NNGroup: AI glossary for design and product teams
Caleb Sponheim at Nielsen Norman Group published this glossary on August 21, 2026, as a practical reference for designers and product people who encounter AI terminology in their daily work but are not researchers or engineers. The resource covers approximately 80 terms, arranged alphabetically, with cross-references to related concepts and links to deeper NN/G articles where they exist.
The vocabulary covered spans the range a practitioner would actually encounter. Core concepts include LLMs, foundation models, and multimodal models. On the prompting side, the glossary covers chain-of-thought, few-shot and zero-shot prompting, and prompt engineering without assuming prior ML background. System-level terms — embeddings, tokens, context windows, inference, RAG — get straightforward definitions that explain what they mean for how an AI product behaves, not just how the engineering works. Product-level entries cover agents, tool use, generative UI, and memory. The safety and quality section defines hallucination, bias, guardrails, alignment, evals, and prompt injection in terms relevant to deciding what to build and how to test it.
A section on UX risks is particularly useful for practitioners. Automation bias, overreliance, sycophancy, and anthropomorphism each get their own entries, framed in terms of what can go wrong when users interact with AI systems rather than in terms of model behavior. This makes the glossary relevant not just for building AI features but for evaluating them and communicating about them to stakeholders.
The value of a shared vocabulary in this domain is hard to overstate. Design teams that cannot consistently distinguish between a model and an agent, or between retrieval-augmented generation and fine-tuning, tend to make poor scoping decisions and struggle to communicate requirements to engineering partners. This glossary does not resolve those problems on its own, but it gives teams a common reference point when terminology disagreements slow down a conversation or create misalignment around what a feature will actually do.
A useful reference to bookmark for design and product professionals at companies building on AI systems. Teams working on AI-native products will likely find they already know many of the terms, but the UX risk section is worth reviewing regardless of experience level.