LiveChat AI: Complete Guide to Features, Pricing & Limitations (2026)

So you want to know if LiveChat's AI can just handle your chats for you. Fair question. I've dug into this a fair bit, and the answer is more tangled than the marketing lets on.

LiveChat sells the AI hard. Its AI page promises to "supercharge your customer service" and resolve up to 80% of cases (I'll come back to that number).

But the AI that ships inside LiveChat is mostly there to speed up the people on your team, not to close tickets on its own. And the bit that actually deflects conversations without a human? That's a second product you buy separately.

I've spent the last few years building AI support agents, so I've watched a lot of teams hit this exact wall. You're usually in one of three spots.

You already run LiveChat and someone said "switch on the AI", and you can't tell whether Copilot and Reply Suggestions are the AI or whether you also need the chatbot. Or you're weighing up LiveChat because of the AI pitch, trying to work out what's included versus what's a separate bill. Or you're modeling the real monthly cost, and the pricing is scattered across three or four different product pages.

Either way, I've got you. This guide covers what LiveChat's AI actually is, how the pieces fit, what it can and can't do alone, what it really costs, and how it stacks up against a fully autonomous AI agent.

What is LiveChat's AI?

LiveChat's AI is really three things bundled together: agent-assist features baked into LiveChat plans, a separately-bought chatbot, and a new platform umbrella that launched in 2026.

The agent-assist layer is sold as "Text Intelligence" and comes with every paid LiveChat plan. It includes Text Copilot (a private helper for each agent, previously called "One"), Reply Suggestions (AI-drafted answers pulled from your knowledge), and AI Insights (a weekly summary of top questions), plus Tag Suggestions, Chat Summary, Text Enhancement, and Sentiment Analysis in early access. You can see the full set on LiveChat's AI page and pricing page.

The autonomous, customer-facing layer (the bit that actually answers customers and deflects chats) is a separate product called ChatBot by Text, with its own subscription. LiveChat is run by parent company Text, Inc., and in May 2026 Text announced a rebrand that folds LiveChat, ChatBot, HelpDesk, and KnowledgeBase into one platform, adding Shopify-native "AI Selling Agents" and "Custom Skills" in the launch release.

The tell is in the language. LiveChat says its AI helps you "instantly boost agent productivity" and "automate routine daily tasks", which is help for a human rather than autonomous resolution.

G2: LiveChat scores 4.4/5 from 799 reviews on G2. "The AI assistant effectively manages large volumes of chats... I'm particularly happy about being able to train the AI assistant." via a G2 reviewer (Mid-Market).

This is worth a pause, because it's the single most important thing to get straight before you buy. In my experience an AI copilot's job is to make your human agents more efficient, so they reply faster and more consistently.

That's genuinely useful, and it fits two moments well: when you're not ready to let AI reply to customers directly yet, and on chats that have already been escalated where a person still needs the help. But it's a different job from an AI agent that resolves the ticket end to end, and LiveChat's core AI is the former.

How easy is it to set up LiveChat's AI?

Basic setup is genuinely quick. The AI takes a little more thought, because "the AI" is two products with two setup paths.

For the core product you paste a script tag into your site's code before the closing body tag, or use the direct install if you're on WordPress or Shopify. LiveChat says on its AI page that "most teams go live in under ten minutes" with "no technical skills" needed. For the chatbot, ChatBot says you "paste your website URL, and ChatBot automatically builds a knowledge base", then drop the widget in with a snippet.

Knowledge lives in the Automate section, in a Knowledge Hub where you add website URLs (crawling nested pages or just one) and drag in PDFs, as LiveChat's Reply Suggestions docs lay out. Deeper work runs through the developer platform. Fair warning: the advanced automation rules and analytics take a bit of learning before they click.

What channels does LiveChat's AI work in?

LiveChat has strong messaging coverage, with one real gap on voice.

The native channels on its channels page are the website widget (deployable on unlimited sites under one license), Facebook Messenger, WhatsApp Business, Instagram, Apple Messages for Business, SMS via Twilio, Telegram, and email through an async availability feature. In-chat voice, video, and screen-sharing come via the Snapcall integration.

The gap is real: there's no native phone or voice channel at any tier, since the in-chat voice option is an escalation feature rather than full inbound or outbound phone. There's also no native X, TikTok, or LinkedIn. Ecommerce coverage, on the other hand, is excellent through a 200+ marketplace spanning Shopify, BigCommerce, and WooCommerce.

What are the limitations of LiveChat's AI?

This is the section I'd spend your evaluation time on, because it's where the marketing and the reality part ways. LiveChat's AI has four hard edges.

The first is structural: the AI is agent-assist rather than autonomous resolution. It's built to help your human agents, who still handle every conversation, and even LiveChat's own Copilot docs warn that "Copilot can make mistakes. Always double-check the information it shares."

The second is the one that surprises new buyers: real deflection means buying a second product. The autonomous chatbot, ChatBot by Text, is a separate purchase that LiveChat's own pricing page lists as starting at $52/mo on top of your per-agent LiveChat fees. Sign up expecting AI automation and you find you need a second subscription to get it.

The third is one I'd weigh heavily: there's no simulation mode and no agentic tool-use on the LiveChat side. You can't test the AI against your historical tickets before going live, and Copilot can retrieve information but can't act in other systems, so it can't look up an order or process a refund.

Knowledge Hub is capped too, as the Reply Suggestions docs spell out: 3 sites and 10 PDFs on Team, 10 sites and 30 PDFs on Business and Enterprise, and 2,000 pages per source.

The fourth is that per-agent pricing bills idle seats and scales painfully. You pay per agent, so seats cost you whether they're busy or not, and it compounds as you grow. One Capterra reviewer puts it bluntly: "the price has increased dramatically over the last few years and this means it is becoming less viable as a customer support tool."

What knowledge sources can I train LiveChat's AI on?

LiveChat runs two knowledge systems, one for each side of the AI.

The LiveChat Knowledge Hub feeds Copilot and Reply Suggestions. It takes websites (enter a URL, crawl all nested pages or a single page, mark sources internal-only or public, set an auto-update cadence) and PDFs. The caps matter, and the Reply Suggestions docs list them: up to 10 files and 3 websites on Team, up to 30 files and 10 websites on Business and Enterprise, a max of 2,000 pages per source, and a recommended 300 to 400 pages total for speed and accuracy.

The ChatBot AI Knowledge system (the autonomous side) is wider. It takes websites, files (PDF, DOCX, PPTX, TXT, CSV, MD up to 50 MB each), editable in-platform articles, an import from the sibling KnowledgeBase product, and Zendesk help-center scanning, up to 5,000 sources.

One quirk worth knowing, from ChatBot's AI Knowledge docs: scripted flows take priority over the AI, so "the AI chatbot will first search for an answer within the bot structure. When there is no answer on the bot structure, the bot searches the AI modules."

What AI features does LiveChat have?

It helps to think of LiveChat's AI in three layers.

The first layer is agent-assist. Text Copilot is a private helper for each agent (opened with cmd+G or ctrl+G) that explains features, surfaces account metrics, and answers questions from your Knowledge Sources.

Reply Suggestions sits inline in the chat: an agent clicks "suggest reply", reviews it, and sends. Around those sit Text Enhancement (rephrase, summarize, fix grammar), Tag Suggestions, Chat Summary for handoffs, AI Insights, and Canned Response Suggestions, all documented on the Copilot help page.

The second layer is the autonomous chatbot. ChatBot runs on a visual drag-and-drop builder with blocks for welcome messages, AI Assist, fallbacks, and Shopify or HelpDesk actions. It offers Custom Instructions up to 10,000 characters to set role and personality, tone presets, and a built-in Testing tool so you can chat with the bot before you publish.

The third layer is the new agentic tier from the May 2026 release: AI Selling Agents that recommend products and handle Shopify transactions, and Custom Skills, which Text describes in its launch coverage as "structured workflows that guide how AI agents behave... resolving issues, collecting information, offering incentives, triggering follow-ups, or escalating conversations to humans." There's full REST and RTM API access on the developer platform too, and even a Text MCP server for AI assistants.

How do I improve LiveChat's AI responses?

LiveChat's advice on improving responses is light, and it leans on curating your knowledge rather than on feedback loops.

The clearest tip in the Reply Suggestions docs is to be picky with sources: "resist the temptation to make the AI model learn everything, as this will not only slow it down, it will also make the suggestions much less relevant... generally, the lower-level, concrete the source, the better." The advice is to keep to 300 to 400 pages and prioritize product pages, help-center articles, pricing, and FAQs.

On the ChatBot side, its docs name two levers: Custom Instructions, and a separate Training module for short Q&A pairs that return exact answers.

What's missing is a real feedback loop. There's no thumbs-up/down customer signal that retrains the model, and no simulation mode for testing against past tickets, both of which newer agentic platforms offer.

What resolution rate can I expect from LiveChat's AI?

LiveChat doesn't publish a measured resolution rate, so treat the headline numbers as marketing ceilings rather than benchmarks.

The claims move around depending on the page: "resolve up to 60% of customer queries" on the AI page, "up to 80% of cases" on the chatbot page, and a "74% AI resolution rate" in the 2026 Text rebrand release. All three are best-case figures rather than measured averages.

LiveChat's own 2025 Customer Service Report, drawn from over 2 billion chats, measures satisfaction rather than resolution, and puts chatbot-handled CSAT at 64.7%.

For a sense of where the field actually sits, our own benchmark data across 195 deployments and 55+ vendors puts the median AI-handling rate at around 70%. That's a directional aggregate rather than a like-for-like comparison (every vendor defines its metric differently, and a "deflection" number isn't a "resolution" number), but it tells you "up to 80%" is a ceiling rather than a typical result.

There's a deeper point here, and I'd hang onto it: a resolution rate is only as truthful as the escalation path behind it. A high number means very little if the customer can't easily reach a human when the AI falls short; it only reflects reality when anything the AI can't handle goes cleanly to a person.

The vendor case studies bear this out, landing well below the ceiling: Funded Trading Plus automated 125,000 chats a year at 93% CSAT with an 18% workload cut, and Hairlust reported around a 20% drop in support time and ticket volume.

What AI model does LiveChat use?

This is the murkiest part of LiveChat's AI, and there's a genuine contradiction worth knowing about.

LiveChat's official line denies using third-party model providers. Its pricing FAQ states that "ChatBot is a standalone AI. It doesn't rely on third-party providers like OpenAI, Google Bard, or Bing AI."

The ChatBot AI page doubles down, promising "no extra configuration needed with AI providers, like Anthropic, Google or OpenAI. We handle that for you."

But here's where it gets interesting. LiveChat's own sub-processor list, dated October 2025, names both OpenAI and Google under "AI-enabled functionality", so the "standalone, in-house AI" claim doesn't line up with the disclosure.

My read is that LiveChat wraps third-party LLM APIs while building its own orchestration and small specialized models on top. No specific model is named, and you can't pick between models.

What languages does LiveChat's AI work in?

The chat widget's interface is translated into 48 languages, with agents groupable by language and the widget shown per page in the matching one.

The AI's language support is documented less clearly. ChatBot built with AI Knowledge claims in its multilingual docs that the "AI model will handle multiple languages automatically", adjusting to whatever the customer opens in.

But no exact count is published, some ChatBot features are English-only, and cross-language retrieval (an English knowledge base answering a Spanish question) isn't spelled out. For reference, Intercom's Fin advertises 45+ languages and Tidio's Lyro 12, so LiveChat's disclosure is thin next to them.

How secure is LiveChat's AI?

LiveChat's security is solid but mid-tier, with a couple of real gaps for enterprise buyers.

On the strong side, it landed SOC 2 Type I in December 2025 and has said it will go on to a Type II audit, so Type II is still pending as of writing. The company is GDPR compliant with an EU data-center option, offers HIPAA "readiness" via a Business Associate Agreement on Enterprise, and encrypts connections with 256-bit TLS, all set out on its security page.

On AI data specifically it has a genuinely strong stance, opting out of third-party LLM training. Its AI trust policy says "our partners, like OpenAI, do not use the data from LiveChat to train their models."

The gaps: ISO 27001 isn't listed, and the security page still shows a Privacy Shield badge, a framework that was struck down back in 2020. LiveChat also notes it trains small in-house models on customer data for features like tag suggestions.

Who is using LiveChat's AI?

LiveChat markets 35,000+ customers and a logo wall with names like PayPal, IKEA, Toyota, and Mercedes-Benz. Those are live-chat customers, though, and LiveChat doesn't publicly show which of them use the AI (a distinction I always check). The named AI case studies lean SMB and mid-market.

The specific AI results LiveChat does publish include Funded Trading Plus, a financial-trading firm that automated 125,000 chats a year at 93% CSAT with an 18% workload cut, and Hairlust, an ecommerce brand that trimmed support time and ticket volume by around 20%.

It also names Wembley Stadium, which credits ChatBot with $1.5M in added revenue over six months, and The Chat Shop, an outsourced chat provider reporting 16× higher average session value and a 98% CSAT.

How much does LiveChat's AI cost?

This is where the "separate product" reality shows up on the invoice. The agent-assist AI comes with your LiveChat plan; the autonomous AI is a separate bill.

Play video

LiveChat (per agent)

LiveChat is priced per agent per month, billed annually, on its pricing page: Starter at $19, Team at $49, Business at $79, and Enterprise custom. All of them include the Text Intelligence agent-assist features, with richer ones unlocking higher up (Text Enhancement and Sentiment Analysis are gated to Business, for example).

ChatBot by Text (per resolution)

To get autonomous deflection you add ChatBot, priced on its own pricing page: Essential at $19/user/mo (or $52/mo on legacy pricing), Growth at $79/user/mo, and Enterprise custom, with 10, 200, and 2,000+ AI resolutions a month. Go over your limit and 50-resolution packs auto-refill at $49.50, which is about $0.99 per resolution. ChatBot counts a resolution as a conversation that "ends without the customer requesting more help after the AI's final reply."

Put together, a full suite stacks up: LiveChat (per agent), plus ChatBot (per resolution), plus optionally HelpDesk at $29/user/mo and KnowledgeBase at $49/mo. Here's a worked example for a 10-agent team wanting LiveChat plus ChatBot with around 500 monthly resolutions:

Does LiveChat's AI have a free trial?

Yes. LiveChat and ChatBot both offer a 14-day free trial with no credit card, confirmed on both pricing pages (and I'd start both, since they're separate products).

Neither has a permanent free plan, though. After 14 days you need a paid subscription, and the "first AI bot free" line refers to the trial rather than a forever tier.

Self-serve demos live on the LiveChat demo and ChatBot demo pages, with 24/7 support during the trial.

Is LiveChat's AI worth it?

LiveChat is a mature, reliable, genuinely well-liked live-chat platform. Whether its AI is worth it, in my view, comes down to what you want the AI to do.

I'd say it's worth it if you want a polished live-chat tool with light AI assist for a lean SMB or ecommerce team: five to fifty agents who lean on real-time human chat for sales and support, and who value reliability and a deep integration ecosystem over deep automation. The agent-assist features genuinely help a human team move faster and stay consistent.

It's less of a fit if you want autonomous resolution, if you need native voice and omnichannel, or if you're scaling past around 30 agents where per-agent pricing plus a separate ChatBot bill starts to compound. That's the same friction the Capterra reviews keep circling back to: a great chat tool, with an AI story that costs more the bigger you get.

What are the pros and cons of LiveChat's AI?

Pros

• A mature, reliable live-chat product with a deep integration ecosystem. LiveChat's been refined since 2002, and its 200+ marketplace integrations and polished agent tools get consistent praise.
• Genuinely useful agent-assist in every plan. Text Copilot and Reply Suggestions make human agents faster and more consistent at no cost above the LiveChat plan.
• Fast basic setup. Under ten minutes to get the widget live, no developer for the standard install.

Cons

• The AI is agent-assist rather than autonomous resolution. LiveChat's own AI helps humans; it doesn't close tickets alone, so you still staff a person for every conversation.
• Real deflection means a separate paid product. Autonomous automation needs a ChatBot by Text subscription on top of the per-agent LiveChat fees.
• Per-agent pricing bills idle seats and scales painfully, with reviewers repeatedly flagging price hikes.
• A model-transparency gap. LiveChat publicly claims a "standalone" AI while its own sub-processor list names OpenAI and Google.

FAQs