4 / 16
Chapter 4 · Step 2

Step 2: Writing and production

The step where most agencies confuse "AI content" with "lazy content" and their blogs quietly become a liability.

The instinct is to paste the keyword into ChatGPT, accept whatever comes out, and call it done. Every client I have talked to who tried this abandoned the experiment within 60 days. The articles were bland, full of factual errors, stripped of any point of view, and read like every other AI blog on the internet. Google noticed.

The agencies that actually win with AI-assisted content treat the AI as a production accelerator, not an author. The output still has to sound like a human wrote it. It still has to be factually correct. It still has to argue a point. The AI shortens the time from zero to draft by 80%. It does not eliminate the editorial layer.

The production framework

Every article moves through five stages, in order, without skipping any of them:

StageWhat happensWho owns it
1. BriefAngle, target keyword, outline, voice rules, must-include factsStrategist (you or a senior editor)
2. DraftAI generates first pass against the briefAI (ChatGPT, Claude, Gemini)
3. Structural editRework the flow, cut fluff, tighten the argumentHuman editor
4. Fact and voice passVerify claims, match client tone, add data/quotesHuman editor
5. Final polishHeadlines, CTAs, internal links, metaHuman editor

Skip any of these stages and the article either fails to rank (no editorial depth) or ranks briefly then gets outranked by a better-written competitor (no retention).

The brief is 60% of the work

The single highest-leverage artifact in the entire production pipeline is the brief. A good brief makes the AI draft 70% usable. A bad brief means your editor is essentially rewriting from scratch, which defeats the entire point of using AI.

A real brief contains:

Writing a brief this complete takes 15-25 minutes per article if you know what you are doing. For a 25-article retainer, that is 6-10 hours per client per month before a single word gets written. For five clients, 30-50 hours of briefing alone.

The DIY tool stack for production

ToolRoleCost
ChatGPT Plus, Claude Pro, or bothDraft generation$20-40/month
Frase or SurferSEOSERP analysis, content optimization$45-89/month
Grammarly BusinessCopyediting pass$15/month
Originality.ai or GPTZeroAI detection check (for clients who care)$15-30/month
2-3 VA editorsStages 3, 4, and 5 of the production framework$1,200-1,500/VA/month
Google Docs or NotionWorkflow managementFree-$20/month

The DIY workflow, article by article

  1. Write the brief (15-25 minutes)
  2. Generate the draft in ChatGPT or Claude using your brief as context (5-10 minutes)
  3. Run it through Frase or Surfer to score against the top-ranking articles for your keyword (5 minutes)
  4. VA handles structural edit: rework flow, cut AI tells, sharpen transitions (20-30 minutes)
  5. VA fact-checks and voice-matches: verify stats, add client-specific angles, match tone (20-40 minutes)
  6. Senior review: you or a lead editor reads, flags issues, requests fixes (10-15 minutes per article)
  7. Revisions: VA cleans up the flagged issues (10-20 minutes)

End to end, an article takes 85-145 minutes of human time even with the AI doing the heavy lifting on the draft. At 25 articles a month per client, that is 35-60 hours of production work per client.

Where it breaks at scale

The writing stage is where VA management becomes its own full-time job. Two different VAs on the same client produce two different tones, two different depth levels, two different handling of the client's product mentions. Clients notice. They always notice.

You also run into the briefing bottleneck. The only person who can write a good brief is you (or a senior editor who costs $6-10K per month). At one or two clients, briefing is manageable. At five, you spend your entire Monday and Tuesday writing briefs instead of running your business.

The specific trap: VAs get sharper at spotting AI tells after a few months, but they also start taking shortcuts. They stop re-reading the brief carefully. They reuse phrasing from previous articles. They over-rely on the AI-generated outline instead of restructuring when the SERP calls for it. The content looks "fine" on a skim but the keyword rankings stall. You lose clients in month 5-7 because the results curve flattens at exactly the moment clients expect it to climb.

What scales this step

The leverage is in systematizing the brief. If the brief is generated from the keyword map plus SERP analysis using a consistent template, and the draft is generated against that brief using consistent prompts, and the edit passes are guided by a consistent quality rubric (see Chapter 5), the output stays consistent across every client and every article. That is what RankStack does. The production time per article drops from 85-145 minutes to under 5 minutes of your time, with the same quality floor.

Quality review is the step that actually enforces the consistency. That is what we cover next.