Teardown: Spot the Brand Voice Drift Before It Publishes

Marketing Academy · Field Work●AI in Marketing

All projects·From: Automating Your Content Pipeline with AI

CoreTeardown· 40 minutes

Peloton Interactive

Objective: Given three AI-drafted social posts generated by a content repurposing pipeline, apply the lesson's brand-voice and factual-accuracy risks to correctly identify which posts are safe to schedule and which contain a real defect versus a merely different but acceptable stylistic choice.

You're the content lead reviewing Peloton's automated pipeline output before it reaches the Buffer approval queue. Three drafts came out of this week's instructor Q&A podcast repurposing run.

Review each draft against the source transcript excerpt, flag real defects using the lesson's failure modes (transcription errors compounding downstream, brand voice drift, unreviewed factual claims), and don't flag a stylistic choice that isn't actually wrong.

Before you start

What you'll need

Free path (everything below is enough to finish)

Google Sheets

FreeLog each draft against the source transcript with a pass/flag decision

Free, simple enough for a per-episode review log

The process

Specimens to review

Compare the drafted post to the source transcript. Is this safe to schedule as-is?

Sample output

SOURCE TRANSCRIPT EXCERPT: "...and we've now got about 40,000 people who've completed the Century Ride challenge since it launched..."

DRAFTED LINKEDIN POST: "Huge milestone: over 100,000 members have now completed our Century Ride challenge! The community keeps showing up. #PelotonProud"

Specimen: synthetic, realistic

Compare the drafted blurb's tone and language to the source transcript's actual voice. Is this safe to schedule as-is?

Sample output

SOURCE TRANSCRIPT EXCERPT: "...honestly some days you just don't want to get on the bike, and that's fine, showing up tired still counts..."

DRAFTED EMAIL BLURB: "Unlock unparalleled performance synergy with every ride. Our best-in-class instructors are committed to maximizing your fitness ROI, session after session."

Specimen: synthetic, realistic

Compare this drafted script to the source transcript. Does this one contain a real defect, or is it safe to schedule?

Sample output

SOURCE TRANSCRIPT EXCERPT: "...honestly some days you just don't want to get on the bike, and that's fine, showing up tired still counts..."

DRAFTED SHORT-FORM SCRIPT: "Some days you don't want to show up. Do it anyway. Showing up tired still counts. That's the whole workout today."

Specimen: synthetic, realistic

Final deliverable

A reviewed decision log marking each of the three drafts as approved, flagged for a factual fix, or flagged for a voice rewrite, with the specific reason cited.

See a reference example

Sample output

REVIEW LOG, Stitch Fix Styling Podcast Repurposing, Week 12

Draft 1 (LinkedIn): FLAGGED, factual. Source says 'about 15 stylists,' draft says '50+ stylists.' Do not schedule until corrected.
Draft 2 (email blurb): FLAGGED, voice drift. Source is warm and specific; draft reads as generic corporate copy, rewrite needed.
Draft 3 (video script): APPROVED. Accurate condensation of the source quote, tone matches.

Success criteria

You're done when you can:

Correctly flags the number-drift defect in draft 1 as critical and cites the source mismatch
Correctly flags the voice-drift defect in draft 2 as a tone/language problem, not a factual one
Correctly approves draft 3 without flagging a non-defect