$ man how-to/how-to-use-ai-content-creation

AI content quality is a function of input quality. Generic input ("write a blog post about AI tools") produces generic output. Specific input produces specific output. The method: give the AI three things before asking it to write. First, your experience - the specific story, result, or insight from your actual work. Second, your voice - the patterns, vocabulary, and style you want. Third, the format - the platform, length, structure, and audience. With all three inputs, AI produces content that reads as yours. Without them, it produces content that reads as everyone else's.

PATTERN

The most common AI content failure: asking it to write about something you have not done. AI can research and synthesize, but it cannot fabricate authentic experience. The fix is to provide your experience as input. Before generating content, write a rough brain dump. Not polished writing - just notes. "This week I tried waterfall enrichment in Clay. Used Apollo, Prospeo, and Dropcontact in sequence. Got 82% email coverage vs 55% from Apollo alone. Took 2 hours to set up. Credits cost about $1.50 per lead. The gotcha was MX validation - without it, 8% of emails were bad domains." That brain dump is 60 words of raw experience. The AI turns it into a polished 500-word blog section with your specific numbers, your specific tools, and your specific gotcha. It cannot make up the 82% figure or the MX validation insight. Those come from you. Rule: never publish AI content about something you have not personally done. Use AI as a writing accelerator, not a writing replacement.

PATTERN

Before the AI writes, load your voice system into the conversation context. In Claude Code, this means the CLAUDE.md references your voice DNA file, or you explicitly load it at the start of a content session. Minimum viable voice loading: "Write in short sentences. Average 10 words. No hedging language. No phrases like 'it is worth noting' or 'in today's landscape.' Use specific numbers over vague qualifiers. Start with claims, then prove them. Lowercase energy - no exclamation marks, no hype." That is 50 words of voice instruction that dramatically changes the output quality. The AI stops writing in its default polished-but-generic voice and starts writing in your patterns. For better results: load your full voice DNA file with 30+ specific rules. The more patterns you provide, the closer the output matches your natural writing. The sweet spot is 50-100 rules that cover vocabulary, sentence structure, formatting preferences, and content patterns.

PATTERN

The AI gives you a draft. Now make it yours. The editing pass is where AI content becomes authentic content. First pass - truth check: is every claim accurate? Did the AI add details you did not provide? AI sometimes confabulates - adds plausible but false specifics. Remove anything you cannot verify. Second pass - voice check: read it aloud. Does it sound like you? Mark every sentence that sounds like "AI wrote this" and rewrite it. Usually it is the transitions and the opening/closing sentences that need the most human touch. Third pass - add the texture: inject the details only you know. The unexpected difficulty. The tool that almost worked but had a deal-breaking bug. The metric that surprised you. These textures are what make content feel human. Fourth pass - cut the fat: AI writes more than necessary. It explains things readers already know. It adds caveats that weaken the argument. Cut 20-30% of the draft. Tighter content performs better on every platform. The editing takes 15-30 minutes per piece. The writing took 3 minutes. Total time: 20-35 minutes for a piece that would take 90-120 minutes to write from scratch.

PRO TIP

Before publishing, run the authenticity test. Three questions: 1. Could anyone else have written this? If yes, it needs more personal experience. Add your specific results, your specific tools, your specific failures. Make it impossible for someone who has not done your work to produce this content. 2. Would you say this out loud to a colleague? If the language feels stiff or formal, it still has AI residue. Rewrite the stiff sentences in the way you would actually explain it in conversation. 3. Does it make a claim you would defend? If there is a statement you would hedge on in a live conversation, either strengthen it with evidence or remove it. AI loves to make broad claims. Your content should only make claims you have earned through experience. If the piece passes all three, publish it. If not, one more editing pass focusing on the questions it failed. The goal is not to hide that you used AI. The goal is to produce content that reflects your actual expertise, delivered faster than you could write it manually. When the experience is real and the voice is yours, the tool used to produce it is irrelevant.