A free chat widget is a tier, not a product. Three shapes exist in the market: a working product with a hard ceiling, a working product with no ceiling but a branded widget, and a timed trial. The label "free" tells you nothing until you check which one you are getting.
Five things decide whether a free tier is actually free to run: what the cap counts, what happens when you hit the cap, who owns the transcript after the chat, what the widget writes into the visitor's browser, and what the page has to fetch to keep it working.
Most pages that list "free chat widgets" sort the results by logo and stop there. The list is not the hard part. Picking the right one from the list is, because the term covers three different products with three different costs. The rest of this page walks through what a free tier actually is, what to check before you paste a script tag into your site, and how the maths lands on Asyntai's published allowance.
The three shapes that "free" can take
"Free" gets used for three different business models in chat widgets, and each one stops working in a different place.
A working product with a hard cap. The widget is the same one paying customers use, with a smaller allowance. You install it on your real site, real visitors chat with it, and at some number of messages a month it stops answering until you pay. This is the most useful shape, because the widget itself behaves like the paid product on every metric except volume.
A working product with no cap and a branded widget. The chat works forever, but the visitor sees a logo or text on the widget that advertises the vendor. You trade on-brand presentation for zero recurring cost. This shape is common on hosted platforms where the vendor monetises the impression rather than the message.
A timed trial. The widget is free for fourteen days or thirty days, then it locks. Some trials convert into a paid plan on the same data, others wipe the conversation history and the configuration when the clock runs out. A timed trial is not a free tier. It is a sales motion.
If you cannot tell which of the three you are installing, the pricing page will not say so. You find out when you read the terms, or when the widget does the unexpected thing in front of a real visitor. That is why the rest of this page is about checking, not picking.
What the cap counts, and what happens at the cap
This is the question that decides whether the free tier is usable on your site, and almost no comparison page asks it. The published cap is one number, and the behaviour at the cap is a separate product decision.
Some vendors count messages. Some count chats, where a chat is a single conversation regardless of how many messages it contains. Some count resolved conversations, where a resolved conversation is one a customer marks as done. Some count agent replies only, so a visitor can type all they like and it does not cost you until a human (or the AI) writes back.
Behaviour at the cap varies in the same way. The widget can stop answering new questions, hide itself entirely, keep answering but refuse to send the transcript anywhere, or roll over into the next month and charge you on the overflow. Each of those is a different product.
If the cap counts conversations and your visitors ask one question per conversation, your allowance in messages is roughly the same as your allowance in chats. If your visitors write long, multi-turn exchanges with the AI, the same allowance in messages will cover far fewer conversations. The unit the cap is published in is what you have to model on.
What to check before installing a free chat widget
Five checks before you paste a script tag. None of them takes more than a few minutes, and each one answers a question you would otherwise discover the hard way.
- What does the cap count? Messages, chats, resolved conversations, or agent replies. The unit decides whether the published number matches your traffic.
- What happens at the cap? Stops answering, hides the widget, or bills for overflow. The behaviour at the cap is the behaviour your visitor sees.
- Who owns the transcript afterwards? Can you export the conversation history, and in what format. Can the vendor train on it, or is that opt-out, or is it opt-in. This matters the moment a chat contains a complaint.
- What does the widget write into the visitor's browser? Cookies, local storage, fingerprinting, session identifiers. Anything the widget stores client-side is part of your cookie policy and your consent banner, whether or not the vendor's docs mention it.
- What does the widget cost the page in requests? A script tag on every page adds a network round-trip before the visitor sees anything. Some widgets fetch on every page load. Others lazy-load when the widget opens. The difference shows up in your Core Web Vitals.
These are checks you run on the vendor's own documentation and on your own site. None of them require comparing one widget against another. The point is to install a chat widget you can defend if a visitor, a regulator or your own web team asks why it is there.
What the widget writes into the visitor's browser
A chat widget is a piece of third-party JavaScript that runs on your domain. Whatever it does in the visitor's browser, your site is responsible for under data protection law.
Three things to look at in the documentation and in your browser's storage inspector.
Cookies. First-party cookies set by the widget, third-party cookies set by the vendor's domain, and any session cookie used to keep the chat open across page loads. Each one is a separate line in your cookie banner if you run one in the EEA or the UK.
Local storage and session storage. The widget may store a visitor identifier, a chat thread, or unread message state outside of cookies. Browsers do not surface these in the same UI as cookies, so most consent banners miss them. Inspect the storage panel in DevTools after a single chat to see what was written.
Fingerprinting. Some widgets collect enough client-side signals to identify a returning visitor without a cookie. The vendor's docs may not describe this, because it is usually implemented as a side effect of an analytics library the widget ships with.
If your visitor base is in the EEA, the UK, California or any other jurisdiction with a consent regime, each of those items changes what you have to disclose, and each one is decided by the widget, not by your site.
What the widget costs the page in requests
A chat widget that loads on every page is a network round-trip before your content renders. On a fast site, that round-trip is invisible. On a slow mobile connection, it is the difference between Largest Contentful Paint at one second and at three.
Two patterns in use today.
Load on every page. The script tag ships a small loader that pulls the widget bundle and connects to a websocket. Every page pays the cost of the loader, even if the visitor never opens the widget. This is the most common pattern and the most expensive.
Lazy-load on intent. The widget is not fetched until the visitor scrolls, idles, or moves the pointer towards a corner. The cost is paid once per visitor who shows intent, not once per page view. This pattern is less common, and it tends to be the one a free tier supports, because the vendor is paying for bandwidth out of revenue they are not collecting from you.
To see which pattern a widget uses, open the Network panel, hard-reload your page, and filter by the vendor's domain. If a request fires on a page where the widget is never opened, the pattern is load-on-every-page. If nothing fires until you scroll or hover, it is lazy-load.
Arithmetic on Asyntai's free allowance and the first paid step
Asyntai publishes a free plan with 50 messages and a Starter plan at $39 with 800 messages. The conversion from "messages" to "chats" depends on your visitor behaviour, so the maths here is yours, not the vendor's.
50 messages is a small allowance. It is meant for a site that wants to see whether a chat widget fits the page, not for a site with steady traffic. To know whether the free plan covers a month of chats on your own site, convert the published unit into the unit your visitors use.
Assume a visitor opens the widget, asks one question, and the AI answers. That is one chat and roughly two messages (the visitor's question and the AI's reply). At that shape:
A multi-turn exchange is more expensive in messages. Assume the visitor asks a question, the AI answers, the visitor asks a follow-up, and the AI answers again. That is four messages per chat, and the same 50-message allowance covers:
The cap is a message cap, not a chat cap, and the message cap does not stretch as visitor behaviour stretches. The right way to read the published number is to measure your own exchanges, not to assume a chat is one message.
Where the free plan stops covering traffic
The Starter plan is published at $39 for 800 messages. The break-even point between free and Starter is the monthly message volume at which paying $39 buys more headroom than the free plan's 50 messages can give. That is a number you compute on your own traffic, because the free plan does not roll over.
At the same per-chat conversion above, $39 buys on the Starter plan, read in messages and then converted to chats:
The Standard plan publishes at $139 for 15,000 messages, and the Pro plan at $449 for 50,000. The shape of the step is the point. Going up a tier multiplies the allowance by roughly an order of magnitude each time, not by a small factor, because chat volume does not grow linearly with site traffic.
The mistake most people make with a free chat widget
The mistake is treating the free tier as a free product. The free tier is a paid product with a $0 entry point and a message cap, and the cap is the part of the product you have to plan around.
Two ways the mistake shows up in real installs.
Install on launch day, replace on day 100. The free plan covers the first month. By the third month, traffic has settled above the cap, the widget has stopped answering for new visitors, and the team is reading the chat history to find out how many leads went unanswered while the widget was over budget. The fix is to estimate the cap on day one, not to discover it on day ninety.
Install across multiple sites under one account. The cap is per account, not per site. Three small sites on the same free plan share the same 50 messages. If the sites are unrelated, one of them will run out of budget and the others will keep working. The visitor experience depends on which site they landed on.
The free plan is meant for evaluation. To use it in production, you have to know your own message volume, the shape of the messages your visitors write, and what happens on your site when the cap is hit. The plan does not tell you any of those things, and the widget does not warn you when the answer changes.
When the free tier is the right choice
Three cases where the free tier is what you actually want.
You are choosing between two widgets. Install both on staging, point each at the same content, and see which one answers your real questions well. The free tier is the right tool for a comparison, and you should run the comparison before you sign up for a paid plan on either side.
Your traffic is genuinely below 50 messages a month. A niche site, an internal tool, a microsite that does not get indexed. If your real traffic is well under the cap, the free plan is the right plan, and you do not need a paid tier until traffic changes.
You are testing a single page before rollout. Drop the widget on one high-traffic page, measure the deflection rate and the answer quality for a month, then decide whether to roll out. The free tier is the right budget for a single-page pilot.
In every other case, the free tier is a step on the way to a paid plan. The work before you install is to know which step you are on, and what the next one costs in messages and in dollars.
What changes the answer
The published free allowance is a number. Whether it is enough is a function of your own site, and four things move it.
Volume. A site that gets 500 visitors a month will not hit 50 messages. A site that gets 100,000 will. The cap is in messages, not in visitors, so the conversion rate from visitor to chat matters more than the pageview count.
Conversation length. A visitor who asks one question and leaves costs two messages. A visitor who stays for a multi-turn exchange costs four or six. The same site, the same traffic, can be on either side of the cap depending on what visitors do once they open the widget.
Time of day. A widget that answers out of hours picks up the conversations a human team would otherwise miss. That is the point of having a widget. It is also the reason the message count climbs faster than visitor count suggests.
Page weight. If the widget loads on every page, the cost in network requests is paid by visitors who never open it. If the widget lazy-loads, the cost is paid only by visitors who show intent. The same free plan on the same site is a different product depending on the load pattern.
Before you install, do this
Six concrete steps. None of them requires comparing one widget against another, and each one produces a fact about the widget you are about to install.
- Read the cap in the pricing page, then read it again in the terms. The pricing page says messages. The terms say what counts as a message.
- Open the widget in a real browser, send one message, and inspect the storage. Cookies, local storage, session storage, indexedDB. Each one is a separate disclosure if you run a consent regime.
- Open the Network panel and load a page where you never open the widget. Count the requests to the vendor's domain. If the number is more than zero, the widget is loading on every page.
- Take the published cap and convert it into your own unit. Measure a few real conversations, count the messages, and divide the cap by your own average.
- Decide who owns the transcript. Export a chat and look at the file. If you cannot export, the transcript is the vendor's, not yours.
- Plan the step from free to the next tier before you go over the cap. Know the price of the next plan in dollars, the allowance in messages, and the conversion from messages to your own chats.
Free is a tier, and a tier is the bottom of a ladder. The work before you install is to see the ladder, name the rung you are on, and decide what the next rung costs in messages and in dollars before the cap decides for you.
Try Asyntai on the free plan: 50 messages a month, no clock on the trial, no widget branding on the visitor side. When the cap is close, the published next step is $39 for 800 messages. See the published plans.
Sources and dates. Asyntai plan names, message allowances and prices are from Asyntai's own pricing page, accessed September 2026. Behaviour at the cap, cookie and storage behaviour, and network request patterns are vendor-specific and were not verified against any third-party vendor on this page; check each widget on your own site before installing.