We Ran 258 Comments Through AI. Here's What Steven Bartlett's Audience Is Really Thinking.

258 people told Steven Bartlett exactly what they think about AI. Not in a survey. Not in a focus group. In the comments section of "AI Whistleblower: We Are Being Gaslit By The AI Companies!" featuring Karen Hao. We used AI to listen.

Here is what his audience is really thinking.

The Numbers

Three Tribes, One Comment Section

The 258 comments are not a monolith. They split into three distinct audience clusters, each with different emotions, different vocabulary, and different needs that Steven Bartlett is not currently serving.

Cluster 1: Fearful Workers (45%, 116 commenters)

Sentiment: 72% negative

These are working professionals who watched the episode and felt their stomach drop. They are teachers, marketers, designers, customer service reps. They talk about their jobs in the present tense and AI in the future tense, as if waiting for a verdict.

“"I already see AI replacing half my team's work. My manager pretends it is not happening."”

“"What are we supposed to do? Just accept that our careers are over?"”

This cluster generates the most personal anecdotes. They mention their children, their mortgages, their colleagues. They demand government action. They tag friends to watch. They are not angry at AI itself. They are angry at the silence from the people who could help them prepare.

What Steven Bartlett is not doing: Zero content in his entire catalogue serves this audience practically. No "here is how to actually prepare" episode. No action framework. 116 people told him exactly what they need, and the response was a pinned subscribe request.

Cluster 2: Analytical Skeptics (30%, 78 commenters)

Sentiment: 55% neutral

These are the contrarians. They push back on the hype AND the fear. They have technical knowledge. They cite papers, reference specific model capabilities, and correct other commenters. Their comments are the longest in the dataset. They get the fewest likes but produce the highest quality discourse.

“"Every generation has had a technology panic. The printing press. The loom. The calculator. We adapted every time."”

“"Karen Hao is excellent but this framing makes it sound like we have zero agency. We do."”

This cluster self-moderates. They reply to fearful comments with nuance. They are the comment section's immune system.

The paradox: These comments get 50% fewer likes than the fearful ones, but they are the ones that actually move the conversation forward.

Cluster 3: Anti-Corporate Activists (25%, 64 commenters)

Sentiment: 85% negative

For this group, AI is not a technology story. It is a power story. They frame every point through inequality: who benefits, who pays, who decides. They reference Elon Musk, Sam Altman, and Mark Zuckerberg by name. They use words like "oligarch," "exploitation," and "dystopia."

“"Of course they are gaslighting us. That is what corporations do. Tobacco. Oil. Now AI."”

“"Sam Altman sits in front of Congress and lies through his teeth while his net worth doubles."”

This cluster has the highest share rate. Their comments are built for virality: short, emotional, quotable. They are not trying to learn. They are trying to recruit.

The Sentiment Map

What is Driving the Narratives

The Top 10 Influence Leaders

Not the most frequent commenters. The most influential, measured by likes.

The Paradox

The most liked comment on a video about AI fear is not fearful. @KiteRheinRuhr (410 likes) posted a Jurassic Park reference: "Your scientists were so preoccupied with whether they could, they didn't stop to think if they should."

It is intellectual, not emotional. Cultural, not political. The audience rewarded intelligence expressed through shared cultural literacy over raw outrage, even though negative comments overall get 2x the likes of neutral ones.

This tells us something important: the DOAC audience is literate, referential, and rewards cleverness. The loudest comments are not the most valued.

The Spam Ring

Buried in the 258 comments, 12 accounts were running a coordinated crypto promotion.

The pattern:

• All accounts created within the same recent window
• All posted within a 2-hour span
• All follow the same template: "[Fortune 500 company name] + Caddun protocol + financial claim"
• None engage with the video content

Plus @Mickey.steph4, whose book promotion comment received 380 bot-boosted likes, making it the second most liked comment on a video about AI ethics. The irony is not subtle.

Time to moderate: Under 5 minutes, if you know what to look for. 12 accounts, one template, same posting window. An automated system catches this in seconds.

What Steven Bartlett Should Do Next

Six data-backed recommendations from the intelligence report:

1. Follow-up episode: "I Showed This Report to Sam Altman"

42% of comments reference Big Tech deception. The audience is begging for confrontation. Bring the data. Show Sam Altman what 258 people said about him.

2. The AI Survival Guide for Normal People

45% of his audience (116 commenters) expressed job displacement anxiety. Zero content in the DOAC catalogue serves them practically. This is the most underserved segment in his audience. An episode titled "What to Actually Do If AI Takes Your Job" would be the highest-engagement video of the quarter.

3. Pin @KiteRheinRuhr, Not the Subscribe Request

The pinned comment (116 likes, subscribe request) is outperformed by a community comment (410 likes, cultural reference) by 3.5x. Pinning the top community comment signals that DOAC values its audience. A subscribe request signals that it values metrics.

4. Clean the Caddun Spam Ring

12 coordinated bot accounts. 5 minutes to remove. Every hour they stay up, they erode trust in the comment section for real participants.

5. Double Down on the 14-Language Dub Strategy

8+ languages detected in the comments. The audience is already global. The dub strategy is validated by the data, not just the hypothesis.

6. Engage Karen Hao's Audience

Karen Hao has a book-reading audience (her MIT Technology Review following). The DOAC audience praised her specifically. A crossover event, co-authored piece, or book club episode converts her followers into subscribers.

The Missing Creator

Perhaps the most striking finding: zero creator replies in 258 comments. The only engagement from @TheDiaryOfACEO is a pinned subscribe request with 116 likes, outperformed by 7 community comments.

87 reply threads happened in the audience. 87 conversations that Steven Bartlett was not part of. Some of those threads are better than most podcasts.

The comment section is not a wall. It is a room. And right now, nobody from DOAC is in it.

FAQ

Q: How many comments were analyzed?

A: 258 comments on the Diary Of A CEO episode "AI Whistleblower: We Are Being Gaslit By The AI Companies!" featuring Karen Hao as the guest.

Q: How were the audience clusters identified?

A: Comments were grouped by semantic similarity using embedding-based clustering, then validated by sentiment analysis and narrative theme extraction. Three distinct clusters emerged: Fearful Workers (45%), Analytical Skeptics (30%), and Anti-Corporate Activists (25%).

Q: What tool was used for this analysis?

A: This report was generated by NAWA's Intelligence Engine. The full analysis of 258 comments was completed in under 60 seconds. Learn more at trynawa.com.

Q: Is the spam detection automated?

A: Yes. The 12 Caddun crypto bot accounts were flagged automatically based on account age, posting time correlation, and template matching. The @Mickey.steph4 book promotion was flagged by like-count anomaly detection (380 likes on a promotional comment with no engagement replies).

Q: What about comments in other languages?

A: 8+ languages were detected in the dataset. NAWA's Arabic Intelligence engine processes Gulf, Egyptian, Levantine, and Maghreb Arabic natively using IBM ALLaM, the most advanced Arabic language model available.

This report was generated by NAWA's Intelligence Engine. 258 comments. 60 seconds. trynawa.com