A website AI chatbot is the cheapest employee a small business can hire in 2026. It answers pricing, shipping, and "do you serve my zip code" questions the second they land, grabs the email of the visitor who would have bounced, and does it at 2 a.m. without complaining. The category has finally matured to the point where a non-technical owner can ship one in an afternoon.
What hasn't matured is the pricing. Every builder in this field meters something different: message credits, per-resolution outcomes, "billable conversations," AI chats vs. regular chats, contacts, seats. The sticker on the pricing page usually isn't the number that lands on your card. We compared the current, official plan pages of the five builders a small business would realistically shortlist, graded them on the buyer decisions that actually move the needle, and picked one to recommend. No hands-on lab this round. This is a documentation and pricing comparison, and the "How We Tested" plan below spells out exactly what that means.
A note on how we got here, because the ranking surprised us.
We went in expecting Intercom Fin to run away with this. On raw AI resolution quality it still does.
Intercom Fin pricing is $0.99 per resolution, with a minimum of 50 resolutions per month
, and the outcome-based model is genuinely one of the more honest billing structures in this category. But sit with the pricing page for a while and the shape of Fin’s bill changes what it’s a good fit for.
You’re only charged once per conversation, even if multiple questions are answered
, which is great, but the rate is flat forever with no cap, and
qualification outcomes cost $9.99 each
. That’s fine at a scale where support is already a real cost center. For a small business trying to catch website leads on evenings and weekends, it’s a bill shape that punishes you for growing.
Chatbase wins because it does one job,
accepts multiple content sources (website URLs, PDF uploads, plain text, Notion, and direct text input), chunks your content, generates embeddings, and stores them in a vector database for retrieval
, and prices it in a way a small-business owner can defend. The credit system needs a little care:
standard models consume 1 credit per response, GPT-5.2 and Gemini 2.5/3.1 Pro cost 2 credits, Grok 3 costs 3 credits, Grok 4 four credits, Claude Sonnet 4.5 and 4.6 burn 3 credits per response, and Claude Opus 4.5 and 4.6 five credits
. Pick an economy model, feed it your real content, and the Hobby or Standard tier is a fair deal for what you get. Its real gap, and it’s honest to name it, is that
it lacks the broader customer support features (ticketing, routing, analytics, team inbox) that Intercom or Zendesk provide. If a customer needs help beyond what your documentation covers, Chatbase has no built-in path to a human agent. For marketing-oriented chatbots or documentation portals, this is fine. For primary customer support, the lack of human handoff is a serious gap.
Tidio slots into second because it’s the one tool here that treats “bot + live chat + inbox” as a single job.
Tidio pitches Lyro as solving up to 67% of customer problems by automatically answering common questions and performing recurring tasks so your team can focus on growing your business.
That’s the vendor’s own number, treat it as a ceiling, not a promise, but the underlying product is well built. What the pricing page hides is that
for more than 2,000 billable conversations, 1,000 Lyro AI conversations, or 100,000 visitors reached with Flows, you have to contact Tidio for a custom plan
, and the three separate meters (base plan, Lyro, Flows) mean the sticker is rarely the bill. If you already know you want a shared inbox and human agents in the same tool as the AI, Tidio’s worth the read. If you don’t, Chatbase is cheaper and simpler.
Botpress lands fourth despite arguably having the most sensible pricing philosophy in the category. Its May 2026 pricing update is worth reading straight from Botpress:
“Previously, we charged per incoming message. We now charge per conversation: any exchange with at least two end user messages, whether handled by a bot, a human, or both. Why? A conversation is the unit that maps to how you actually think about your support volume.”
They also make a specific argument against Fin’s model:
“Conversations are predictable — you know your monthly cost before the month starts. Resolution pricing creates disputes over what ‘resolved’ means and, perversely, raises your bill when your bot gets smarter. With conversation pricing, a better bot means lower headcount cost, not a higher invoice.”
That’s a genuinely good take. It only ranks below Chatbase and Tidio because it rewards a technical builder. If you want a wizard and a widget, this isn’t the wizard.
Landbot is the one to skip unless WhatsApp is genuinely your main channel. The website side of the product is fine, but
WhatsApp automation requires separate Landbot plans, WhatsApp Starter (€80/month) and WhatsApp Pro (€200/month) are complete replacements for standard plans if WhatsApp is your primary channel, not add-ons to existing subscriptions, and the costs extend beyond Landbot’s fees to include Meta’s per-message charges, which vary by message category and geography
. If WhatsApp isn’t your primary channel, you’re paying for a two-track pricing model you don’t need.
One process note before you shop. Every price we quoted here is from the vendor’s own pricing page. No G2, no Capterra listings, no analyst spreadsheets. That matters, because pricing in this category shifts every quarter (Botpress in May, Intercom’s Salesforce acquisition still pending, Chatbase re-tuning credit rates). If you’re signing anything longer than month-to-month, read the pricing page yourself the day you sign. And if you want to build the knowledge base your bot will actually run on, our guide to the best AI SEO tools covers how to produce the source pages, and our roundup of the best AI browsers is worth a read if you want an AI layer sitting on the customer-facing side of your site too.
Pick the tool whose meter matches your traffic. That’s the whole game.
FAQ
What's the best AI chatbot builder for a small business website in 2026?
Chatbase, for most small businesses. It scored 89 on our bench and took Editors' Choice because it's the fastest path from 'we have a website with an FAQ' to 'we have a working AI chatbot on the site,' and its flat monthly plans stay defensible at small-business volume. Tidio is the runner-up if you want live chat and a shared inbox in the same tool.
Which chatbot has the most predictable pricing?
Botpress, by a wide margin in this field. It repriced in May 2026 around conversations rather than incoming messages, bundled AI spend into every plan, and made bots unlimited on paid tiers. The tradeoff is that Botpress rewards a technical builder. If nobody on your team wants to open a flow editor, Chatbase gets you further faster.
Is Intercom Fin worth $0.99 per resolution for a small business?
Only if you already run a serious support operation. Fin is genuinely the strongest AI resolution engine in this roundup, and its outcome-based billing is honest, you're only charged when Fin does the work. But there's no ceiling: a busy month grows the bill in lockstep with volume, and a 'qualification' outcome is $9.99, ten times the resolution rate. For a five-person business with a contact form, Chatbase is a better fit.
Do any of these actually have a free tier that's usable?
Yes, but with caveats. Chatbase's free tier gives you 50 message credits and one agent, but agents get deleted after 14 days of inactivity. Botpress's Pay-as-you-go tier is genuinely free with 500 messages and a small AI credit. Tidio's free plan gives you 50 conversations a month plus a lifetime pool of 50 Lyro conversations. Landbot's Sandbox is capped at 100 chats a month. Every free tier here is a test sandbox, not a production home.
How did you actually test these?
We didn't run a hands-on deflection benchmark. This round was a documentation and pricing comparison. Every score reflects what each vendor's official pricing page and product docs disclosed as of August 2026, graded on five buyer-relevant metrics: time to a working bot, pricing predictability, AI resolution and answer quality, channels and handoff, and small-business value. Every product's official page is linked in Sources.