Can AI creators match human creators through posting volume alone?

The authors describe a "scale-over-preference" (SoP) dynamic in AI-generated content (AIGC), meaning videos made with a platform's own AIGC tools, compared with human-generated content (HGC). AIGC creators upload more, and their aggregate engagement is comparable to HGC creators, even though consumers prefer HGC. The evidence is a matched comparison on the Local Life Channel of a Chinese short-video platform, June 2024 to May 2025. Among 2,497 matched creators, AIGC creators uploaded more videos (HL median difference = 4, p < 0.001), and the authors say this increase is "driven primarily by AIGC video creation rather than HGC." Total valid views and full views showed "no meaningful differences" (valid views: HL median difference = 0, p = 0.003; full views: HL median difference = 0, p = 0.363). The valid-view comparison is significant at p = 0.003 on a zero median, so "comparable" is the authors' reading rather than a null result. On the consumer side, 47,288 matched user-video pairs show lower valid-view and full-view rates for AIGC videos than for HGC.

The supply-side mechanism is volume. AIGC tools let production "scale at substantially lower marginal effort than HGC," which the authors say amplifies "variation in content supply across creators." They call the result an "asymmetry between AIGC supply scale and consumer preference" and define a SoP index, SoPI = ln(S/P), where S and P are AIGC's relative supply and relative consumer preference against HGC. Over April and May 2025, daily values cluster at P ≈ 0.30–0.40 and S ≈ 0.55–0.70, with most SoPI values above ln(1.5). The authors' explanation is that creators "may maximize engagement returns by proliferating AIGC, thereby diluting the density of content that aligns with consumer preferences." AIGC is identified by the platform's metadata label; external tools are checked with Sightengine, and the unlabeled set is found to be predominantly HGC (SI-1.1).

This is a different route from the Nextdoor result in Does better summary writing actually increase user engagement?. There, one summary's informativeness removed the reason to open it. Here the gap appears at the level of platform supply, and the excerpt gives no reason consumers prefer HGC, so the Nextdoor mechanism is a candidate this excerpt neither tests nor rules out. The preference side is also read from implicit signals: the excerpt says views serve as both performance measures and distribution feedback. The implicit-feedback argument in Can implicit feedback reveal both preference and confidence? says such signals need splitting into preference and confidence, and this excerpt makes no such split. The feedback loop also resembles Do online ratings actually reflect independent customer opinions?, though no compounding is modeled here.

The excerpt does not establish several things. The creator-level comparison sums each creator's views across all their videos, so higher volume can match totals even when per-video engagement is lower, and no per-video result is given. The creator matching variables are not listed, and the consumer-side magnitudes are cut off mid-sentence. The data come from one platform, one country and twelve months, and the matched design is observational, so the excerpt does not show that AIGC causes the gap. The SoP pattern is documented in one platform's logs. Claims that it generalizes across platforms, or that it drives consumer disengagement, need evidence this excerpt does not supply.