Last updated 2026-08-12
Every tool here does the visible job well: a bubble in the corner, a conversation, a history, a handover to a human. Where they differ is what happens when the visitor asks something the script did not anticipate, which is most of the time.
A scripted bot answers off topic, politely. A model plugged into your pages answers better, provided your pages actually contain the answer, and that is exactly what they rarely do: your opening hours during the holidays, the deposit you ask for, whether you take that job at all. That knowledge is in your head. The comparison that matters is how much work each tool asks to get it out.
The one that knows your business, because you told it in your own words.
With AskMojo, the chatbot is not the starting point, your knowledge is. You describe how you answer today, including the sentences you repeat and the ones you refuse to say, and that becomes a lab you can put on your site or send as a link.
So the visitor gets your answer, not a generic one, and the questions you never anticipated come back to you as a list worth reading. It runs under your name, in the language of the site, and it exports the day you want it elsewhere.
Best for: a French team that wants one inbox for everything
French company, French support, and a widget that gathers site chat, email and messaging in one place. The handover from bot to human is clean, which is what breaks first elsewhere.
Watch out: It gives you the conversation, not the answers. What the bot knows is still yours to write.
Best for: an online shop with recurring questions
Built around commerce: order tracking, product questions, cart recovery. Installs on the usual platforms without a developer.
Watch out: Outside commerce, the prepared scenarios cover less, and the off-script questions are the ones that lose you a customer.
Best for: a support team that already has volume
The most complete of the list: AI agent, help centre, routing, measurement. On real support volume, the machinery earns its keep.
Watch out: Priced for teams that measure support. For a business of two, it is a lot of platform for four questions a day.
Best for: building the conversation exactly as you want it
A visual builder where each branch is visible, plus connectors. When the conversation has real rules, you can express them instead of hoping.
Watch out: It is a workshop, not a finished product, and billing follows usage: the bot that works is the bot that costs.
What it knows on day one
A chatbot with nothing in it answers nothing. Ask what it reads, and how you correct it when it is wrong.
What it does when it does not know
The honest answer is to say so and pass on. A bot that invents to stay helpful costs more than the call it saved.
Whose name it carries
The widget speaks in your place, in front of your clients. Check what shows, in which language, and whose logo is at the bottom.
What you learn from it
The list of questions nobody could answer is the most useful thing a chatbot produces. Make sure you can read it.
Almost all of them work the same way: create the bot, copy a script tag, paste it before the closing body tag of your site or install the plugin for your CMS. Twenty minutes. The other ninety per cent of the work is writing what it should answer.
Free tiers exist at Crisp, Tidio and Botpress, capped by number of conversations or seats. They are enough to see whether visitors use it. They stop the month it starts working, which is the right moment to decide what it is worth to you.
Two models. A monthly subscription per seat, which is predictable and independent of traffic, or usage billing per conversation, which is cheap while nothing happens and grows with your success. The real cost is neither: it is the time spent writing what it knows, and keeping it true.
It is worth it when the same four questions come back every week and you answer them at 11pm. It is not worth it if your visitors mostly want to book, or to speak to someone: then you need booking, or a phone number that is visible. Count your questions before installing anything.
Turn your way of working into a reusable magik, run it on any model, and export everything with one click.
Browse the labsA general tool answers everyone. These pages start from the job, with its own words and constraints.