Your Brand Style Guide Grows in the Review Comments
Some review comments belong in your brand style guide. Others mean the AI draft had nothing to work with. Here’s how to sort them and fix each at the source.
It’s Thursday, and you’re reviewing this week’s AI-assisted draft. It follows every rule your team has written down, but you find yourself highlighting the third paragraph, just like last week: “What’s our angle here?” You almost assume next week’s won’t have an angle at this point.
Nobody gave the model anything better to work from, so it drafted from its own defaults, and that’s how the AI Ouroboros got started.
Style guides are great, but at the speed content moves, the rules your team actually follows aren’t even the ones in the brand style guide anymore.
Your team is operating under an evolving style, from last month’s brief doc to this week’s blog revisions. You add a few more every week. That’s the working style guide. It grows in the review comments, one round at a time.
You’re not the only one drafting with AI ahead of clear guidance. In an Adweek and NewtonX survey of 500 marketers, 43% said their company’s AI guidelines were unclear or only slightly clear. Most teams are writing the rules as the drafts come in.
Sorting Review Comments on AI Drafts
Review comments on AI drafts come in two kinds.
Style comments are about how the draft reads, and they belong in your brand style guide. Substance comments are about what the draft says, and those go back to the brief.
How it sounds looks like this.
- Too many em dashes.
- The same sentence shape, paragraph after paragraph.
- Subheads written to sound clever, which leave a skimming reader guessing what each section covers.
Substance comments look like this:
- “Source?”
- “What’s our angle?”
- “Every competitor has said this.”
- A stat with a question mark next to it, because nobody can find where it came from.
Picture last month’s comments pasted into one doc. You sort them into a style pile and a substance pile. By the end of the first draft’s comments, the substance pile is already taller, and there are still three drafts to go. Every note in it marks a spot where a reviewer stopped to chase down a claim.
Writing substance comments up as rules makes the guide longer and leaves the next draft exactly as thin as the last one. Sort before you write anything down, and your reviewers stop spending every round on comments no rule was going to fix.
What Belongs in the Brand Style Guide
Style comments make good rules, because you can check a draft against them line by line. When the same note keeps coming back, write it down once and stop typing it.
Give the countable habits a ceiling, so the device stays available for the sentence that needs it. Roughly one em dash per 300 words is a workable limit. Sentence shapes work the same way: if every paragraph opens on a short claim and closes on a punchline, name the pattern and cap it. When the notes keep stacking up, build a feedback system around those comments so each rule gets written on purpose.
Expect the habits to come back after the rule is written. Inflating why a subject matters, or tacking a vague “-ing” clause onto the end of a sentence, are habits deeply embedded in how AI models are trained and deployed. A rule makes them rarer, so keep checking for them every round.
Keep the rules where the drafting tool reads them first, in the instructions it gets before every draft. Then use them again at the end. A style guide does its best work once the substance is settled, which is the moment to hand the style guide to AI for the late revisions.
Why Style Rules Can’t Fix Substance Comments
A model drafts from what it’s given. Ask for a post with no source, no angle and no reader in mind, and you get the average of everything it has already read on the subject. It’ll be fluent and well organized, and it’ll be close to the post your competitors got when they asked the same question.
Now add “Always cite a source” to the style guide. The next draft still has nothing to cite, so it skips the citation or reaches for something that sounds right. “Be original” fails the same way, because the prompt gave the model nothing to be original about.
This is where AI drafts start sounding like everyone else’s. The model read the same coverage your competitors’ models read, and it hands back the same recycled takes. No amount of line editing adds a source or a position the draft never had.
The fix sits before the first prompt, because the outline has to come from outside the model. Put the substance in before drafting, and fewer review rounds go to asking for it after the fact. Your reviewers’ time goes back to judging whether the argument holds up.
Fixing Substance Comments Before the Draft
Give the drafting tool a brief before it writes. Keep it short, as long as it covers what the model can’t supply for itself: a current source it can link to, the angle you’re taking on that source, and the reader the piece is for.
Then check the brief before anyone drafts. A missing angle is quicker to spot in an outline than in 1,300 words of finished copy, and cheaper to fix. Whoever does the drafting, keep a human review step on the brief and on the draft.
When a “What’s our angle?” comment still comes back, skip the reply in the margin. Trace it to the brief, fix the angle there, and draft again from the corrected version. If the angle was never written down, write it in one sentence. If it was in the brief and the draft ignored it, it was probably buried under the background, so move it to the top.
Before the next draft starts, check the brief for:
- A source: current, real and linked, so every claim can be checked against it.
- An angle: the position you’re taking on that source, in one sentence.
- A reader: one role or person, and what they should do after reading.
If one of the three is missing, the draft will come back with the comment that would have caught it.
Giving the Drafting Tool Better Material
Chatter takes on the slowest part of that checklist, which is finding a source worth responding to. It surfaces current stories filtered to your brand, keywords and personas, each one linked to the original. When a story is worth a response, Chatter turns it into a Topic: a brief built from that story, tagged with the audience and keywords it’s for, with the sources linked. Add your angle in the Topic if it isn’t there yet, then export it and hand it to your drafting tool. The draft starts from a source and an angle instead of the model’s defaults.
Chatter doesn’t enforce style rules or review drafts, and it isn’t in the business of writing your content for you. The brand style guide and the review stay with your team.
Sort Your Last Three Drafts
Start with the comments from your last three AI drafts. Sort them into style and substance, and see which pile is bigger.
Write each style comment into your brand style guide once. Send the substance comments back to the brief. Then sort again next month. When the substance pile is the shorter one, your briefs are doing their job.