# BASIC UX principles UX design principles for systems that people and AI agents use. Five principles, each asked of both users: Beauty, Accessibility, Simplicity, Intuitiveness, Consistency. First published by Dan Smith in UX Magazine on June 29, 2016. Version 2.0, updated 2026-10-15. License CC BY 4.0. https://basicux.com Each principle asks its questions of two users: people, who see the screen, and agents, which read the accessibility tree, the HTML and any machine paths the site offers. A third question covers products where an agent acts on a person's behalf. Reader standard: Text is written for a reader with ADHD, dyslexia or another learning difference, and for the least schooling a reader might have: reading grade 6 or below, about 15 words a sentence and none over 25, paragraphs of 1 to 3 sentences, headings of 8 words or fewer, literal words, every term explained once. ## Position - One build serves both users. The accessible page is the agent-ready page. - The people questions always come before the agent questions. - No separate layer for agents: people and agents get the same content, prices and labels. ## Beauty **People: Is it aesthetically pleasing?** A good-looking design feels easier to use. Designers use style guides so small flaws do not make the product look broken. - Is it aesthetically pleasing? - Does it follow the style guide? - Are high-quality visuals used? - Is it properly aligned? **Agents: Is the page code clean?** AI agents read page code and look at screen pictures. Clean code and a steady layout help agents find buttons without mistakes. - Does the screen order match the code order? - Does the layout stay still with nothing covering any button? - Is the HTML page code valid with 1 h1 heading and a title? **When an agent acts for a person** - Can a person check the agent's plan and results fast? Maggie Appleton wrote in 2026 that people need plans they can see. Sources: [web.dev, Build agent-friendly websites (2026)](https://web.dev/articles/ai-agent-site-ux); [Maggie Appleton, Planning with Agents (2026)](https://maggieappleton.com/planning-agents) ## Accessibility **People: Can everyone use it?** Accessibility means all people can use the product. Designers must build for people with poor sight, hand limits or other disabilities. - Can everyone use it? - Does it comply with standards? - Is it cross-platform compatible? - Can people with ADHD, dyslexia or other learning differences read the text? **Agents: Can an agent read and use it?** In a 2026 AudioEye study, AI agents finished 96% of tasks on an accessible page. They finished 31% on a page that was not. - Does every button have a role and name in the accessibility tree? - Is the text in the HTML code before JavaScript runs? - Can an agent do public tasks without CAPTCHA or robots.txt blocks? **When an agent acts for a person** - Can the person stop the agent or take over anytime? Luke Wroblewski wrote in 2025 that people must start and stop agents. Sources: [Nielsen Norman Group, AI Agents as Users (2026)](https://www.nngroup.com/articles/ai-agents-as-users/); [AudioEye, AI agent task completion on inaccessible websites (2026)](https://www.prnewswire.com/news-releases/audioeye-study-finds-up-to-68-drop-in-ai-agent-task-completion-on-inaccessible-websites-302888585.html); [Microsoft, Guidelines for Human-AI Interaction (2019)](https://www.microsoft.com/en-us/haxtoolkit/library/); [Luke Wroblewski, Agent Management Interface Patterns (2025)](https://www.lukew.com/ff/entry.asp?2106=) ## Simplicity **People: Does it make life easier?** More choices make people slower to decide. Show only what people need for their task, and use short, plain words. - Does it reduce the user's workload? - Is it free of clutter and repetitive text? - Is its functionality necessary? - Is the text short and plain, with 1 idea per sentence? **Agents: Does it take the fewest steps?** Elnaffar and Rashidi cut agent steps by 30% in a 2026 store test. You can cut steps with an API, which is a direct software link. - Is there a short path like llms.txt, JSON-LD, an API or WebMCP? - Does each form or tool do 1 job? - How much text must the agent read to finish the task? **When an agent acts for a person** - Is checking the agent's work easier than doing the work yourself? Anthropic found in 2026 that skilled users watch agents instead of approving steps. Sources: [Elnaffar and Rashidi, Designing Agent-Ready Websites for AI Web Agents (2026)](https://arxiv.org/abs/2607.12056); [The /llms.txt file](https://llmstxt.org/); [WebMCP draft specification](https://webmachinelearning.github.io/webmcp/); [Anthropic, Measuring AI agent autonomy in practice (2026)](https://www.anthropic.com/research/measuring-agent-autonomy) ## Intuitiveness **People: Is it easy to learn?** A product should be easy to learn so people do not have to learn it again. Use common button shapes and link colors. - Is the functionality clear? - Can the user reach their goal with little or no instruction? - Can the task be repeated without further instruction? - Can the user predict the outcome? - Can a reader with the least schooling understand it the first time? **Agents: Does it say what it does?** AI agents need every action named in plain text. In a 2026 MERJ test, a bad label made an agent click Cancel 25 times. - Does every button, link or tool name its exact action? - Does text stay on the page to say what happened? - Are actions that cannot be undone clearly marked? **When an agent acts for a person** - Does the person see the plan first, with a way to undo? Victor Yocco wrote in 2026 that users need to see plans and undo actions. Sources: [MERJ, How AI Agents Use Websites, Where They Fail, and What to Fix (2026)](https://merj.com/blog/agent-interaction-research); [Victor Yocco, Designing for Agentic AI, Smashing Magazine (2026)](https://www.smashingmagazine.com/2026/02/designing-agentic-ai-practical-ux-patterns/) ## Consistency **People: Does it match the system?** Keep colors, spacing, fonts and forms the same across the product. Reuse the same parts so people know what to expect. - Does it reuse existing patterns and designs? - Is its language, imagery, and branding consistent with the system? - Does it appear in the right place at the right time? - Does it perform consistently every time? **Agents: Does it act the same every time?** Jakob Nielsen wrote in July 2026 that people and AI agents need the same view. Giving both the same prices stops mistakes. - Do people and agents get the same text, prices and labels? - Does the same name always go to the same web address? - Does the API or tool do the same thing as the screen? **When an agent acts for a person** - Does it say it is AI and ask before spending, deleting or sharing data? OpenAI says a person should check paid or permanent agent actions. Sources: [Jakob Nielsen, UX Roundup: Web Design for AI Agents (2026)](https://www.uxtigers.com/post/ux-roundup-20260731); [OpenAI, A practical guide to building agents (2025)](https://cdn.openai.com/business-guides-and-resources/a-practical-guide-to-building-agents.pdf)