It's tempting to paste the embed snippet the moment you sign up and call it done. Most launches that go wrong don't fail because the technology broke — they fail because a step got skipped: the bot was never tested against real questions, the fallback message was left generic, or nobody checked what visitors were actually asking in week one. This chatbot setup checklist walks through every step worth doing before, and right after, you launch an AI chatbot on your site.
Work through it in order — each step below builds on the last, so skipping ahead usually means backtracking.
Why a checklist beats "just paste the snippet"
The embed itself takes under a minute. Everything that makes the bot actually useful — accurate answers, a fallback that doesn't sound broken, branding that matches your site — happens before that minute, in training and testing. Rushing straight to the snippet usually means your first real visitors become your testers, which is the wrong order. If you haven't yet decided whether a chatbot fits your site, our explainer on what an AI chatbot for your website actually does is worth reading first.
The 7-step launch checklist
- Write and gather the content you'll train on. Before crawling anything, pull together what customers actually ask you: your returns policy, shipping details, pricing questions, common "do you offer X" questions. You can feed this in as uploaded documents, pasted plain text, or direct Q&A pairs — the crawler alone won't catch things that live only in your head or in a PDF you never put online. For a full walkthrough, see how to train a chatbot on your own data.
- Crawl your site. Paste your website URL into the dashboard and let the crawler read your pages — product pages, FAQ, about page, policy pages. This becomes the bulk of the bot's knowledge base. Note the page you crawled and when, so you remember to re-crawl after future content changes.
- Test the bot in the dashboard before anyone else sees it. Ask it the real questions your customers send you — not just easy ones. In strict mode it answers only from your trained content, so if it gets something wrong or doesn't know, that's a content gap to fix now, not a bug to debug later.
- Customize the widget branding. Set the widget's color to match your site, and if removing the "Powered by Askvo" badge matters to your brand, note that's a paid-plan feature — check pricing for which tier includes it.
- Set your fallback message deliberately. This is the line visitors see when the bot genuinely doesn't know an answer. A generic "I don't know" reads as broken; a fallback like "I don't have that answer yet — tap 'Talk to a human' and we'll help directly" turns a dead end into a path forward. Since the bot never invents answers it isn't sure of, this message will get seen — make it count.
-
Publish the embed snippet. Copy the snippet — a custom element plus two script tags — from your bot's install section, and paste it right before the closing
</body>tag on your site. This is the same universal snippet regardless of platform (WordPress, Shopify, Wix, Webflow, Squarespace, React, or plain HTML), so there's no platform-specific setup beyond finding where to paste it. If you're on Shopify specifically, see our guide to adding an AI chatbot to a Shopify store. - Monitor the Unanswered Questions report for the first week. This is the step people skip, and it's the one that matters most right after launch. Every question the bot couldn't answer is logged there. Check it daily for the first week: it tells you exactly what real visitors are asking that your training content doesn't cover yet, which is far more useful than guessing.
Consider a soft launch first
If your site gets meaningful traffic, it's worth a short soft-launch window: publish the snippet, watch the Unanswered Questions report and a handful of real conversations for a day or two, and patch obvious gaps before telling customers about the new chat option in an email or social post. This costs you almost nothing — the bot is already live and answering — but it catches embarrassing gaps (a wrong price, a missing shipping detail) before they reach your widest audience.
What to keep watching after week one
Launch week is the highest-signal period, but the habit shouldn't stop there. A few ongoing checks worth folding into a monthly routine:
- Re-crawl after content changes — new products, updated policies, a redesigned FAQ page. The bot only knows what it was last trained on.
- Watch handoff volume. If "Talk to a human" gets tapped constantly on one topic, that's a sign the bot's content on that topic needs work, not that handoff is failing.
- Check message-credit usage against your plan if you're getting meaningful chat volume — credits are a monthly account-wide pool with an 80% warning email and no overage charges, so there's no surprise bill, just a heads-up in time to upgrade if needed.
- Review leads if you have lead capture on. Follow up while the conversation is fresh — see our chatbot lead generation guide for a simple follow-up playbook.
Getting started
All of this is doable on the Free plan, which includes one chatbot forever with no credit card required, so you can run through this entire checklist before spending anything. When you're ready to compare tiers for lead capture, branding removal, or higher message limits, see pricing or create an account to start training your bot today.
Frequently asked questions
How long does it take to go from signup to a live chatbot?
Most businesses can work through this checklist in under an hour: crawling a typical site takes a few minutes, testing and setting the fallback message takes another chunk, and pasting the embed snippet takes seconds. The variable is usually how much extra content (documents, Q&A pairs) you want to add beyond what the crawler picks up.
What's the single most important step on this list?
Testing the bot with real customer questions before launch, and checking the Unanswered Questions report after. Everything else (branding, fallback wording) is polish; those two steps are what determine whether the bot is actually useful.
Do I need to redo this checklist for every new bot?
Yes, each chatbot has its own training content, branding, and fallback message, so each one deserves its own pass through this list — though subsequent bots go faster once you know the workflow.
What if the bot gives a wrong answer after launch?
Check what it was trained on for that topic — a wrong answer in strict mode almost always traces back to outdated or missing content rather than the bot inventing something. Update the source content or add a Q&A pair, then re-crawl or save, and re-test the same question.