AI Prompt for Troubleshooting Guides

Fix-it guides that work: this AI prompt builds troubleshooting docs — symptoms, causes, and fixes ordered by likelihood. Works with top AI tools.

The prompt

You are a support engineer who writes fix-it guides that actually fix things.

Context: I need a troubleshooting guide for [PRODUCT_OR_SYSTEM] used by [AUDIENCE] ([SKILL_LEVEL]). The most common issues: [KNOWN_ISSUES]. Tone: [TONE] — calm and clear, for frustrated users.

Task: Write the guide: a "before you start" box (what to check first: power, connections, updates, restarts), then issues organized by symptom — each with: symptom description (how the user experiences it), likely causes ordered by probability, fixes ordered easiest-first (each fix: steps + how to verify it worked + "if this didn't work, try..."), and when to escalate (contact support / call a pro). Add a quick-diagnosis table at the top mapping symptoms to sections.

Format: Markdown with diagnosis table first, then headed issue sections, escalation criteria bolded.

Constraints: Easiest fixes first — never start with "reinstall everything". Each fix must include verification ("you'll know it worked when..."). Honest escalation: say clearly when it's beyond DIY. No blame language ("user error") — frame as system behavior. Mark version-specific steps [VERIFY VERSION]. Write for frustrated readers: short sentences, no jargon without definition.

Variables to fill in

Replace each [variable] in the prompt with your own details:

[PRODUCT_OR_SYSTEM] [AUDIENCE] [SKILL_LEVEL] [KNOWN_ISSUES] [TONE]

Example input

Example: PRODUCT_OR_SYSTEM = 'home Wi-Fi router', AUDIENCE = 'homeowners', SKILL_LEVEL = 'non-technical', KNOWN_ISSUES = 'slow speeds, drops in one room, device won't connect', TONE = 'patient and plain-spoken'.

How to use this prompt

List real issues from support tickets or forums, generate, then test the top 3 fixes yourself. Order by actual frequency, not your guesses.

Tips for the best results

The diagnosis table is the highest-value element — frustrated users scan, not read.
'Restart it' belongs first, unapologetically — it fixes 50% of issues.
Escalation criteria protect users from making things worse.