Chatbot Platform, Inside Out: Trace One Customer Message Through Page Rules, Bot Answers, Handover and Tickets (Free Message Path Tracer)
Chatbot platform pages list features in a grid: widget, AI, flows, inbox, tickets. What they rarely show is the order those parts act in when a real visitor types a real question. That order decides whether the visitor gets an answer, a person, an away form or silence. This guide follows one message from the moment it is typed to the moment someone deals with it, layer by layer, and shows what can go wrong at each step. A free tool traces your own messages through your own settings.
⚡ In short
- Every message passes the same gates in order: is chat shown here, is this visitor allowed, is there room in the plan, what does the visitor want, can the bot answer, and who takes over if not.
- The first three gates fail silently. A visitor who is blocked, over a limit or on a page where chat is hidden leaves no trace in your inbox.
- Bots should step aside on two signals: a request for a person, and a question about one customer’s own order or payment.
- A handover is only as good as what happens next: a person online, an away form outside hours, or a visitor left waiting.

Why the order matters
Features don’t fail. The joins between them do
When a small business says “the bot isn’t working”, the bot itself is rarely the problem. The AI answered what it was taught. The keyword bot matched what it was told to match. The inbox delivered what reached it. What broke was the join between two parts: chat was hidden on the very page people asked from, the bot tried to answer a question about someone’s own parcel, or a visitor was handed to “the team” at 9 pm when nobody was there.
You only see those joins when you follow a single message from start to finish. Feature lists can’t show them, because a feature list is flat. A real conversation is a path. Each part either handles the message or passes it to the next one, and the outcome depends on the order of the steps as much as on the steps themselves.
That’s also the most useful way to compare products. Two tools with the same features can behave very differently if one tries the bot before checking whether the visitor asked for a person, or if one sends after-hours messages into a void while the other collects a phone number.
So instead of another feature checklist, this guide walks the path. If you want a scorecard for comparing vendors, our AI chatbot platforms scorecard covers that, and 12 written questions for chatbot companies helps before a demo. This one is about what happens inside, after you’ve signed up.
The path
Seven layers, in the order they act
Is the chat widget shown on this page, device and country at all?
Is this visitor or IP address blocked?
Does the account have room for this visitor this month?
Did they ask for a person, or ask about their own order or payment?
Can the AI answer from its knowledge, or does a saved keyword match?
Is it inside opening hours, and is anyone online?
A person replies, an away form takes a contact, or the chat becomes a ticket.
Layers 1 to 3
The silent gates: chat that never starts
The first three layers decide whether a conversation can happen at all. They matter most because when they block a message, nothing appears anywhere. You can’t fix what you never see.
Page rules
Most tools let you choose where the widget appears: which pages, which devices, which countries. That’s useful. You might hide chat on the checkout page so it doesn’t distract, or show it only to visitors from India. The risk is a rule that’s broader than you meant. Hide chat on every URL containing “/order” and you might also hide it on “/order-tracking”, the one page where people most need help. After any change, visit your key pages on a phone and check the bubble is there. Our guide to showing chat on specific pages walks through the settings.
Block list
Blocking a visitor or IP address stops spam and abuse. It also stops everyone else behind that address. In offices, colleges and some mobile networks, many people share one public IP. Block one troublemaker by IP and you may block a whole office. Prefer blocking the visitor where you can, and review the list every few months.
Plan limit
Every product has some limit: seats, conversations, visitors or AI replies. What matters is what happens at the limit. Some tools charge automatically. Some stop the widget entirely. On LiveChatBot’s free plan, each website gets 100 visitors a month; past that, tracking and alerts continue but chat pauses for new over-limit visitors until you add a visitor pack or the month resets, and nothing is charged automatically. Whatever tool you use, know which of these happens, and check your usage before a busy week.
Layer 4
The intent check: when the bot should step aside
Once a message arrives, the next question is not “what’s the answer?” but “should the bot answer at all?” Skipping this step causes most of the bad bot experiences people complain about.
Two signals should send a message straight past the bot.
The visitor asked for a person. “Can I talk to someone”, “human please”, “call me”, “real person”. However good your bot is, answering a request for a person with another bot reply feels like being ignored. Hand over immediately and say what happens next.
The question is about the visitor’s own order, payment, booking or account. “Where is my order”, “my payment failed but money was deducted”, “I haven’t received my refund”. Unless your bot is connected to your order system, it can’t look these up. A general answer like “orders usually arrive in 3 to 6 days” sounds helpful but doesn’t answer the question, and it annoys someone who is already worried. The bot’s job here is to hand over and, if nobody is online, collect a way to reply.
Everything else, the general questions with stable answers, is the bot’s territory: delivery areas, return rules, cash on delivery, hours, sizes, prices that don’t change daily.
How well a tool handles this layer varies. Some detect handover requests automatically; some need you to set it up with flows or keywords. When you test any chatbot platform, type “can I talk to a person” and “where is my order” in your first five minutes. The replies tell you more than any feature page. Our guide to handing a chat from bot to person covers the wording.
Free tool
Trace your own messages through the layers
Set the situation: page, visitor, plan, hours, team and AI. Add what your bot knows and a few real visitor messages. The tracer shows the path each message takes, where it ends, and what to fix.
1. Describe the moment the message arrives
2. What the bot has been taught, one per line as keyword: answer
3. Visitor messages to trace, one per line
A simplified model that runs only in your browser; nothing is sent or saved. Real AI matching is looser than the word matching used here, and every product orders its steps a little differently.
Layer 5
Two kinds of bot answer, and why you want both
Most small-business tools now offer two ways for the bot to answer, and they behave quite differently.
AI answers from your knowledge
You give the AI your information: delivery times, return rules, hours, product notes. When a question comes in, it looks for the most relevant part and writes a reply from it. Its strength is flexibility. “kab tak deliver hoga”, “how many days to Pune” and “shipping time?” can all be matched to the same delivery answer, even though they share few words. Its weakness is that it can only be as accurate as what you taught it, and it usually has a usage allowance. Our guide to training your AI chatbot shows what to give it.
Keyword answers from saved replies
A keyword bot checks whether a saved keyword appears in the message and sends the matching saved reply. It’s predictable, free to run, and never says anything you didn’t write. Its weakness is that it only matches the words you anticipated. If your keyword is “delivery” and the visitor types “deliver”, it misses.
Why both matter
The useful setup is AI first, keyword bot as the safety net. In LiveChatBot, when AI replies are unavailable or used up for the month, the free keyword bot answers from your saved replies instead, so the website never goes completely quiet. That only works if your keywords cover the wording people actually use. The tracer above flags messages the AI answers but the keyword bot would miss: those are the keywords to add. See using saved replies for setting them up.
When the bot can’t answer either way, the message should move on to a person, not end with “Sorry, I didn’t understand.”
Example
Six messages, one evening, one shop
| Message | Path | Ends as |
|---|---|---|
| “delivery to pune?” | Widget → intent check passes → AI matches delivery | Answered by AI; keyword bot would also match |
| “kab tak deliver hoga” | Widget → AI matches delivery | Answered by AI; keyword bot would miss “deliver” |
| “is cod available” | Widget → AI matches cash on delivery | Answered by AI |
| “where is my order” | Widget → own-order signal → handover → closed | Away form, then ticket |
| “can i talk to a person” | Widget → person request → handover → closed | Away form, then ticket |
| “do you take bulk orders for weddings” | Widget → nothing in knowledge fits → handover → closed | Away form, then ticket; a gap to fill |
Illustrative example: a clothing shop at 9 pm, outside opening hours, AI replies on. It’s the same example the tracer loads.
Layers 6 and 7
The handover: four ways it can end
A handover isn’t an ending. It’s the point where the message stops being the software’s job and becomes yours. What happens next depends on two facts: is it inside your opening hours, and is anyone actually online?
Inside hours, someone online. The best case. The chat appears in the team inbox and, if you use it, the Android app. Someone replies. How fast matters a lot here; our guide to live chat response time has realistic targets for small teams.
Inside hours, nobody online. The quiet failure. Your opening hours say you’re available, so the visitor expects a person, but everyone is at lunch or with a customer. The visitor waits, then leaves. Two fixes: phone alerts so someone notices, and a chat rescuer, a follow-up message in Flows that keeps the visitor engaged or asks for a contact if nobody has replied after a set time.
Outside hours. If you’ve set opening hours, the away form collects an email or phone number with the message. That’s not a failure: the visitor knows when to expect a reply, and you have a way to give one. See setting opening hours.
The follow-up. An away-form message or a chat that needs work later can become a ticket with an owner and a status, so it isn’t lost in the inbox. Customers can check their ticket page without logging in. Our guide to turning a chat into a ticket covers this.
The pattern to avoid is a handover with no destination: the bot says “connecting you to our team” and nothing happens. Trace your after-hours messages in the tool, and if any end as “waiting, nobody online”, fix that before anything else.
Common breaks
Where the path most often goes wrong
Using the trace
How to test any chatbot platform in 20 minutes
You can run the same trace on a real account, whether you’re trying a new tool or checking one you already use. You need a second device (a phone on mobile data is ideal, since it has a different IP address) and about twenty minutes.
- Open your three busiest pages on the phone. Is the bubble there on each one? If not, check the page rules.
- Send a general question you’ve taught the bot, worded the way a customer would, with a typo or a Hindi word. Does the right answer come back?
- Send the same question with different wording. This tests how flexible the matching is.
- Switch AI replies off, if the tool allows, and repeat. Does the keyword bot catch it, or does the chat go quiet?
- Type “can I talk to a person”. Does the bot step aside immediately?
- Type “where is my order”. Does it hand over, or guess?
- Send a message inside opening hours with nobody watching. Time how long until someone notices.
- Send a message outside opening hours. Does the away form appear and ask for a contact? Does the message show up as something you can follow up, such as a ticket?
Write down where each test ends. If every one ends somewhere sensible, the joins are working. If any end in silence, you’ve found the layer to fix. Repeat the test after any big change: a website redesign, new page rules, new opening hours, or a change of plan.
If you’re building flows for specific pages or visitors, our chatbot builder targeting guide shows how to test those conditions on sample visitors too.
Checklist
Before you trust the path
Getting started
Set up the whole path on the free plan
Every layer in this guide is included in LiveChatBot’s free plan: page, device and country rules, visitor and IP blocking, an AI Agent with a small free monthly allowance, the free keyword bot answering from saved replies, Flows with a chat rescuer, opening hours with an away form, the team inbox, the Android app and tickets. You get two websites with 100 visitors per website each month, and no per-seat charges. See pricing for visitor packs, then start at livechatbot.in and run the eight-step test on your own site.
Frequently asked questions
What does a chatbot platform actually do with a message?
It checks whether chat is shown on that page, whether the visitor is allowed and whether the plan has room, then decides whether the bot should answer, tries to answer from what it was taught, and hands over to a person, an away form or a ticket if it can’t.
Why do some visitors never reach my inbox?
Usually a page rule hiding chat on that page, a block on their visitor or IP address, or a plan limit. These fail silently, so check them first when someone says they couldn’t chat.
Should the bot try to answer “where is my order”?
Not unless it is connected to your order system. A general delivery answer doesn’t help someone asking about their own parcel. Hand it to a person, or collect a contact when nobody is online.
Do I need both AI replies and a keyword bot?
It helps. AI matches varied wording; the keyword bot is predictable and keeps answering when AI replies are off or used up. Give the keyword bot the wording your customers actually use.
Does the tracer show exactly how LiveChatBot works?
It’s a simplified model of the common order of steps, with basic word matching. Real AI matching is more flexible. Use it to spot gaps, then test on your own site.
Keep reading
Related guides
Chatbot Options, Narrowed Down: Cross Off the Routes Your Limits Rule Out and Keep a Shortlist of Two (Free Route Eliminator)
Chatbot Price Quotes, Side by Side: Turn Three Different Offers Into One 24-Month Number
Chatbot Online 24/7, Team Online 9 to 6: Map Who Answers Every Hour of Your Week (Free Coverage Planner)
Follow every message to a sensible end
Page rules, bot answers, handover, away form and tickets, all on the free plan. Start free on livechatbot.in.