How to Write SEO Content With AI Without Tanking Your Rankings

The "AI content tanks SEO" debate is mostly people fighting about the wrong question. The actual rule: Google rewards helpful, source-rich, original content. AI can produce that. It can also produce the opposite. The workflow matters more than the tool.

Here's the workflow that's worked for me — twelve posts ranked in the top 5 over the last six months, all AI-drafted, all human-edited.

Step 1: Research is human

This is the non-negotiable. Before AI touches anything:

• I pull the current top 10 results
• I read three of them top to bottom
• I note what they all say (table stakes), what only one says (their differentiator), and what NONE of them say (my opening)
• I find one real source — a primary research paper, a verified case study, an interview — that nobody else cites

If I can't do step 4, the post isn't worth writing. AI won't fix the absence of a fresh angle.

Step 2: Outline is human

Five minutes. Three columns: question, what makes my answer different, proof I'll cite. Most AI-content workflows skip this and produce drift.

Step 3: Draft is Claude

Now the AI does what it's good at. The prompt I use:

Draft a [word count] post for the keyword "[keyword]".
Audience: [specific reader description]
Angle (must use this, not generic): [angle from step 1]
Sources to cite (these MUST appear): [list]
Outline: [paste outline]

Voice rules:
- Active voice, contractions, short sentences
- No "delve", "leverage", "robust", "comprehensive"
- One specific number, name, or example per H2
- No "the future is bright" type closers

Two passes usually. First draft, then "make this 20% shorter and replace abstractions with specifics."

Step 4: Edit aggressively

This is where most AI-content workflows fail. The first draft is 80% there. The last 20% is what separates "AI slop" from "actually useful." Things to check:

• Cut every sentence that doesn't advance the argument. AI loves filler.
• Replace adjectives with numbers. "Significant improvement" → "23% lift."
• Make claims falsifiable. "Many marketers" → name the marketers or kill the claim.
• Restructure if needed. Sometimes the AI's section order is wrong. Move things.
• Read the opener and closer aloud. These are where AI tells live.

Step 5: Ship with proof

Cite sources inline. Link to primary research. Quote experts. Show your data. Google's quality raters are explicitly looking for evidence of expertise — AI content without it gets buried, AI content with it competes.

What kills rankings fastest

• Publishing the AI's first draft
• Using AI to "produce 50 posts on related keywords" — that's pure spam signal
• Letting the AI invent citations (it will, and Google's quality raters will catch you)
• Identical structure across every post (intro / problem / 5 bullets / conclusion every time)
• No author byline / no expertise signals

What this workflow scales to

About 8-12 posts/month per editor. AI doesn't actually save you the writing time — it shifts the time from "blank page" to "aggressive edit." The total is similar. What it saves is the friction of starting.