Top WordPress AI Chatbot Plugins for Small Businesses (2026)
Chatbot vendors price by conversation, which creates a problem nobody mentions in the sales material: the better your chatbot works, the more it costs you. A bot that successfully answers 500 questions a month is doing exactly what you bought it for, and on several plans here that success will cost more than the salary equivalent of the time it saved.
The numbers are not subtle. One popular plan includes ten AI resolutions a month and sells top-ups at roughly a dollar each. Another charges around five and a half cents for the same thing. A plugin using your own API key costs a fraction of a cent. Same underlying technology, a hundredfold difference in price. This article prices the conversation rather than the plugin, and covers the two things that actually determine whether a small business chatbot works: what it knows, and what it does when it does not know.
Pricing verified August 2026. Conversation and resolution definitions vary by vendor, so check what counts as billable before committing.
What a conversation actually costs
Take a small business whose bot handles 500 customer questions a month, which is roughly 17 a day and entirely ordinary for a site with real traffic.
| Tool | Plan needed | Effective cost per AI conversation | Roughly, at 500 a month |
|---|---|---|---|
| AI Engine, own API key | Pro $79 once | Fractions of a cent | Under $5 a month |
| Crisp | Plus $295/mo | ~$0.055 marginal | $95 to $295/mo by tier |
| Tidio | Growth from $49.17/mo | ~$0.24 on Starter | $49+/mo, Lyro billed separately |
| ChatBot.com | Growth $79/user/mo | ~$0.40, top-ups $0.99 | $376/mo with top-ups |
The spread between the cheapest and most expensive way to answer the same 500 questions is roughly seventy-five times. That is not a difference in quality; all of these ultimately call similar large language models. It is a difference in how much of the underlying API cost the vendor keeps.
What the expensive options sell alongside the model is real: a maintained knowledge base pipeline, a live chat inbox your staff can work from, human handoff, mobile apps, and someone else’s uptime. If you have staff answering chats, that infrastructure is worth paying for. If you want a bot that answers questions from your documentation and nothing more, you are buying a very expensive wrapper.
Grounding and handoff decide everything else
Grounding means the bot answers from your content rather than from the model’s general knowledge. A bot that has ingested your product pages, shipping policy, and FAQ will answer accurately. A bot connected to a raw model will confidently invent a returns policy you do not have, and a customer will hold you to it. Ask exactly what sources a tool ingests, how often it re-crawls, and whether it cites the page it answered from.
Handoff is what happens when the bot cannot answer. The failure mode small businesses actually experience is not a wrong answer; it is a customer trapped in a loop with no way to reach a human, who then leaves and does not come back. Every tool should let a user escalate at any point, capture their email if nobody is available, and pass the conversation transcript to whoever picks it up.
The metric to judge all of this by is deflection rate: the share of conversations resolved without a human. Not conversations handled, not messages sent. If a vendor reports volume rather than deflection, they are reporting the number that grows their invoice rather than the one that measures your benefit.
The 6 options compared
1. AI Engine by Meow Apps
The cheapest credible chatbot for a small business by an enormous margin, and the one to start with unless you need a staffed live chat inbox. AI Engine builds a chatbot on your own provider account and your own API key, so you pay the model provider directly at cost rather than paying a vendor’s markup on every conversation.
It supports OpenAI, Anthropic, Google Gemini, Mistral, Perplexity, and OpenRouter, so you can pick a cheap fast model for routine questions rather than being locked to whatever your vendor resells. There is a capable free version and Pro is $79. Realistic total cost for a small business is $79 once plus a few dollars a month in API usage. The tradeoffs are genuine: you configure the knowledge base yourself, there is no staffed inbox or mobile app for your team, and handoff means capturing an email rather than routing to a waiting agent. For a business where nobody is sitting on live chat anyway, none of that is a loss.
- Pricing: Free version; Pro $79, plus your own API usage
- Model: Bring your own key, multi-provider
- Best for: Businesses with nobody staffing live chat
- Watch out for: No agent inbox; you own the knowledge base setup
2. Crisp
The best value among the full hosted platforms, and the one whose pricing structure most favours small teams. Crisp bills per workspace rather than per seat, which is unusual and matters enormously: a five-person team pays the same as one person, where most competitors multiply.
Plans are Free, Mini at $45 a month with around 90 automated conversations and $5 of AI credits, Essentials at $95 with around 450 conversations and $25 of credits, and Plus at $295 with around 1,350 and $75 of credits. Those figures imply a marginal cost near five and a half cents per automated conversation, which is the cheapest of the hosted options here by a wide margin. All paid plans include an omnichannel AI chatbot with a visual workflow builder alongside a proper shared inbox, so you get automation and human chat in one product. For a small business that does have someone answering messages, this is the pick.
- Pricing: Free; Mini $45, Essentials $95, Plus $295 a month, per workspace
- Model: Hosted, AI credits included per tier
- Best for: Small teams wanting chat and automation together
- Watch out for: Credits are consumed by automation, so budget by volume
3. Tidio
The most polished experience in this group and the easiest to get looking professional quickly, with a widget that small businesses consistently rate well and a Lyro AI agent that handles ecommerce questions capably. If presentation matters and your chat widget is customer-facing on a storefront, this is the strongest product here.
The pricing needs care because it operates on two meters. Plans are Free, Starter at $24.17 a month with 100 billable conversations, Growth from $49.17 with up to 2,000, and Plus from $300 plus usage. Separately, Lyro AI conversations are billed on their own scale, starting at 50 a month and rising, with the free tier giving 50 Lyro conversations once rather than monthly. So the plan price is not the AI price, and a busy bot consumes both meters at once. The Premium tier offers pay-per-resolution billing with a guaranteed 50% resolution rate, which is a more honest structure and priced accordingly.
- Pricing: Free; Starter $24.17, Growth from $49.17, Plus from $300 a month
- Model: Hosted; Lyro AI conversations billed separately
- Best for: Storefronts where widget polish matters
- Watch out for: Two separate meters; free Lyro allowance is one-off, not monthly
4. ChatBot.com
A capable platform with a strong visual builder, multiple AI agents, and proactive campaigns, from the team behind LiveChat. The product is good. The pricing is the reason to read carefully before buying.
Essential is $19 per user per month on annual billing, or $25 monthly, and includes ten resolutions per month. Not a thousand, ten. Growth is $79 per user per month annually, or $99 monthly, for 200 resolutions. When you exceed your allowance, top-ups are 50 additional resolutions for $49.50, which works out at 99 cents per resolution. And because pricing is per user, a three-person team triples the base cost. For a business expecting 500 AI conversations a month, this is comfortably the most expensive option in this article, plausibly by a factor of five. It suits organisations with low volume and high complexity, and it does not suit a small business hoping to deflect routine questions at scale.
- Pricing: Essential $19/user/mo annual, 10 resolutions; Growth $79/user/mo, 200 resolutions
- Model: Hosted, per user, resolution-metered
- Best for: Low-volume, high-complexity conversation design
- Watch out for: Top-ups at $0.99 per resolution; per-user billing multiplies
5. WPBot
The long-standing native WordPress option, with a free core on WordPress.org and paid tiers above it. Its appeal is that it lives entirely inside WordPress: conversations, settings, and data stay on your site rather than in a third-party dashboard, which is a genuine consideration if you are cautious about routing customer messages through an external service.
The free version handles rule-based question and answer flows without AI, which for many small businesses covers the top ten questions perfectly well and costs nothing. Paid tiers add AI integration and richer features; confirm current pricing on the vendor’s own site rather than trusting a comparison table, since this plugin is widely resold through unofficial channels. Judge it as a solid, self-contained option rather than the most capable one, and note that a rule-based bot answering your ten commonest questions often deflects more than a clever AI answering badly.
- Pricing: Free core; paid tiers, confirm on the vendor site
- Model: Self-hosted inside WordPress
- Best for: Keeping conversation data on your own site
- Watch out for: Widely resold unofficially; buy direct
6. Self-hosted open source, such as Botpress
The route that removes per-conversation billing entirely. Open-source conversational platforms can be self-hosted, connected to whichever model provider you choose, and embedded into WordPress, so your only costs are hosting and API usage. For a business with high conversation volume, that changes the economics completely, since the marginal cost of the thousandth conversation is the same fraction of a cent as the first.
The cost is engineering. You are running a service, maintaining it, and owning its uptime, and a chatbot that is down during a sale is worse than no chatbot. Realistically this suits businesses with a developer already on hand, or those with volume high enough that hosted per-conversation pricing has become genuinely painful. For most small businesses, AI Engine gets you most of the economic benefit with a fraction of the operational burden, which is why it is the recommendation above and this is not.
- Pricing: Free software; you pay hosting and model usage
- Model: Self-hosted, any provider
- Best for: High volume with technical staff available
- Watch out for: You own uptime; downtime during a sale is expensive
Comparison table
| Tool | Entry price | Billing basis | Agent inbox | Data location | Best for |
|---|---|---|---|---|---|
| AI Engine | Free, Pro $79 | Your own API usage | No | Your site plus provider | Cheapest capable option |
| Crisp | Free, Mini $45/mo | Per workspace plus credits | Yes | Hosted | Best hosted value |
| Tidio | Free, Starter $24.17/mo | Conversations plus Lyro | Yes | Hosted | Most polished widget |
| ChatBot.com | $19/user/mo annual | Per user plus resolutions | Via LiveChat | Hosted | Complex flows, low volume |
| WPBot | Free core | Licence | Limited | Your site | Keeping data in WordPress |
| Self-hosted open source | Hosting only | Infrastructure | Varies | Your server | High volume with a developer |
Privacy, and what you are sending where
A chatbot receives whatever customers type, which routinely includes order numbers, email addresses, phone numbers, and occasionally things nobody should have typed into a chat window. That data leaves your site.
- Name the processors in your privacy policy. A hosted chatbot with an AI backend usually means two: the platform and the model provider. Both belong in your policy.
- Check data residency if you are in the EU. Ask where conversations are stored and processed, and whether the vendor offers EU hosting. This is a routine question and any serious vendor answers it plainly.
- Confirm training terms in writing. Whether your conversations can be used to train models varies by provider and by plan, and the default is not always what you would want.
- Set a retention period. Conversation logs are useful for improving the bot and a liability once they are old. Ninety days suits most small businesses.
- Never let the bot handle payment details. If a customer starts typing a card number, the bot should stop them. Some platforms redact automatically; most do not.
Related guides
Frequently asked questions
What counts as a billable conversation?
It varies enough to change your bill several times over. Some vendors count any visitor interaction, some only conversations the AI handles, some only ones it resolves. Ask for the definition in writing before choosing a tier, because the same traffic can produce wildly different invoices under different definitions.
Will a chatbot actually reduce my support load?
Yes, if it is grounded in real content and your questions are repetitive, which for most small businesses they are. Delivery times, returns, opening hours, and stock questions are ideal. It will not handle anything requiring judgement, and pretending otherwise produces angry customers.
Should the bot pretend to be human?
No. Say it is an assistant, and make reaching a person easy and obvious. Customers forgive a bot that says it does not know; they do not forgive discovering they wasted ten minutes on something they thought was a person.
How much content do I need for grounding?
Less than you think. A good FAQ, shipping and returns policies, and your main product pages cover the majority of routine questions. Start by writing down the twenty questions you answer most, make sure each is answered somewhere on the site, and ingest that.
Do I need AI at all, or will a rule-based bot do?
For many small businesses a rule-based bot answering ten common questions deflects more volume than an AI answering badly, and it costs nothing per conversation. Start rule-based, measure what falls through, and add AI for the long tail.
What deflection rate is realistic?
It varies too much by industry for any published figure to be meaningful, which is why vendors quote conversation volume instead. Measure your own: count conversations that ended without a human, divide by total, and track it monthly. That is the only number that tells you whether the spend is working.
The verdict
If nobody staffs live chat: AI Engine Pro at $79 with your own API key. Under $5 a month in usage for volumes that cost hundreds elsewhere, and you can switch model providers freely as prices change.
If someone does answer chats: Crisp, starting at $45 a month. Per-workspace billing means your whole team is included, the marginal cost per automated conversation is the lowest of the hosted platforms, and you get a real shared inbox alongside the bot.
If the widget is part of your storefront experience: Tidio, budgeting for both meters. It is the most polished product here, and worth the premium if your chat widget is customer-facing on a shop.
Think hard before ChatBot.com unless your volume is genuinely low. Ten resolutions a month on the entry plan and 99 cents per top-up resolution make it the most expensive way to answer a question in this article, and per-user billing compounds that with every staff member you add.
Whichever you pick, write down your twenty most common questions first and make sure your site answers all of them. A chatbot grounded in good content is useful on any platform; a chatbot grounded in nothing is expensive on all of them.
