Skip to main content

Core Concepts

CustomGPT.ai Researcher operates through two powerful modules - Deep Research and Content Generation - working in harmony to transform raw topics into comprehensive, well-researched articles. This guide explains how these modules work together and the technology that powers them.

CustomGPT.ai Research Process

Our platform's architecture is designed to:

  • Process 200+ authoritative sources per article
  • Analyze over 1.4M words of content
  • Generate deep research-based articles in approximately 38 minutes
  • Ensure factual accuracy through RAG technology
  • Provide properly cited, hallucination-free content

Research Methods​

CustomGPT.ai Researcher offers two powerful approaches to content generation:

Google Research​

Our AI agent conducts comprehensive research across the internet using advanced search techniques:

  • Analyzes 200+ authoritative sources
  • Processes 1.4M+ words per article
  • Uses advanced Retrieval Augmented Generation (RAG) technology.
  • Ensures factual accuracy via anti-hallucination.
  • Provides proper inline citations

Custom Knowledge​

Use your own documents and data as the foundation for research:

  • Upload internal documents
  • Add specific webpage URLs (eg: your client's or brand's website)
  • Build private knowledge base
  • Maintain brand consistency
  • Ensure confidentiality

Research Methodology​

Our research process follows a systematic approach:

  1. Topic Analysis

    • Breaking down topic research query
    • Identifying key aspects via diverse perspectives
    • Planning search strategy across sources and databases.

  2. Source Gathering

    • Collecting relevant sources from different public Internet sources.
    • Adding sources from diverse perspectives (e.g. "think like Einstein")
    • Adding topic-specific sources (eg: Youtube videos, etc) [Premium Plan]

  3. Content Synthesis

    • Download and process source content (e.g. webpages, PDFs, Youtube Videos, podcasts, etc)
    • Build RAG for semantic search (using CustomGPT.ai)
    • Build Visual Search engine (using Google Images, GPT Vision)

Content Generation Process​

The deep research agent has been optimized to create articles based on the deeply-research sources. Unlike task-based deep researchers (from OpenAI or Perplexity), the CustomGPT.ai has been optimized for article writing - especially when focussed on your custom data.

For example: A marketing agency acting on their client's brand content can greatly benefit from such deep researchers focussed on custom knowledge.

1. Outline Generation​

  • Plan research outline (using o3-mini for highest reasoning)
  • Tune using Chain-of-Thought and diverse perspectives
  • Optimize for language-specific sources and perspectives.

2. Processing Stage​

For each sub-section in research outline:

  • Generate section-specific content.
  • Maintain progressive narrative
  • Add inline citations and sources.
  • Add contextually relevant visual images (using vision)

3. Output Stage​

  • Final document creation
  • Format conversion

Output Formats​

Your research is delivered in multiple formats and notification methods.

Document Types​

  • Browser (.html)
  • Microsoft Word (.docx)
  • Markdown (.md)
  • Google Docs (.gdoc)

Notification Methods​

  • Dashboard : You can check progress in your dashboard app.
  • Email: You will get a notification email with quicklinks (check your spam folder)
  • Google Drive: The final article will be shared with you via Google Docs.
Understanding Processing Time

The 38-minute average processing time allows the deep research agent to:

  • Thoroughly build a plan and find sources
  • Analyze source data and build semantic search index (RAG)
  • Progressively build narrative and deep-researched sections.
  • Generate comprehensive content with visual images.

Anti-Hallucination Technology​

CustomGPT.ai Researcher ensures accuracy through:

  • Source-based generation
  • Semantic search with anti-hallucination.
  • Inline citations.