The obvious answers were the first ones I checked. Open any one to see the test and the result.
Topic
Do women just pick lower-traffic topics?
No, within the same categories, women still get about 20% fewer views.
Comparing only within the same 40 content categories removes just ~11% of the gap. Nearly 9 in 10 of it is within-category.
Output
Do women post fewer videos?
Partly, but a 22% gap remains after holding it constant.
Video count is the single largest ordinary factor in the raw gap. Hold it and the other basics constant and about 22% fewer views still remains.
Titles & descriptions
Is it how women title their own videos?
No, their own word choices move the gap by under 3%.
Scoring every title and description and holding them constant removes only a few percent of the gap.
Transcripts
Is it what women actually say in their videos?
No, spoken content runs in women's favor: holding it constant leaves the gap intact, even slightly larger.
Transcript language, scored with the same tools as everything else, explains essentially none of the gap. Transcript availability is equal across genders (about 88% for both), so it is not a data artifact.
Thumbnails
Is it how women package their videos?
No, women's thumbnails are if anything an asset: more positive, warmer, less clickbait.
Packaging is the largest single creator-controlled surface and still explains under about 9% of the remaining gap; in the integrated model it partly runs in women's favor.
Subscribers
Do audiences subscribe to women less?
No, there is no real subscriber gap. The shortfall is only in views.
Run the same models with subscribers as the outcome and the gap essentially disappears.
Engagement
Do audiences value women's content less?
No, women's channels get higher engagement, not lower.
More likes and comments relative to audience size, on women's channels, not fewer.
Who comments
Is it just male commenters?
No, male and female commenters both soften on women's channels.
Measured separately, both groups address women's channels with warmer, lower-heat language. The pattern does not depend on who is commenting.
Self-promotion
Do women promote themselves less elsewhere?
No, women link out more, and it does not change the gap.
Women link to other platforms more often (about 20% vs 16%); accounting for it leaves the view gap intact.
Backlash
Is it backlash against women in male-typed fields?
No, the backlash test comes back null.
The three-way interaction is not statistically significant (beta=0.026, p=.73), so the visibility penalty attaches to the affective signal itself, not specifically to women as a category.
It also holds up under the usual robustness checks: outliers, missing data, comment length, alternative codings, and alternative ways of computing the statistics.
