Streaming Preview: Why You See Two Formats

Great question! What you're seeing is actually a feature called Streaming Preview — and it's working exactly as designed. When GPTWeb begins generating a response, it streams the raw AI-generated text to your screen in real time as tokens arrive from the language model. This gives you immediate feedback that the system is working and lets you start reading the answer before it's fully complete. During this phase the text appears in a simpler, flowing format because it's a live preview of content that hasn't finished rendering yet.

Once the AI finishes generating the full response, GPTWeb seamlessly transitions from the streaming preview into the final fully structured, rich-text format — complete with styled tables, headers, bullet lists, Mermaid diagrams, code blocks, charts, and all the visual formatting you'd expect. This two-phase approach (raw stream → rich render) is intentional: it dramatically reduces perceived wait time, especially for longer, more detailed responses. The transition is designed to be smooth with no visible flash, so ideally you shouldn't notice a jarring jump — just a natural upgrade from preview to polished output.

During the streaming phase, GPTWeb also renders tables incrementally — showing styled HTML tables on desktop and stacked card layouts on mobile as the data streams in. A small footer during streaming shows the token count, generation speed, and a stop button if you want to halt a long response early. So the "terrible format" you're seeing first is really just the raw ingredients, and the rich text is the finished dish — GPTWeb just lets you watch it being assembled in real time rather than making you wait for the whole thing to be ready. If you'd like to explore more about how GPTWeb's Knowledge Base (RAG) and Agentic Workflow power these responses, those are great next steps!