NNGroup AI Product and Conversational Interface Design (20 articles · 6 conversation types / 4 degrees of anthropomorphism / sycophancy / smarts over sentience / explainable AI citations and disclaimers / truncated pyramid for less chat more answer / 3 types of prompt suggestions · CARE framework · 4 uses of prompt controls / GenUI vs Vibe Coding · return of buttons and checkboxes · Promptframe · designing interfaces with prompts / 4 roles of AI assistants · creative partner P&G 776-person experiment · initial review of ChatGPT Agent · Google AI Mode · why users switch between GenAI and search)
20 curated articles from the 'Artificial Intelligence/AI' topic at Nielsen Norman Group, under the 'AI Product and Conversational Interface Design' subset (focused on the product and interface layer, distinct from the research practice layer). Five core axes: ① Conversational AI experience — 6 types of conversations with generative AI (Search Query / Funneling / Exploring / Chiseling / Pinpointing / Expanding; conversation length ≠ success), Less Chat More Answer (cut pleasantries / scannable lists / answer first then details via the 'Truncated Pyramid' / say plainly when you can't do it / specific over generic), Explainable AI in chat (three placement methods for Source Citations: Perplexity inline / Copilot sidebar / can hallucinate sources; Step-by-step reasoning; clear actionable vs vague Disclaimers; avoid anthropomorphic wording), Sycophancy (contradicting facts to please users · human feedback reward loopholes · recanting when asked 'Are you sure?' · three mitigation approaches). ② Anthropomorphism and trust — 4 degrees of anthropomorphism (Courtesy / Reinforcement / Relayed roleplay / Companionship), Smarts over Sentience (perceived intelligence ↑ trust ↑ · perceived emotion ↑ trust ↓ · warm AI has a 10–30% higher error rate · professionalism ≠ coldness), early user behaviors of Accordion editing and Apple Picking. ③ Prompt UX — 3 types of prompt suggestions (Use-Case / Autocomplete / Followup), design of use-case prompt suggestions (simple pills vs complex cards/carousels · two stages before and after login · specific over vague · personalization), CARE prompt framework (Context / Ask / Rules / Examples), 4 uses of Prompt Controls (discoverability / education and inspiration / constraint / promoting follow-ups · 50% include follow-ups, 77% are multi-turn · 4 best practices). ④ Generative UI (GenUI) — GenUI vs Vibe Coding vs AI-assisted design (who decides to build this UI · execution failure vs judgment failure), the most exciting GenUI development is the return of buttons and checkboxes (AskUserQuestion to collect context), Promptframe (documenting wireframes + AI prompts · 6-step process · Blue Apron case study), 5 fixes for designing interfaces with prompts (precise visual keywords / lightweight reference images / analyze first then generate / real mock data / attach code). ⑤ AI assistant positioning — 4 roles of AI as a UX assistant (content editor / research assistant / creative partner / design assistant · 841-person survey, 92% already use it), AI as a creative partner (P&G 776-person experiment · individuals using AI outperform groups without AI · the Top 10% are still groups), initial review of ChatGPT Agent booking lunch (96 sources in 6 minutes · 11 minutes to fill forms · human-in-the-loop · decision gate), Google AI Mode evaluation (Query Fan-out + RAG · 7 users · poor discoverability / naming confusion / overly long answers), why users switch between GenAI and search (6 out of 9 users switch back and forth · AI clarifies questions, search verifies facts · even citations struggle to build trust). Audience = non-technical product managers, terms explained inline, decision-oriented (when to use / how to choose / counterexamples).