The best AI chatbot for ecommerce, ranked on real evidence

The best ai chatbot for ecommerce is the one whose billing unit matches how you actually get asked questions, and which tells a customer it does not know rather than guessing. On prices published on 4 September 2026, chatfor.site, Tidio, Chatbase and Intercom cover four different shapes of shop. This ranking shows which is which.

Every price below was read off the vendor's own pricing page on 4 September 2026 and is quoted with its billing basis. Nothing is from memory, and where a figure was not published we say so. For the category explained before you shop, start with our ecommerce chatbot guide.

Disclosure: chatfor.site is our own product and is ranked first here. The criteria are stated in full below, so you can disagree with the weighting and re-rank the list yourself.

What counts as the best ai chatbot for ecommerce?

An AI chatbot for ecommerce is software that answers a shopper's typed question on your own website, using your delivery, returns, sizing and product information rather than a scripted menu of buttons. The good ones are judged on four things: cost per answered question, availability, honesty when a question is not covered, and how quickly you can correct a wrong answer.

Those became the five criteria for this ranking:

1. Is the price knowable without a sales call? A tool you cannot price is a tool you cannot compare.
2. What is the billing unit, and what does it convert to per answered question? Seats, resolutions, conversations and message credits are not the same thing.
3. What happens when the supplied material does not cover a question? A confident wrong answer about a returns window costs more than the subscription.
4. Can the shop owner see why an answer was given, and overrule it?
5. Does it fit a shop with no support team? Most searches for this phrase come from people who are the support team.

Worth knowing: criteria 3 and 4 are scored only for chatfor.site, because they cannot be established from a pricing page and we did not run controlled trials of the other three. They are marked "not verified" in the table rather than guessed at. That is less satisfying than a full grid of ticks and considerably more honest.

Best ai chatbot for ecommerce: the ranking at a glance

The four tools, ranked

1. chatfor.site, for the shop whose answers are already written down

chatfor.site turns documents you already have into an assistant on your own site. You upload policies, price lists, FAQs and product guides, or paste text straight in, and visitors get answers drawn from those uploads. Install is one script tag, the same shape as adding Google Analytics.

The design decision that puts it first on criteria 3 and 4: it answers only from material you supplied, and is instructed to say it does not have that to hand when the documents do not cover a question. Ask it who the chief executive is with nothing about that in the uploads and it declines rather than improvising. That is refusal by design: it narrows the surface for invention by restricting the source material. It does not repeal how language models work, and no tool can promise that.

You can also open any past conversation in the dashboard, see which passages an answer was drawn from, and overrule it. When a customer tells you the bot said something wrong, that turns a week of argument into a fifteen minute fix.

Pricing, from the plan catalogue:

Free carries a small chatfor.site badge; paid plans remove it. Growth adds conversation history and search plus the option to place the assistant inside a page rather than only in a bubble. Every plan starts with a free account, so nothing needs a card until you choose to pay.

Best for: a shop fielding the same handful of questions daily, and agencies running assistants for several client sites from one account. Weakest for: live order lookups, ticket routing, or a support team of ten.

2. Tidio with Lyro, for a shop that wants human chat and automation in one window

Tidio publishes a free tier with 50 billable conversations, a Starter plan at "$24.17/mo." and a Growth plan from "$49.17/mo.", with Plus starting at "$300/mo. + monthly usage" and Premium on request. Its Lyro AI agent is sold as a separate add-on starting at "$32.50/mo." from 50 Lyro conversations (Source: Tidio pricing page, read 4 September 2026).

Two things follow. The billing unit is a conversation rather than a message, so a shopper asking six questions in one thread consumes one unit. And the AI is priced separately from the chat product, so a realistic monthly figure for automated answering is the plan plus the add-on. Do that sum before comparing it with anything.

Best for: a shop that wants staffed live chat and automation in the same widget and is comfortable with conversation-based billing. Check before buying: whether the displayed monthly prices are annual-billing prices, and how a Lyro conversation is counted against the allowance.

3. Chatbase, for building an assistant on your own documents

Chatbase publishes a free plan with 50 message credits a month, one agent and 1 MB of training content, then Hobby at $40 a month for 700 credits and 10 MB, Standard at $150 a month for 4,000 credits and 20 MB, and Pro at $500 a month for 15,000 credits and 40 MB, with Enterprise on request and 20 per cent off yearly plans (Source: Chatbase pricing page, read 4 September 2026).

The billing unit here is a message credit, and the training limit is in megabytes rather than a document count. Translate both before you commit: work out roughly how many megabytes your policies and product notes come to, and ask how many credits one answer consumes on the model you intend to use. That ratio decides your real cost per answered question.

Best for: someone who wants an assistant trained on their own documents and is happy to reason in credits and megabytes. Check before buying: the credit-to-answer ratio, and the behaviour when a question falls outside the training content.

4. Intercom with Fin, for a team that already has support agents

Intercom publishes Essential at "$29 per seat/mo", Advanced at "$85 per seat/mo" and Expert at "$132 per seat/mo", with its Fin AI agent included on all plans "From $0.99 per Fin outcome", a standalone Fin option with no seat costs at the same "from" rate, a Copilot add-on at "$29 per agent/mo" on annual billing, and a 14 day free trial with a minimum commitment of one full seat (Source: Intercom pricing page, read 4 September 2026).

This is the only entry in the list whose base price scales with your headcount rather than your customers, and the only one that prices automation per resolved outcome. Both are rational choices for a support team of several people with an existing helpdesk. For a one-person shop, paying per seat and then paying again per outcome is a shape that gets expensive at low volume, which is why it ranks fourth on criterion 5 and not on quality. If per-seat pricing is your sticking point, our list of what else you can run in its place goes through the options in detail.

Best for: a support team with agents, tickets and an existing helpdesk. Weakest for: a solo shop owner whose entire support function is one person and an inbox.

Which one fits which kind of shop?

Convert every quote into cost per answered customer question at your own volume, then pick on shape rather than on headline price. Using only the figures published on 4 September 2026 and each vendor's own unit:

• chatfor.site Starter: $29 for 2,000 answers is $0.0145 per answer.
• Chatbase Hobby: $40 for 700 message credits is $0.057 per credit, and a credit is not guaranteed to be one answer.
• Tidio Starter: $24.17 for 100 billable conversations is $0.24 per conversation, and a conversation can hold many messages. Lyro is charged separately from $32.50 a month.
• Intercom with Fin: $0.99 is the published floor per Fin outcome, on top of at least one seat at $29 a month.

Those units are not equivalent, so this is not a like-for-like table and should not be read as one. What it does show is that the cheapest published unit of automated answering in this set is chatfor.site's $0.0145 per answer, and the most expensive is a per-outcome charge roughly 68 times that figure. Whether that matters depends entirely on how many questions you get and how many of them a person needs to handle.

Worth knowing: a shop getting 200 questions a month is a different buyer from a shop getting 20,000, and no single tool is best for both. If your volume is in the hundreds, a free tier plus a $29 plan is the whole decision. If it is in the tens of thousands with a support team attached, per-seat and per-outcome pricing starts to make sense.

What should you test before you commit?

Trials are short, so spend them on the things a pricing page cannot tell you. Five tests, in order of how much they will save you:

1. Ask it something your documents do not cover. Something plausible and absent, such as your warranty length if you have not uploaded a warranty. A tool that invents an answer here will invent one for a customer.
2. Ask the same question three different ways. Including one with a typo and one phrased rudely. Scripted tools fall over on the second and third.
3. Ask a two part question. "How long to Manchester, and can I return it if it does not fit." One answer should cover both.
4. Find the transcript afterwards. Can you see the answer, the passages behind it, and correct it? If not, you have no way to fix a bad answer except emailing support.
5. Check the overage behaviour. When you pass the allowance, does the assistant stop answering or does the bill keep growing? Get that in writing.

For a sense of what good answers look like in practice, our worked examples of real shop questions and answers runs through the common ones, and if you have not yet decided whether you want automation or a person on the other end, the argument for each on the questions that actually arrive settles it faster than a trial will. Anyone specifically weighing up Tidio should also read our notes on where conversation-based billing helps and where it does not.

What we could not verify, and why that matters

Three deliberate gaps in this ranking:

• Answer quality. We did not run a controlled benchmark across four tools on identical documents. Anyone publishing one without describing the corpus and the scoring is publishing marketing.
• Behaviour on uncovered questions for Tidio, Chatbase and Intercom. This is criterion 3, and the most important one. It cannot be read off a pricing page, so it is marked "not verified" rather than assumed either way. Test it in the trial; it takes two minutes.
• Feature grids. Vendor feature lists change weekly and are not comparable across different vocabularies. The prices above are quoted as rendered, with the date, and nothing else about the other three tools is asserted.

One feature of our own is worth flagging as absent rather than implied: domain locking appears in the Starter feature list but is not yet self-serviceable from the dashboard, so treat it as unavailable when you compare.

Frequently asked questions

What is the best ai chatbot for ecommerce for a small shop?

For a shop with no support team and answers already written down, a per-answer plan is the right shape: chatfor.site is free for 100 answers a month and $29 for 2,000. Per-seat pricing charges for staff you do not have, and per-resolution pricing gets expensive at low volume.

How much should an ecommerce chatbot cost?

On prices published 4 September 2026, self-serve entry plans in this set run from $0 on a free tier to $40 a month, with per-seat products starting at $29 per seat plus usage. Convert each into cost per answered customer question at your own volume before comparing headline figures.

Do these chatbots work with Shopify and WooCommerce?

chatfor.site installs as a single script tag, which works on any platform that lets you add one, including Shopify, WooCommerce, WordPress, Squarespace and BigCommerce. It answers from documents you upload rather than reading your product catalogue live, so it will not quote real-time stock levels.

What happens if the chatbot gives a customer wrong information?

With chatfor.site you open the conversation in the dashboard, see which uploaded passages the answer was drawn from, fix or add the document, and overrule the answer. Tools without transcript and source visibility leave you unable to work out why a wrong answer happened, which is the expensive version.

Is a free chatbot good enough for an ecommerce site?

For testing, yes. chatfor.site's free plan gives 100 answers a month, one assistant and 25 documents with no card required, and Tidio and Chatbase both publish free tiers of 50 conversations and 50 message credits respectively. A hundred answers is enough to find out whether it handles your real questions.

Sources

• Intercom pricing page, read 4 September 2026 (Essential $29 per seat/mo, Advanced $85 per seat/mo, Expert $132 per seat/mo, Fin from $0.99 per outcome, Copilot $29 per agent/mo on annual billing): https://www.intercom.com/pricing (nofollow)
• Tidio pricing page, read 4 September 2026 (free tier 50 billable conversations, Starter $24.17/mo., Growth from $49.17/mo., Plus from $300/mo. plus usage, Lyro from $32.50/mo.): https://www.tidio.com/pricing/ (nofollow)
• Chatbase pricing page, read 4 September 2026 (Free 50 message credits and 1 MB, Hobby $40/mo with 700 credits and 10 MB, Standard $150/mo with 4,000 credits and 20 MB, Pro $500/mo with 15,000 credits and 40 MB): https://www.chatbase.co/pricing (nofollow)
• chatfor.site plan catalogue, lib/plans.ts, 4 September 2026