The Sameness Problem
By mid-2025, AI content had homogenized the internet. The same phrasings, the same structures, the same fake authority. Brands that adopted AI without a strategy ended up sounding like every competitor that also adopted AI. Generic content gets shared less, ranks less, and earns less trust.
The Framework
- Voice first — codify the brand voice in writing before any AI touches a draft.
- Ideas stay human — AI doesn't decide what to write about.
- Production with AI — structured drafting against voice rules.
- Human review — every published piece gets human editor pass.
- Measure differently — not just output volume, but reader engagement.
Voice First
The brand voice document is the single highest-leverage asset in AI content strategy. It includes:
- Brand personality (3–5 adjectives, with anti-adjectives).
- Tone (formal/casual, warm/distant, direct/diplomatic).
- Sentence patterns (length, rhythm).
- Banned phrases.
- Required structures.
- Examples (good vs. bad).
- • Print three of your AI-drafted paragraphs alongside competitors' AI content.
- • Can a reader tell which is yours?
- • If no, your voice rules aren't tight enough.
Ideas Stay Human
AI is excellent at executing on an editorial direction. AI is mediocre at deciding what direction to take. Topic selection, angle, point of view — human decisions. AI helps research them; AI doesn't pick them.
Production With AI
The production loop: brief in → AI draft against voice rules → editor refines → publish. The brief is dense (audience, goal, angle, key points). The AI does the heavy lifting on first draft. The editor brings the brand back into the prose.
The Review Layer
No AI content goes live without human editor review. The editor's job: enforce voice, add specificity, cut generic phrases, ensure facts. Editor time per piece: 30–45 minutes for a 1500-word post.
Metrics That Matter
- Time on page — flat or up vs pre-AI content.
- Scroll depth — flat or up.
- Newsletter subscribes per published piece.
- Backlinks earned — the real signal of quality.
What not to track: word count, posts per week. Both incentivize the wrong behavior.
The brands winning at AI content in 2026 ship less than the early enthusiasts did in 2023 — and earn more attention with what they ship. Quality compounds; quantity decays.
See making AI blog posts useful.
FAQ
How many AI-assisted pieces per week? Start at 2–3. Quality matters more than cadence.
Should we disclose AI use? Reasonable people disagree. We disclose; many don't. Trust matters either way.
Specialist or generalist editors? Specialists for technical topics. Generalists fine for everything else.