Site Chatbots: Answer Fast, Skip the Chat
Users treat site AI chatbots like search bars—short queries demand direct, scannable answers without small talk, fluff, or overload. Use truncated pyramid: essentials first, details via prompts.
Match Short, Imperfect Queries with Direct Responses
Users approach site AI chatbots expecting instant answers, typing minimal, keyword-like prompts without greetings, politeness, or perfect grammar. In a study of 9 participants across 8 chatbots (2–3 per user), queries started as full sentences but quickly shortened to phrases like "Need a car for three people. Going to Orlando, FL, from Hampton, Georgia" (Turo), "What are the fees?" (Scouting America), or "Do you sell pavers?" (Home Depot). Typos didn't hinder understanding, building trust for even briefer followups.
Avoid sycophantic filler like "great question!"—it annoys users seeking tools, not relationships. Home Depot's Magic Apron excelled by delivering answers without pandering, earning praise: "I just want the information." This directness respects typing effort and mirrors search bar behavior, boosting efficiency.
Format for Scannability: Bullets, Bold, Short Paras
Chat viewports amplify text density, so apply web-writing rules strictly: sentences under 20 words, paragraphs 2–3 sentences max, plus lists, bold, headers, and whitespace. Mississippi's Ask MISSI overwhelmed with unformatted paragraphs filling the viewport, especially during streaming, causing users to disengage: "The pouring in of information made me feel overwhelmed."
Contrast with successes: Scouting America's Scoutly gave concise fee breakdowns without preamble, using bullets for fine print. Williams Sonoma formatted long cooking tips as bulleted lists, prompting delight: "I love that they're bulleted, not one big paragraph." Being concise trims nonessentials while retaining utility—formatting prevents even helpful content from feeling exhausting.
Truncated Pyramid: Essentials Upfront, Details on Demand
Ditch inverted pyramid for chatbots; use truncated pyramid—deliver only the asked-for answer plus accuracy caveats first, then suggest prompts for extras like context or steps. Olympic site's overload on a simple "Who did the flip?" (scores, background) frustrated users wanting just a name, unlike ChatGPT's bullet-first approach.
For ambiguity, ask sparse clarifications to avoid wrong answers, then stick to basics. When unable, state plainly upfront without padding: Turo wasted time vaguely explaining manual search instead of admitting limits; Redfin buried filtering options after a "can't help" opener. Specifics shorten responses—e.g., Scoutly's startup costs: National fee $85, uniform $50–$100, dues ~$100/year, gear $50–$150, total $300–$450. Turo could improve generic plans with ranges: Premium $25–60/day ($595/2 weeks), Standard $10–$30 ($280), Minimum $5–$15 ($140). Vague replies erode trust, pushing users to humans; specifics build reliability.
Audit responses ruthlessly: every word must serve the query. User testing identifies essentials vs. extras for progressive disclosure.