Build a no-code chatbot with AI in 2026: how far it works, which tools, what it costs
No-code platforms can put a chatbot trained on your help docs or website live in an afternoon. Getting it to answer only from those documents, resist manipulation and stay within consumer law is harder. A tribunal has held a company to what its chatbot said, so plan for testing, guardrails and a human hand-off before launch.
Where AI stands: Build a no-code chatbot · September 2026
As of September 2026, you can build a useful support chatbot from your own documents without code, but it will still invent some answers, it can be manipulated, and you are responsible for what it says.
- Bots trained on your docs and website
- Answers limited to your sources
Backed by trials or large deployments
- Custom GPTs with knowledge files
Promising, limited outcome data
- Guaranteed factual answers
- Protection from prompt injection
- Passing legal responsibility to the bot
Evidence says avoid or add safeguards
What works today
- 1
Proven Train a support bot on PDFs, Word and text files, a website crawl, Notion pages and custom Q&A pairs without writing code.
Evidence
Chatbase documentation (checked Sept 2026). Each agent is capped at 33 MB of training content, and files need selectable text. Source ↗
- 2
Proven Ground answers in up to 25 public websites or SharePoint URLs, plus uploaded documents, and block answers when no source is found.
Evidence
Microsoft Learn, Copilot Studio knowledge sources (updated August 2026). Answers are blocked when 'Allow ungrounded responses' is off. You must verify website ownership. Source ↗
- 3
Early Build a custom GPT in ChatGPT with instructions and up to 20 knowledge files, using no code.
Evidence
OpenAI Help Center (2026). Files can be up to 512 MB each. OpenAI advises testing in Preview and using text-forward files because complex layouts are harder to use. Source ↗
What doesn't work yet
- 1
Can't Guarantee factual answers. Even when given source text, current models sometimes add unsupported details.
Evidence
Vectara leaderboard (May 2026). Hallucination rates when summarising documents ranged from 1.8% to 24.2%. The risk is higher for questions your docs do not cover. Source ↗
- 2
Can't Fully prevent prompt injection, including hidden instructions in crawled pages or uploaded files. Retrieval over your own docs does not solve it.
Evidence
OWASP LLM01:2025 names prompt injection the top LLM risk and states that RAG and fine-tuning 'do not fully mitigate' it. Source ↗
- 3
Can't Shift legal responsibility onto the bot. Companies can be held liable for what their chatbot tells customers.
Evidence
Moffatt v. Air Canada (BC tribunal, February 2024). The tribunal rejected the argument that the chatbot was a separate entity and awarded about CA$650 for a wrong bereavement-fare answer. Source ↗
Where to start
Small business
Start with a bot that answers only from your FAQ and policy pages, with an "ask a person" fallback.
Support team
Test with off-topic and edge-case questions before launch, and route price, refund and contract questions to staff.
IT or security lead
Crawl only sources you control, and limit the tools and actions the bot can call.
Tools for this task, by pricing model
Freemium
| Tool | Pricing model | Free tier | Starting price | Compliance claims | Official site | Last verified |
|---|---|---|---|---|---|---|
 Botpress - Faster build chatbots with ChatGPT | Freemium | Yes: $0: 25 conversations, no top-ups or overages; 3 seats, 3 AI agents; community support | $189 / month | 4 : SOC 2 Type II, GDPR, CCPA, DORA | Online (2026-09-26) | 2026-09-26 |
 DocsBot | Freemium | Yes: Free for personal use: 1 DocsBot, 50 source pages, 100 monthly AI credits, 1 user.… | $49 / month | 2 : SOC 2, GDPR | Online (2026-09-24) | 2026-09-26 |
Usage-based
| Tool | Pricing model | Free tier | Starting price | Compliance claims | Official site | Last verified |
|---|---|---|---|---|---|---|
 Quidget | Usage-based | See tool page: 'Start for Free' CTA; pay only for AI-resolved conversations. Free allow… | Not verified | 2 : GDPR, CCPA | Online (2026-09-24) | 2026-09-26 |
Paid
Pricing not verified yet
| Tool | Pricing model | Free tier | Starting price | Compliance claims | Official site | Last verified |
|---|---|---|---|---|---|---|
 Dialogview - AI customer support provider | Not verified | Not verified | Not verified | 1 : GDPR | Online (2026-09-24) | 2026-09-26 |
Typical cost
Chatbase Hobby is $40/month (700 message credits), and DocsBot Personal is $49/month. Botpress Plus is $150/month billed yearly ($180 monthly) for 250 conversations plus a $25 AI credit. Copilot Studio is $200/month per 25,000 Copilot Credits, or pay-as-you-go through Azure.
Usage limits, not the headline fee, usually drive the bill. Prices were checked in September 2026.
Data & compliance requirements
Laws, rules and official guidance that commonly apply when AI is used in Build a no-code chatbot.
| Name | Applies to | Official source | When it applies |
|---|---|---|---|
| California BOT Act (B&P Code §17940) | US-CA | leginfo.legislature.ca.gov | Bots interacting with consumers must disclose that they are bots |
| EU AI Act Art. 50 | EU | digital-strategy.ec.europa.eu | Chatbot transparency |
| TCPA | US | fcc.gov | AI voice agents making outbound calls |
| GDPR / CCPA-CPRA | EU / US-CA | cppa.ca.gov | Conversation logs and personal data |
| PCI DSS | Industry standard | pcisecuritystandards.org | Collecting card numbers during support conversations |
| Accessibility (ADA / WCAG) | US / International | w3.org | Public-facing service interfaces |
This list is a starting point for your own compliance review, not legal advice. Last reviewed 2026-09-24.
Failure modes
- Confident answers to questions your docs don't coverRestrict the bot to grounded answers, add an 'I don't know, contacting a human' fallback, and test with off-topic and edge-case questions.
- Bot states wrong prices, refund rules or policiesKeep policy pages current, exclude outdated files, and route money-related or contractual questions to staff. Your company is bound by what the bot says.
- Users or poisoned pages override the bot's instructionsCrawl only sources you control, limit the tools and actions the bot can call, and review logs for manipulation attempts (OWASP LLM01).
- Chat logs retained longer than you or customers expectCheck the vendor's retention and training settings and its data processing agreement. A 2025 US court order temporarily required OpenAI to keep deleted consumer and API chats.
- Costs spike when conversations exceed plan limitsSet caps or alerts on auto-recharge. Several platforms add overage packs automatically at 95% of quota.
Recent changes: tools for this task
Last verified: 2026-09-25 · auto-checked 2026-09-26 · Reviewed by Toolsfine editorial
- Riley · AI alternative note · — → A general LLM can answer parenting qu… ·
- Riley · editor verdict · — → Riley: baby tracker and AI parenting … ·
- Riley · integrations · — → ["iOS app","Android app"] ·
- Riley · pricing page URL · — → https://rileyapp.com/pricing ·
- Riley · starting price · — → 9.99 ·
- Riley · free tier · — → {"exists":true,"limits":"Riley Basic:… ·
- Riley · pricing model · — → Freemium ·
- Quidget · AI alternative note · — → A general LLM cannot sit in your webs… ·