Diagnosing this requires segmenting audience behavior by traffic source in YouTube Analytics, specifically comparing how viewers who arrive at long-form videos via Shorts-driven traffic (Shorts feed, “from Shorts” attribution) behave against viewers arriving through other sources (suggested videos, search, browse features), and looking for a systematic gap in retention and session behavior between those groups. There’s no single confirmed metric YouTube publishes that names this phenomenon; it’s a pattern you infer from comparing segments, not something the dashboard labels for you.
Why this is even plausible: two partially distinct recommendation systems
YouTube has been explicit in creator-facing communications that Shorts are surfaced through a dedicated short-form feed with its own engagement logic, distinct in important ways from the recommendation logic that drives long-form suggested videos and browse features. The Shorts feed is built around rapid, low-friction, high-volume consumption: swipe-based browsing, completion and rewatch on very short videos, and session patterns that look more like a vertical-video feed than a video-on-demand session. Long-form recommendation, by contrast, has historically weighted signals like session watch time, click-through rate on thumbnails/titles, and sustained retention across a longer viewing session.
Because these are meaningfully different consumption contexts, a viewer who discovers a channel through Shorts is being trained (in the sense that YouTube’s systems build a viewing profile from behavior) on a very different signal set than a viewer who discovers the channel through a long-form suggested video. If a channel’s Shorts succeed at pulling in large volumes of viewers who are optimized for fast, low-commitment scrolling, and the channel’s recommendation profile increasingly reflects that audience, it’s a reasonable hypothesis that YouTube’s systems could start surfacing the channel’s long-form content to viewers whose demonstrated behavior pattern is mismatched with what long-form retention requires, i.e., people who came for a 30-second hook and don’t have the session habits or intent to sit through a 12-minute video.
It’s important to be precise about what is and isn’t confirmed here. YouTube has not disclosed a named “Shorts cannibalization” mechanism, and there’s no public statement that says audience overlap between formats mechanically degrades long-form recommendation eligibility. What is grounded in YouTube’s own documentation is (1) that Shorts and long-form use different discovery surfaces and algorithms, and (2) that YouTube Analytics provides traffic-source and audience segmentation reporting specifically designed to let creators see where viewers of a given video came from. The cannibalization hypothesis is a pattern to test with that reporting, not a confirmed YouTube effect to assume is happening.
Diagnostic approach using YouTube Analytics
Start in YouTube Studio Analytics at the channel level, then drill into individual long-form videos, using these actual reporting categories:
Traffic source: Shorts feed. In the Reach tab, traffic sources include a “Shorts feed” category distinct from “Suggested videos,” “Browse features,” and “YouTube search.” For long-form videos published after the channel became Shorts-active, check what percentage of impressions/views are attributed to Shorts feed traffic (this typically shows up when a Short links to or drives discovery of a long-form video, or when the channel’s overall audience graph has shifted). A rising share of Shorts-originated traffic reaching long-form content is the first flag worth investigating.
Audience retention, segmented by traffic source. The Engagement tab’s audience retention graph can be cross-referenced against traffic source data (export the data if the in-dashboard segmentation isn’t granular enough) to compare average view duration and percentage retention for viewers arriving via Shorts-adjacent discovery versus viewers arriving via long-form-native sources like suggested videos or search. If Shorts-sourced viewers show a steep early drop-off (leaving in the first 15-30 seconds at a meaningfully higher rate) compared to viewers from other sources, that’s a direct behavioral signal of the mismatch the question describes.
Returning viewers vs new viewers, and subscriber status. The Audience tab breaks down views by subscribers vs non-subscribers and returning vs new viewers. If Shorts have driven substantial subscriber growth, check whether those newer subscribers (who likely subscribed off Shorts) show different long-form watch-time-per-session and video-completion behavior than subscribers acquired before the channel invested heavily in Shorts. A widening gap here, tracked over time as the Shorts-acquired subscriber cohort grows as a share of the audience, is the clearest longitudinal signal.
Impressions click-through rate on long-form videos over time. If long-form CTR from suggested/browse surfaces is declining in a period that correlates with Shorts growth, it can indicate YouTube is surfacing long-form videos to a broader but less-matched pool of viewers who are less likely to click, consistent with the audience-profile-shift hypothesis, though correlation here should be treated cautiously since CTR can move for many reasons (thumbnail/title fatigue, topic saturation, seasonality).
A hypothetical illustration
As a hypothetical illustration: imagine a cooking channel called Hearthside Kitchen that ramps up Shorts publishing from zero to four per week over three months, alongside its usual long-form recipe videos. Hypothetically, its subscriber count grows quickly during that period, largely driven by viewers discovering the channel through the Shorts feed. Six months later, the channel’s owner notices that average view duration on new long-form videos has been declining, even though the videos themselves haven’t changed in format or quality.
Segmenting the data, suppose Hearthside finds that long-form videos are now receiving a meaningfully higher share of impressions attributed to “Shorts feed” as a traffic source than they were a year earlier, and that viewers arriving via that path show a steep drop-off in the first 20 seconds of the long-form videos, well below the retention shown by viewers arriving through search or suggested videos. That combination, rising Shorts-attributed traffic to long-form content paired with a segment-specific retention gap, would be the diagnostic pattern described above: evidence that Shorts-trained viewing habits are being funneled into long-form content that those viewers aren’t retained by, rather than evidence that the long-form content itself has gotten worse.
What to do if the pattern shows up
If segmentation shows Shorts-originated viewers reliably underperform on retention for long-form content, the practical response isn’t to stop making Shorts (they may still be valuable for channel growth, subscriber acquisition, and brand reach in their own right), it’s to decouple the strategies where they conflict:
Avoid using Shorts primarily as direct funnels into long-form content if the retention data shows that funnel converts poorly; instead let Shorts serve top-of-funnel awareness and channel discovery while long-form content earns its own long-form-native audience through search and suggested placement.
Watch whether the channel’s overall long-form suggested/browse impressions volume changes shape (reaching a wider but shallower audience) after a sustained period of high Shorts output, and if so, treat that as a signal to evaluate whether Shorts publishing cadence or content strategy needs adjustment relative to long-form goals.
Treat this as an ongoing diagnostic rather than a one-time check, since audience composition shifts gradually and the traffic-source and retention segmentation described above needs to be revisited periodically (e.g., quarterly) to see whether a genuine trend is forming or whether an early read was noise.
