How I split copywriting work between AI acceleration and human judgment across a 6-stage workflow, from brief intake to final proofread.
Most conversations about AI copywriting land in one of two camps: “AI writes everything now” or “I’d never let AI near my brand.” Both are wrong. Here’s the workflow I actually use on every Sage Writers project — where AI accelerates the work and human judgment finishes it.
Why “no AI” and “all AI” both fail
The “no AI” freelancer is romantic but slow. They’re typing out research summaries by hand, reformatting bullet points into paragraphs, and losing two hours a week to work a machine can do in ten minutes.
The “all AI” agency is fast but dangerous. They’re shipping first-draft ChatGPT output as final copy, ranking for a fortnight, then confusing their client when the traffic doesn’t convert. AI doesn’t understand your positioning, your funnel, or your reader.
The amalgamation is where the good work happens. AI does the slog. Humans do the judgment. Below is exactly how I split it, stage by stage.
The 6-stage workflow, top to bottom
Stage 1: Brief intake (human only)
Every project starts with a conversation. I read your existing site, your competitors, your last three months of social, and any customer reviews I can find. Then I ask three or four awkward questions on a 15-minute call — the ones the brief didn’t answer because you didn’t know they mattered yet.
AI is not in the room for this. Positioning, tone, funnel stage, buyer psychology — these are things I need to feel, not summarise. If you want to see what a proper brief looks like, I broke that down in the template that saves you 3 calls.
Stage 2: Research + source gathering (AI accelerates)
This is where AI earns its keep. I’ll use it to:
- Summarise long PDFs (industry reports, competitor whitepapers, customer research)
- Cluster search-intent for a target keyword (what people actually mean when they type it)
- Pull out the recurring objections in a set of customer reviews
- Draft a rough outline structure for a long piece so I can react to it instead of staring at a blank page
Every AI output at this stage gets a human sanity-check. Hallucinated statistics get flagged and rewritten. Made-up sources get deleted. But the pace difference is real: what used to be a full day of research is now closer to two hours.
Stage 3: Outline drafting (AI drafts, I approve)
Once I’ve done the human research, I ask AI to structure a one-page outline based on the brief and the research summary. This is scaffolding, not writing.
Then I redraw it. I move sections around, delete the parts that miss the angle, add the sections AI never thought to include (usually the sceptical ones — AI defaults to enthusiasm). What lands with the client is my outline, informed by AI’s scaffolding.
Stage 4: First draft (human writes, AI checks)
The first draft is written by me, not by AI. That’s where the amalgamation gets deliberate: AI drafts read fluent but say nothing that hasn’t already been said. If you want your copy to be yours, a human has to write it.
What AI does at this stage is proofread as I go — catches typos, spots awkward phrasing, questions long sentences. It’s a second pair of eyes that never gets tired.
Stage 5: SEO pass (human strategy, AI suggests)
Every page runs through an SEO layer before it’s final. Keyword mapping, search intent, headings, meta, internal linking, entity coverage. I plan the SEO strategy myself — because SEO is a commercial decision, not a text-formatting task.
AI helps me spot the small stuff: alt text suggestions, meta description variants, related search terms I might have missed, questions to answer in an FAQ. I keep what fits the brief. The rest gets dropped.
If you want the longer take on how SEO has changed in the AI-search era, I wrote about that in SEO copywriting in 2026: what changed, what didn’t.
Stage 6: Edit + proofread (human only)
Final editing is human, always. Reading the copy out loud, cutting the middle 20% out of every paragraph, checking the rhythm, killing anything that sounds like a robot wrote it (even if a human did).
AI is genuinely bad at knowing when a sentence is too long. It’s worse at knowing when a paragraph is boring. It has no ear. Editing is where the writer earns their fee.
What AI never touches in my process
- Positioning decisions. Whether you should say “fast” or “expert” is a strategy question, not a copy question.
- Voice calibration. AI defaults to a bland, upbeat, mid-Atlantic voice. Getting a distinct brand voice out of it takes more effort than just writing in that voice yourself.
- Ethical judgment. Whether to make a claim, source a statistic, or push back on a brief.
- The final read. Every draft that leaves my inbox is read cover to cover, out loud, by me.
What this actually means for you
You get copy that ships in five working days instead of ten, because the research and scaffolding stages are compressed. You get final copy that sounds like a human wrote it, because a human did. And you get SEO baked into the draft from the outline stage, because the strategy is planned by someone who understands why you’re publishing the page.
The amalgamation is why I can run Sage Writers as a solo freelance copywriting agency without sacrificing the depth clients used to expect from a five-person team. AI takes the load off. Human craft finishes the work.
If you’d like to see this workflow applied to a page on your site — whether that’s a homepage rewrite or an SEO blog post — send a brief. I’ll come back with a written approach and a free sample rewrite of one of your pages.