The Real Problem Isn't Quality
A Bynder study of 2,000 consumers found 56% preferred AI-written copy over human-written copy in a blind test. The moment they were told it was AI, 52% immediately disengaged. That's not a quality problem. That's a perception-of-thinking problem. Readers don't mind AI-polished prose. They mind feeling like the author didn't actually think.
This distinction matters for operators building B2B content at scale. The failure mode isn't using AI—it's using AI in a way that produces content that reads like no one had a real point of view before the prompt was written.
What a Non-Slop Workflow Actually Looks Like
Kieran Flanagan published a replicable system the same day Brian Chesky's 'claudeslop' backlash went viral, which is useful timing. His workflow does three things most AI content processes skip:
- Starts with a content audience profile, not an ICP. An ICP tells you who buys. A content audience profile tells you what that person is actively reading, arguing about, and sharing this week. Those are different inputs and they produce different outputs.
- Cross-references the author's proven personal patterns against live Reddit, X, and web signals. The AI isn't generating ideas in a vacuum—it's finding the intersection of what has already worked for this specific writer and what the audience is engaging with right now.
- Returns a ranked idea list, not a draft. The human still makes the judgment call on what's worth saying. AI handles the surface-area problem of idea generation. The author handles the selection problem, which is where voice and credibility actually live.
This is the structural difference between AI-assisted content and AI-generated content. The former is defensible. The latter is what's creating the trust gap the Bynder data is measuring.
The Actionable Takeaway
If you're running a B2B content function—or advising one—audit your current AI content process against one question: At what point does a human with actual expertise make a judgment call?
If the answer is