Frequently Asked Questions About CustomerGPT

CustomerGPT uses strict context-grounded Retrieval-Augmented Generation (RAG). The LLM is mathematically constrained to answer solely from your indexed documentation and provides direct source citation links. If a user asks for discounts or policies not found in your data, the bot politely declines and offers human escalation rather than fabricating answers.

Yes! In the CustomerGPT admin panel, you can enter plain text or copy-paste content directly. The backend immediately processes the raw text, splits it using a sliding character window (800-character window size with 150-character overlap), and vectorizes it to update your chatbot's knowledge base.

CustomerGPT connects to Shopify in under 2 minutes via a single script tag or plugin. It auto-indexes your store FAQs, shipping policies, and return guidelines, resolving up to 80% of repetitive order status, sizing, and return questions 24/7 without agent intervention.

We support a wide range of training content: website URLs, document file uploads (PDF, DOCX, TXT, Markdown), direct text inputs, and automatic integrations with Google Drive, Notion, Confluence, Dropbox, OneDrive, SharePoint.

Yes, automatic retraining depends on your tier. Growth accounts trigger monthly syncs, Scale accounts run daily scans with weekly syncs, and Enterprise plans support instant webhook-triggered syncs that update the knowledge base as soon as your website changes.

Yes, you can upload PDF, DOCX, TXT, or Markdown files. When you upload a file, the backend processes it asynchronously, extracting and cleaning the text, chunking it, generating 1536-dimensional embeddings, and saving them to our database.

After training is complete, the dashboard generates a single copy-paste HTML script block. You can paste this script tag into the head or body of your website. It runs on any HTML/JS site, including WordPress, Shopify, Webflow, and React/Next.js.

Training takes from a few seconds up to a few minutes, depending on the volume of text. Because the crawler and chunking engine process data asynchronously in a background promise, your dashboard remains fully interactive during training.

Intercom charges unpredictable fees ($0.99 per AI resolution on top of seat costs), which quickly escalates support bills as ticket volume grows. CustomerGPT provides predictable flat monthly pricing starting at $49/month with zero per-resolution fees, saving businesses 60% to 80% in operational costs.

Yes! We offer a 14-day free trial with no credit card required, allowing you to build, train, and test your chatbot on your own data. Paid plans also come with a 14-day money-back guarantee.

Yes, you can embed your chatbot across multiple websites or domains under all our plans. The single script block works everywhere and shares the same chatbot knowledge base.

Yes, our agency features are available on the Scale ($349/mo) and Enterprise tiers. These include isolated tenant sandboxes, custom brand customization, white-label options, and configurable rate limits for each chatbot under one master admin.

A 'page' corresponds to a single crawled website URL or an uploaded document containing up to 10,000 characters. Our Service automatically monitors these character and page consumption counts.

Yes, white-labeling is supported on our Scale and Enterprise tiers. You can completely remove 'Powered by CustomerGPT' branding, configure a custom subdomain for your widgets, and override colors/styles.

CustomerGPT utilizes OpenAI's GPT-4o and GPT-4o-mini models for natural language response generation. Document chunking is mapped to 1536-dimensional vector spaces via text-embedding-3-small and indexed using pgvector with HNSW for sub-50ms retrieval.

If a user query falls outside indexed knowledge or receives a low confidence score, CustomerGPT triggers an autonomous fallback: it can collect visitor contact details or route the conversation in real-time to Slack, Zendesk, or Telegram.

Yes, CustomerGPT supports over 95 languages with automatic detection. Even if your documentation is written in English, the AI agent can read, reason, and answer fluently in Spanish, French, German, Japanese, Chinese, or Arabic.

Yes, CustomerGPT natively integrates with Zendesk, Slack, Telegram, HubSpot, Shopify, and 5,000+ web applications via Zapier webhooks and REST APIs.

Yes. All uploaded files and crawled pages are stored in isolated tenant sandboxes with PostgreSQL Row-Level Security (RLS). Your proprietary data is encrypted using AES-256 at rest and TLS 1.3 in transit, and is never used to train public foundation models.

Yes. CustomerGPT adheres to GDPR guidelines (with Data Processing Agreements available on all paid plans) and HIPAA security controls. You retain 100% data ownership with instant purge and export options.

CustomerGPT implements a multi-layer defense topology compliant with OWASP Top 10 for LLM Applications: cryptographic HMAC-SHA256 CAPTCHAs, pre-prompt semantic classification, and strict system boundary prompts that prevent prompt leakage or instruction override.

Our core vector databases and inference gateways operate on secure AWS cloud infrastructure located in redundant US regions, with global edge routing for sub-second response times worldwide.