For businesses still running scripted help chat
SiteChat Lead Widget reads your entire website, answers visitor questions in a voice you choose, qualifies the visitor, and hands you the lead with the full transcript. It runs on current AI models, and we publish exactly what those models cost us, down to the fraction of a cent.
Start with a free 30 day pilot See what a conversation costs
Pilot setup is CA$249, and CA$99 for the first three pilot sites in exchange for a published, measured lead report. No lead lift is promised. The pilot exists to measure it.
Try it live: the chat bubble at the bottom right of this page is a working AI demo of this exact product concept, grounded in this page, with the pilot flow running as an approved script. It is a demonstration, labelled as one, and it makes no promises either.
Illustration
Answers grounded in the business pages. Booking runs as an approved script. A human confirms.
Scripted help chat follows a decision tree. The moment a visitor asks a question the tree did not predict, the conversation hits a dead end, a keyword mismatch, or a loop back to the menu. A 2026 Clutch survey of consumers put hard numbers on what that feels like:
And when a bot finally gives up and transfers the chat, 47% of visitors are annoyed at having to repeat their whole issue to the human who takes over.
Source: Clutch, Fix Your Chatbot UX, 2026 survey, clutch.co/resources/fix-your-chatbot-ux. We quote these figures because they are independent. Vendor claims about chatbot conversion are labelled as vendor claims wherever we mention them.
Frustration is not just a service problem. A frustrated visitor leaves, and a visitor who leaves never becomes a lead. For a business that pays real money for every lead, that leak is the expensive part.
| Scripted failure | What the visitor experiences | What changes with SiteChat |
|---|---|---|
| Decision tree dead ends | The question does not match a menu branch, so the bot offers the menu again | The model reads the actual question and answers from your site content, in plain language |
| Keyword matching | Ask about a tune up price in your own words and the bot hears nothing it recognizes | Questions are matched by meaning against your indexed pages, not by exact keywords |
| Fallback loops | Sorry, I did not understand, repeated until the visitor gives up | Low confidence triggers an honest I do not know plus a human offer, never a loop |
| No memory of the conversation | Transferred visitors repeat everything, the complaint 47% of transferred visitors raised in the Clutch survey | The full transcript travels with the handoff, and the lead arrives with the whole conversation attached |
| The human is hidden | 81% in the survey say AI support made reaching a person harder | A human is one tap away from every conversation, on every plan, by design |
| Stale scripted answers | The tree still quotes last years price because nobody updated branch 14 | Your site is re crawled weekly, so the answers follow your pages, and Script Builder covers the actions where wording must be exact |
Modern models follow instructions and stay grounded in supplied source text far better than the keyword matchers inside scripted bots. SiteChat retrieves the four most relevant passages from your own pages for every question and answers from those passages only. If your site does not cover the question, the widget says so and offers a human, instead of guessing.
A typical small business site holds 25,000 to 200,000 tokens of text, roughly 60 to 500 passages once it is indexed. Current models take that in stride, so the widget can know your services, hours, service area, and policies, not just the ten questions someone had time to type into a script.
Streaming responses start in about a second on modern inference services, so there is no long silent wait. And because a full conversation costs a fraction of a cent (see the published table below), we can run a capable model on every chat. Nobody gets routed to a crippled free model to save money.
One more change matters: a human is one tap away in every conversation, and the human receives the full transcript. Nobody repeats their issue. That single feature answers the loudest complaint in the survey above.
The signature feature
Here is the pattern on almost every business site: one action dominates. Book an appointment. Get a quote. Check a price. Track an order. Most of your chat volume is that one action, asked a hundred slightly different ways.
SiteChat studies your real conversations and ranks what visitors ask for most, with the counts to prove it.
For the number one action, the AI writes a complete, ready made handling script: the questions to ask, the exact wording, the information to collect, and where the lead goes.
You read it, edit it, and approve it before it ever runs. Nothing about your prices, policies, or promises goes live without your sign off.
From then on, that action is handled by the approved script: instant, identical every time, no improvisation, and little to no AI inference cost, because a script does not need a model to think.
Example only, with stated assumptions. Say your site has 300 conversations a month on the Growth plan, and booking a visit is 40% of them, which is common for service businesses. Once booking runs as an approved script, most of the AI inference for those 120 conversations disappears. Your total inference cost falls by roughly that same share, around a third to 40%, and every booking is captured in exactly the format your office wants.
Honest footnote: inference is already only a fraction of a cent per conversation, so the dollar saving is small. The bigger wins are speed, consistency, and control, which is why Script Builder is in every plan instead of being a paid add on.
| Handled by the AI model | Handled by an approved script | |
|---|---|---|
| Response starts | About a second, streaming | Instant, it is a fixed flow |
| Full answer time | Roughly 1 to 3 seconds | As fast as the visitor can tap and type |
| AI inference cost | About US$0.003 per conversation | US$0, no model call at all |
| Wording | Generated each time, grounded but variable | Your approved words, identical every time |
| Best for | The long tail of real questions | The one action your business runs on |
Illustrative, based on the published API rates in the cost section below and typical streaming response behaviour. Your top action and volumes will differ; the pilot report measures yours.
Why approval matters: in 2024, a Canadian tribunal ruled in Moffatt v. Air Canada that a business is bound by what its chatbot tells a customer, even when the chatbot invented the policy. Approved scripts put your most common promises in words you chose, reviewed before a single visitor saw them. The AI handles the long tail of questions, grounded in your site. The actions that matter most run on rails you approved.
Most chat tools show you a transcript dump and wish you luck. SiteChat closes the loop:
Your learning stays yours. The system learns from your business conversations to serve your business. Your chat data is not pooled to train anything for other customers.
| Signal measured | What it usually means | Fix proposed |
|---|---|---|
| Same question asked many times | Your site buries the answer, or never had it | A clearer passage for the knowledge base, and a suggested page addition for your site |
| Visitor asks for a human | The answer was missing, unclear, or the task needs a person | Knowledge fix if it was answerable; a Script Builder candidate if it is an action |
| Chat ends mid task | Friction: too many steps, a confusing question, a missing price | A shorter scripted flow, or the missing fact added to the knowledge base |
| Question the widget could not answer | A real gap in your site content | Draft answer for your approval, added to the knowledge base once you sign off |
| One action dominates volume | Your number one request is ready to become a script | A Script Builder draft, waiting for your approval before it runs |
Nothing ships on a guess. Every proposed fix is regression tested against the questions your visitors already asked, including the edge cases from your launch battery, and the changes that pass appear in your monthly report with before and after counts.
Every public page is read, split into passages, and indexed into a private knowledge base for your business only. Indexing a typical site costs under one cent in embedding fees, and we refresh it weekly so new pages and price changes show up.
Friendly neighbour, crisp professional, warm and unhurried. The persona sets tone only. It never changes facts, prices, or what the widget is allowed to promise.
The widget qualifies the visitor, asks for name and contact details, and sends you the lead instantly with the full transcript. Contact details are collected for the inquiry itself, and any marketing follow up asks for separate, explicit consent, logged with a timestamp as Canadian anti spam law requires.
Onboarding is one embed code on your site. We never ask for your website password or login.
Most chat vendors hide the model cost inside credits, multipliers, and per resolution fees. One rival charges US$0.99 for every resolved chat. Another charges about US$0.65 to US$0.78 per AI conversation. Here is what the raw AI actually costs at published API rates, November 2026:
| Model | Input, per 1M tokens | Output, per 1M tokens | One 10 message conversation |
|---|---|---|---|
| Groq GPT-OSS-20B | US$0.075 | US$0.30 | about US$0.002 |
| Google Gemini 2.5 Flash-Lite | US$0.10 | US$0.40 | about US$0.003 |
| OpenAI GPT-4o-mini | US$0.15 | US$0.60 | about US$0.004 |
| Groq GPT-OSS-120B | US$0.15 | US$0.60 | about US$0.004 |
| Anthropic Claude Haiku 4.5 (premium option) | US$1.00 | US$5.00 | about US$0.028 |
Assumptions, stated plainly: each message carries about 1,500 input tokens (the visitor question plus the retrieved passages from your site) and about 250 output tokens, across a 10 message conversation. Indexing your whole site costs under US$0.01 one time at published embedding rates of US$0.02 per million tokens. Sources are listed at the bottom of this page.
| Model class | 5 messages | 10 messages | 20 messages | 300 conversations a month |
|---|---|---|---|---|
| Gemini 2.5 Flash-Lite class | about US$0.001 | about US$0.003 | about US$0.005 | about US$0.75 total |
| GPT-4o-mini class | about US$0.002 | about US$0.004 | about US$0.008 | about US$1.14 total |
| Claude Haiku 4.5 premium class | about US$0.014 | about US$0.028 | about US$0.055 | about US$8.25 total |
Read that last column again. A full month of Growth plan conversations, every one answered by a capable model, costs roughly a dollar in AI fees. Even the premium model at full volume stays under ten dollars. Anyone charging per conversation or per resolution is charging for the metering, not the AI.
Indexing is a one time cost of under a cent, and weekly refreshes cost the same order of magnitude, because only changed pages are re indexed. Then Script Builder starts eating the volume: if your top action is 40% of conversations and it moves to an approved script at zero inference cost, your monthly AI bill falls by roughly that share, from about a dollar to about sixty cents in the example above. The curve only bends one way, month after month, as the learning loop promotes more repeat actions into scripts.
The conclusion: a conversation costs between US$0.003 and US$0.03 in AI fees, and even the premium model stays under three cents. That is why our pricing is a flat monthly fee in Canadian dollars with a clear conversation allowance, instead of credits that burn at different rates depending on which model answered. And it is why we never route your visitors to a free tier: free API tiers can use conversation content to improve the provider products, which is no place for your customers names and phone numbers. Your chats run on paid tiers with no training on your data.
Scripted chat is cheap or free, and you get the frustration numbers above. AI chat fixes the answers, but most vendors then meter the AI in credits, per conversation fees, or branding add ons. Published prices, checked in October and November 2026, in US dollars unless marked:
| Product | Published price | The catch for lead capture |
|---|---|---|
| tawk.to | Free forever; US$29 per month to remove branding | Human live chat with scripted shortcuts; no grounded AI answers on the free tier |
| Tidio | Starter US$24.17 per month billed yearly; Lyro AI from US$32.50 to US$39 for 50 AI conversations | The AI is a paid add on, priced per conversation at roughly US$0.65 to US$0.78 each |
| Crisp | Mini US$45, Essentials US$95 per workspace; free plan has no AI credits | Built around a human inbox; AI credit is small and metered |
| HubSpot Chat | Free inside the free CRM | Scripted builder tied to the CRM; depth lives in paid hubs |
| LiveChat | US$19 per user per month billed yearly, no free plan | Human first; the AI product is separate |
| Drift | Quote only | Being sunset: announced March 2026 and closed to new customers, so its users are actively migrating |
| Product | Published price, US$ per month | White label and agency terms | Billing unit |
|---|---|---|---|
| Chatbase | 40 (700 credits), 150 (4,000 credits), 500 (15,000 credits) | Badge removal adds US$99 per month; full white label is Enterprise only | Credits, and premium models burn several credits per reply |
| SiteGPT | 59 monthly, 39 billed yearly | Badge removal adds US$39 per month | Messages, with stronger models counting 10 to 1 |
| CustomGPT.ai | 99 (1,000 queries), 499 (5,000 queries) | Branding removal only at the 499 tier | Queries, and the bot stops at the cap |
| FastBots | 33 (2,000 credits) and up | Not verified in our research | Credits with model multipliers up to 10 per reply |
| Wonderchat | Free for 50 replies; 39 for 500 replies; 149 for 3,000 | Not verified in our research | Replies, with stronger models burning extra |
| Dante AI | 40, 120, and 400 | Full white label at the 400 tier; lead capture from 120 | Credits with premium model multipliers |
| SiteChat Lead Widget | CA$79 Growth, CA$149 Pro, CA$49 per client site for agencies | Full white label in the agency plan from the first client | Flat monthly fee with a plain conversation allowance. No credits, no multipliers |
Competitor prices are published vendor prices from our November 2026 research review, listed in the sources section. Vendor pricing changes often; check a rivals own page before quoting it.
CA$0 for 30 days
plus CA$249 setup · CA$99 for the first three pilot sites
The first three pilots pay CA$99 setup in exchange for letting us publish their measured report. The pilot is how we prove value, because published lead lift research for this product category is thin, and we will not borrow vendor numbers and call them ours.
CA$79 per month
1 website · up to 300 conversations a month
CA$149 per month
Up to 3 websites or locations · 1,000 conversations a month
CA$49 per client site, per month
Minimum 3 client sites · full white label
Month to month. Cancel inside the product, no phone call required. Setup is refundable until your widget goes live; subscriptions are pro rated to the current month. Full terms are provided in writing before the first payment.
Businesses in our first verticals already pay for leads through ads. Median advertising cost per lead, WordStream and LocaliQ 2026 benchmarks, in US dollars: legal US$131.63, real estate US$102.51, home improvement US$90.92, dental US$72.97. At those prices, a Growth plan at CA$79 a month is covered by one extra captured lead every month or two. That is the math the 30 day pilot tests on your site, with your traffic.
Now onboarding English language sites in Ontario and the rest of Canada outside Quebec. Quebec service, including French, arrives with our v2 compliance package.
It says it does not know, offers to take a message or connect a human, and logs the gap. That log is gold: it becomes a suggested page update and a candidate fix in your monthly report, so the same question gets a real answer next month.
Yes, through Script Builder. The booking flow is a script you approved in advance, connected to your booking link, so the widget collects exactly the details your office needs and a human confirms. It never invents availability.
Tone and wording style only: friendly, formal, brief, warm. It never changes facts, prices, service areas, or what the widget is allowed to promise. Those come from your site and your approved scripts.
No. Conversations run on paid API tiers whose terms exclude training on your data. Free tiers that reserve the right to use content are never used for live visitor chats.
Only the lead fields you configure, typically a name, a phone number or email, and what they need. The inquiry itself gives you a window to respond under Canadian anti spam law, and any marketing follow up asks for separate explicit consent, unticked by default, logged with the wording and timestamp. Unsubscribes are honoured within 10 business days.
Your site is re crawled weekly on Growth and faster on Pro, and you can request a refresh any time you publish new prices or services. Your monthly report flags answers that relied on pages that changed.
Because the gate is artificial. Rivals charge US$39 to US$99 a month just to remove a badge, and reserve full white label for US$400 to US$500 tiers. Our cost to put your brand on the widget is near zero, so we price it that way.
Then the pilot will show it, honestly, in 30 days, for the cost of the setup fee. A widget cannot create visitors. It converts the ones you have, including the after hours visitors your office never sees.
Tell us where to look. We review every application by reading the site first, checking what your current chat does with real questions, and confirming your business is in a vertical where one lead covers the cost. Pilot places are limited while we onboard carefully.
Submitting asks us to review your site for the pilot. It does not sign you up for marketing email, and nothing is charged at this stage.
