LLMS

How to Batch a Week of Writing Without Sounding Like a Bot

Kevin PlumleeOctober 10, 20263 min read
How to Batch a Week of Writing Without Sounding Like a Bot

If you use AI to write and everything comes out in the same cadence, readers can smell it. The sentences are all the same length. The takes all lean on tired contrasts. The endings all put a hand on your shoulder and tell you to be intentional.

Batch a week’s writing anyway, because shipping beats waiting for a muse that does not respect calendars. The trick is to batch the system, not the voice. Voice still has to show up like a person who has a body and a job.

Don’t batch from a blank prompt

The samey problem starts when you ask the model to “write a blog post about X” fourteen times. Of course it rhymes with itself. You gave it the same job in the same empty room.

Batch from notes that already sound human:

  • a title that’s specific
  • three points you actually believe
  • one concrete example
  • a line you would say out loud to a friend
  • what you refuse to claim

Sometimes those notes are voice memos dumped into text. Sometimes they’re ugly bullets. The point is there is already a point of view before the model types.

Separate research day from prose day

Do the boring collection first: URLs, stats you can verify, competitor pages, call themes, FAQ gaps. Put them in a brief.

Then write. Mixing research and drafting in one chat is how you get confident mush with a fake statistic living in paragraph three.

If a claim needs a source and you don’t have one, cut the claim. Do not ask the model to “find something that sounds right.”

Edit in passes, not vibes

After the draft lands, run the same passes every time:

  1. Cut the opening throat-clearing.
  2. Replace vague words with concrete ones.
  3. Kill banned patterns and corporate filler.
  4. Check facts and links.
  5. Read the ending out loud. If it sounds like a poster, rewrite it into a decision.

Batching works when the edit pass is ruthless. Without that, you just industrialize mediocrity.

Keep a swipe file of your own lines

Save sentences that sound sharp when you’re clear. Not to copy-paste forever. To remind the model (and yourself) what good sounds like for this site.

When a draft drifts into generic AI cadence, paste two of those lines into the brief and say: match this energy, not LinkedIn energy.

Cadence variety is a feature

Short. Then a longer sentence that earns its length because it carries a real condition or tradeoff. Then a list when a list helps. Then stop.

If every paragraph is three sentences of equal weight, readers glaze over. Vary on purpose.

What not to batch

Do not batch legal promises, clinical claims, or anything you would not defend in a room with a skeptic. Those need slower hands.

Also do not batch twenty thin location pages that swap a city name. That is not a writing system. That is a future cleanup ticket.

A practical weekly rhythm

Monday: briefs and sources. Tuesday to Wednesday: drafts from the briefs. Thursday: edit passes and packaging. Friday: schedule what survived, kill what didn’t.

Volume without a kill rate is how calendars fill with regret.

Bottom line

Batch the workflow. Do not batch the personality. Give the model notes with a point of view, edit like you mean it, and cut anything you cannot defend. Shipping on a schedule is fine. Sounding like a template factory is not.

Kevin Plumlee

Founder, RVNOPS

Kevin Plumlee

15+ years running healthcare SEO, paid media, and operations. Now I advise the teams who do that work, across marketing, intake, technology, and AI.

More about me

Tell me what’s not working.

Stalled growth, spend you can’t trace, systems that fight your team. That’s a good place to start.

Start a conversation