ChatGPTClaudePerplexityGemini
AI chat output UX compared: ChatGPT, Claude, Perplexity & Gemini
The useful comparison is the deliverable lifecycle: how it is edited, versioned, inspected, and moved into real work.
Teardowns
A closer look at the interface decisions behind AI products and workflows.
Explore teardowns
45 resources
ChatGPTPerplexity
AI citations UX compared: ChatGPT vs Perplexity sources design
A useful sourcing interface supports both a quick check and an examination of the original evidence.
ChatGPTGeminiPerplexity
AI human-in-the-loop UX compared: ChatGPT, Gemini & Perplexity approval flows
Those are different decisions and should not share one ambiguous confirmation control.
ClaudeGeminiPerplexityGrok
AI model picker UX compared: Claude, Gemini, Perplexity & Grok
Outcome labels can lower the vocabulary burden only if their limits are explained honestly.
Notion AILovableManus
AI onboarding UX compared: Notion AI, Lovable & Manus first-value flows
The best first flow depends on available user context, not on whether a wizard looks polished.
ChatGPTClaudePerplexity
AI personalization UX compared: ChatGPT, Claude & Perplexity memory design
The design decision is how people inspect and correct influence over future answers.
ManusDeepSeekPerplexityLovable
AI pricing & upgrade UX compared: Manus, DeepSeek, Perplexity & Lovable
An upgrade prompt should explain the actual boundary without implying that payment guarantees a better answer.
Alexa for shopping
Alexa for shopping: conversation, price alert, and auto buy
A progression from advice to autonomous purchase increases the need for distinct consent and budget boundaries.
Google Maps
Ask Maps by Google Maps: chat beside the map
A shared map can anchor a conversation, but assistant cards and ordinary place details must refer to the same entity clearly.
YouTube
Ask YouTube: guides built from video
Synthesis can reduce browsing effort if viewers can trace a recommendation back to the relevant clip.
ChatGPT
ChatGPT citations UX: inline chips, popovers & sources
A layered inspection path can support quick reading and deeper verification without treating source count as proof of accuracy.
ChatGPT
ChatGPT composer UX: input bar, tools & voice design
A consistent input anchor helps discovery, while active tools need visible scope before submission.
ChatGPT
ChatGPT Human-in-the-Loop UX: Clarify & Handoff Patterns
Clarification narrows intent, but it is not necessarily approval for the final action.
ChatGPT
ChatGPT Intelligent UI: generative interfaces in the thread
Inline interfaces can reduce context switching, but generation must stay inside validated components and application permission boundaries.
ChatGPT
ChatGPT Output UX: Response Refinement & Editing
An in-thread editing loop benefits from clear version ownership and a way to target only part of an answer.
ChatGPT
ChatGPT personalization UX: tone presets, memory & profile design
Layered preference controls can become contradictory unless their combined effect is understandable.
ChatGPTClaudePerplexityGemini
ChatGPT, Claude, Perplexity and Gemini composer comparison
A shared task reveals meaningful trade-offs more reliably than counting composer buttons.
Claude
Claude artifacts UX: split pane & preview design
A separate workspace can clarify reusable output if the user always knows which artifact and version are active.
Claude
Claude composer UX: model picker, skills & attachments
Separating input context from model selection helps organize complexity, but users may still confuse tools with model capability.
Claude
Claude output UX: replies, retry & feedback
Different refinement actions can be efficient, but their consequences must be distinguishable to a first-time user.
Claude
Claude UX Teardown: Personalization & memory
Users need a coherent explanation of which preference affects the current conversation and which persists elsewhere.
DeepSeek
DeepSeek composer UX: DeepThink, Search & modes
Showing the speed/depth trade-off can help set expectations if the interface explains what changes for the task.
Gemini
Gemini composer UX: tools, Drive & thinking modes
Distributed controls may scale capability, but they also require clear boundaries between input type and execution mode.
Gemini
Gemini Human-in-the-Loop UX: Gmail Draft & Send Gate
The final gate should reveal recipient, content, and account identity, especially when generation and sending are close together.
GLM (Z.ai)
GLM (Z.ai) composer UX: Magic Design, Deep Think & Agent
A familiar default can coexist with powerful tools if mode labels explain the consequence of switching.
Google AI Mode
Google AI Mode Interface Design: Composer UX
An expanding tool surface needs clear mode identity and a path back to ordinary search.
Gemini
Google Gemini Output UX: Tables, Sources & Export
Export can be a useful next step, but users need to understand whether they are copying, creating, or modifying an external artifact.
Grok
Grok composer UX: skills, models & voice input
The composer risks becoming a product menu unless the default task stays obvious.
Kimi
Kimi composer UX: agent pills, Instant/K3 & skills
Multiple depth and capability labels create a teaching burden unless each answers a different user decision.
Lovable
Lovable Interface Design: Build Composer UX
A planning mode is useful only if users can tell whether it changes files or merely proposes work.
Lovable
Lovable onboarding UX: setup wizard & segmentation
Early questions can tailor the workspace, but each adds delay before a user experiences value.
Manus
Manus UX Teardown: Composer
Task-oriented framing may help people state an outcome, but mode changes need a visible authority boundary.
Manus
Manus UX Teardown: Onboarding
Segmentation should improve the immediate task rather than merely collect profile data.
Manus
Manus UX Teardown: Output & artifacts
A deliverable should be visibly different from an agent's work-in-progress update.
Notion AI
Notion AI onboarding UX: prompt starters & first value
Contextual examples can teach capabilities more effectively than a separate tour.
Perplexity
Perplexity Citations UX: Source Chips & Trust Signals
Verification and refinement should remain distinct actions so checking evidence does not accidentally alter the answer.
Perplexity
Perplexity composer UX: search modes & connectors
Context and capability need separate identities so a user can predict what the next search will inspect.
Perplexity
Perplexity Human-in-the-Loop UX: Connector Permissions
Capability-level controls can reduce overbroad authorization, but users need to understand duration and revocation.
Perplexity
Perplexity output UX: tabs, follow-ups & export
Multiple views can reduce repeated searches, but the source-of-truth result must stay consistent between them.
Perplexity
Perplexity personalization UX: vertical profiles, response style & memory
Separate settings can clarify what was explicitly supplied versus inferred or retained, but terminology must explain that distinction.
Perplexity
Perplexity Shopping UX: Discovery Feed & Instant Buy
A short purchase path still needs unmistakable merchant, total price, payment, and confirmation boundaries.
Qwen
Qwen composer: a quiet bar for a multimodal studio
Progressive discovery can protect a simple entry point, but a hidden capability menu may become the real navigation.
Qwen
Qwen output: keep the chain in the thread
Keeping related transformations together preserves context, but versions and generation progress must remain distinguishable.
Knoveya
A library built for discovery
What a multi-format AI reference can teach us about navigation and wayfinding.
Knoveya
Animate UI: let the interface demonstrate
An independent reading of how an animation library explains itself through its own controls.