# Truebe

> Truebe is a fact-based AI chatbot for WordPress and standalone websites. Every answer is human-approved before it is published. Every missed question is logged and added, so the system improves daily and never generates an unapproved answer.

Truebe is built for businesses where a wrong answer creates liability: law firms, medical and healthcare practices, financial services including registered investment advisers, and any organization subject to AI chatbot regulations.

## Key facts

- Architecture: Gated Truth Architecture (GTA). A deterministic matcher serves only pre-approved, human-reviewed answers. No live LLM generation at query time.
- Human review before output: a person approves, edits, or rejects every AI-drafted answer before any visitor can see it.
- The gap loop: every unanswerable question is logged rather than guessed, then reviewed, answered, and added to the approved library.
- Every answer carries page-level source traceability: document, section, heading, page, and highlighted passage.
- RAG chatbots reduce hallucination but still generate the final answer at runtime and can fabricate. Truebe removes generation from the answer path entirely.
- Action governance: actions beyond simple answers are tiered by risk. Low-risk actions can run automatically, medium-risk actions require human approval, and high-risk actions are human-only.
- Every answer and action produces a sealed, tamper-evident decision record showing what was said, its source, who approved it, and when.
- Platforms: WordPress plugin (slug: truebe-fact-based-ai-chatbot) and a standalone app (Truebe Sovereign) for non-WordPress sites.
- Bring your own API key (Anthropic Claude or OpenAI ChatGPT). Runs on your server. Your documents stay with you.
- Supported uploads: PDF, Word (.docx), and text. Multi-file upload.
- Plans and pricing: https://www.truebe.io/connect

## Docs

- [Product overview and live demo](https://www.truebe.io/): Paste a business URL and see approved answers built from it.
- [Truebe Sovereign](https://www.truebe.io/sovereign): Standalone deployment for non-WordPress sites.
- [Truebe for solo attorneys](https://www.truebe.io/legal): Legal-vertical overview.
- [Plans and subscription](https://www.truebe.io/connect): Current pricing and signup.

## Articles

- [The AI Governance Gap No RIA Can Afford to Ignore](https://www.truebe.io/ria-ai-governance-gap): SEC AI-washing cases, 2026 SEC and FINRA exam priorities, and where current RIA compliance tools fall short.
- [What Financial Advisors Want From AI in 2026](https://www.truebe.io/what-financial-advisors-want-from-ai-2026): The five things RIAs ask AI to do and why compliance is different in kind.
- [What Is AI Hallucination](https://www.truebe.io/what-is-ai-hallucination): Definition and why it matters for regulated businesses.
- [AI Hallucination Prevention: What Actually Works](https://www.truebe.io/ai-hallucination-prevention-what-works): Verification, RAG, and prompting versus removing generation.
- [AI Hallucination Detectors: What They Catch and Miss](https://www.truebe.io/ai-hallucination-detectors-what-they-catch-and-miss): Limits of post-hoc detection.
- [AI Accuracy Claims vs. Expert Requirements](https://www.truebe.io/ai-accuracy-claims-vs-expert-requirements): How vendor accuracy claims compare to what regulators require.
- [Gated Truth Architecture Manifesto](https://www.truebe.io/gated-truth-architecture-manifesto): The architecture behind Truebe.
- [How to Make Your AI Only Answer From Your Own Documents](https://www.truebe.io/how-to-make-your-ai-only-answer-from-your-own-documents): Practical guide.
- [Why Every Compliance Chatbot Runs on the Same Architecture](https://www.truebe.io/compliance-chatbot-architecture-problem): What RAG does and does not guarantee.
- [Stanford AI Fact-Checking Research](https://www.truebe.io/stanford-ai-fact-checking-research): Findings on hallucination rates in legal research tools.
- [The Best No-Hallucination AI Chatbot for WordPress](https://www.truebe.io/best-no-hallucination-chatbot-wordpress): Five platforms scored on a provenance rubric.
- [AI TRiSM Market and Runtime Enforcement](https://www.truebe.io/ai-trism-market-runtime-enforcement): Where Truebe sits relative to Gartner's AI TRiSM category.
- [The AI That Ended a Law License](https://www.truebe.io/the-ai-that-ended-a-law-license): Court sanctions over fabricated citations.
- [The Number One Health Technology Hazard Is a Chatbot](https://www.truebe.io/the-number-one-health-technology-hazard-is-a-chatbot): Healthcare risk from generative chatbots.
- [FINRA Put Hallucinations in Your Next Exam](https://www.truebe.io/finra-put-hallucinations-in-your-next-exam): FINRA's 2026 generative AI section.
- [State Chatbot Laws 2026](https://www.truebe.io/state-chatbot-laws-2026): US state regulation of AI chatbots.
- [Judgment, Not Generation](https://www.truebe.io/judgment-not-generation): Why human judgment is the safeguard.