 words covering who the bot is, what it knows, how it should behave, and when to use the fallback message.
- The knowledge base is a static document: it only knows what you put in it and never updates itself. Update it every time your pricing or offers change.
- Enable lead capture to collect visitor contact information even when they don't click through to a sales page or send a DM.
- Always set a fallback message with a specific next step — a DM handle, email, or booking link — so the bot never leaves a visitor stranded.
- Test the bot as a stranger would after every major change: ask the questions your audience actually asks, not the ones you expect them to ask.
Ready to put an AI assistant on your bio page? Create your free UniLink page and add the ChatAI block in a few minutes.
Create Your Free Link-in-Bio Page
Join thousands of creators using UniLink. 40+ blocks, analytics, e-commerce, and AI tools — all free.
Get Started Free