Been building AI prompt systems for SEO/affiliate content for a while now, and the #1 thing I see bloggers get wrong isn't keyword research — it's writing content that sounds like an AI wrote it and getting hammered on E-E-A-T (Experience, Expertise, Authoritativeness, Trust).

Here's the pattern I keep seeing:

1. No first-person experience signals. Google (and readers) can tell when nobody actually used the product. Even in AI-assisted drafts, you need a structured 'personal experience' block — specifics, not generic praise.

2. Outline-first, not prompt-first. If you go straight to 'write me an article about X,' you get generic structure. Building a real outline (H2/H3 hierarchy tied to actual search intent clusters) before generating prose cuts rewrite time in half.

3. Meta descriptions as an afterthought. CTR from SERPs is still one of the highest-leverage, lowest-effort levers on affiliate sites. Most niche site owners batch-write these last and it shows.

4. Zero internal linking strategy. Pillar pages without a deliberate internal linking pass lose 20-30% of the topical authority they could be building.

The fix isn't 'use better prompts' in the abstract — it's sequencing: keyword clustering → outline → E-E-A-T draft → internal linking pass → meta pass, each with its own dedicated prompt instead of one mega-prompt trying to do everything.

Happy to share the exact sequence I use if anyone's stuck on a specific step.