Users watching real-time AI responses often feel anxious when they can't tell if the system is thinking, stuck, or has finished processing their request.
How Apps Solve It
ChatGPT displays a pulsing cursor during generation, then shows "Stop generating" button while text streams in word-by-word. Google's Bard takes a different approach—it shows three animated dots while processing, then reveals the complete response at once with a subtle fade-in effect. Bing Chat combines both strategies: it shows "Searching" and "Generating" status labels, streams text progressively, and displays source citations in real-time as numbered references appear. Indian apps are catching up—Flipkart's new AI shopping assistant shows a typing indicator with "Flippi is thinking..." text, while PhonePe's AI support bot displays percentage-based progress (like "Understanding your query... 40%") before streaming responses.
The Design Pattern
This is called Progressive Disclosure with State Indicators—showing users exactly what stage the AI is in through visual and textual cues. When apps get this wrong, users repeatedly click submit thinking nothing happened, abandon the interface assuming it's broken, or miss partial responses because they don't realize generation is still ongoing.
Why It Matters
Users need constant reassurance that expensive AI operations haven't failed silently. Designers debate whether showing partial responses reduces perceived latency or creates confusion when AI self-corrects mid-stream—especially critical when dealing with financial or medical information where incomplete answers could be dangerous.
Key Terms
- Streaming Response: AI output displayed word-by-word as it's generated
- State Indicators: Visual elements showing processing, generating, or complete status
- Perceived Latency: How long users feel they're waiting versus actual processing time
- Progressive Disclosure: Revealing information gradually rather than all at once
- Token Streaming: Technical term for how AI models output text piece by piece
Practice on ProMode
Compare how ChatGPT's word-by-word streaming differs from Google Bard's complete response approach. Practice explaining the trade-offs on ProMode with AI voice calls.