> ## Documentation Index
> Fetch the complete documentation index at: https://docs.coloop.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Comparison with ChatGPT

| Feature                               | CoLoop                                                                                                                                                                                            | ChatGPT                                                                                                                                                                                                                          |
| ------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Integrations**                      | Recollective; Incling; Excel Formats; Transcript Formats; Audio and Video; Zoom\*; MS Teams\*; G Meet\*; Translation built in Even more coming soon...                                            | Some capabilities here inherently and via plugins with 3rd parties; no file management capabilities; all files are treated as binaries or text so has no notion of: speakers; tasks; activities; questions; segments or filters. |
| **Transcription accuracy**            | 43% fewer WER than Whisper on real world data                                                                                                                                                     | Lower accuracy; uses Whisper; also can’t handle speaker labelling                                                                                                                                                                |
| **Analysis grids**                    | Filter; compare; contrast and analyse across documents. View everything in a familiar UI; Suggested questions and minimal learning curve. Optimised to extract themes; also provides theme counts | Limited to chat UI; Doesn’t reliably cite sources                                                                                                                                                                                |
| **Research Specialisation**           | Uses discussion guides; background information and speaker roles to 10x performance on qual analysis task                                                                                         | Generalised tool                                                                                                                                                                                                                 |
| **Clip Reels**                        | Can easily select and generate clips in 1 click (up to 6Gb per file)                                                                                                                              | Can be done via code interpreter for small video files (less than 512MB per file)                                                                                                                                                |
| **Segmentation**                      | Can handle participant segments and answer questions about them                                                                                                                                   | Has no notion of speaker roles or segments; highly unreliable at distinguishing between them                                                                                                                                     |
| **Support**                           | Highly available support team                                                                                                                                                                     | No support                                                                                                                                                                                                                       |
| **Context limits / accuracy**         | Backed by a qual specific data model and search engine; reduces, enriches and constructs evidence before responding Proven accuracy up to 100 hours worth of interviews.                          | Fills entire context with as much info as possible regardless of its relevance (max 100,000 words \~ 10 hours audio). Good at answering questions about single documents but prone to hallucinate when working across many.      |
| **Vendor lock in**                    | Will always use best available models whoever they are from: Gemini / Claude / Mistral etc. Also includes hybrid / agentic approaches e.g. Gemini + GPT4 or Mistral + Claude.                     | Will only ever use OpenAI models                                                                                                                                                                                                 |
| **Data privacy**                      | Will always guarantee this; Key to our operating model; fully insured                                                                                                                             | Frequently change their terms and are subject to competing interest for their customer facing products                                                                                                                           |
| **Access Management & Collaboration** | Can access manage projects and share resources; files and more across teams selectively; will continue to develop specifically for agency work.                                                   | Limited collaboration capabilities                                                                                                                                                                                               |
