Great videos don't find an audience on their own. YouTube's recommendation system is a behavioral prediction engine, it's constantly asking "what will this specific viewer watch next, and for how long," and it rewards videos that answer that question well before a single view even counts toward growth.
I optimize both the discovery layer (titles, descriptions, search visibility) and the behavioral signals YouTube actually weighs most heavily now: satisfaction, retention, and session value, not just keywords and tags. Long-form and Shorts run on genuinely separate algorithms in 2026, and I treat them that way rather than applying one playbook to both.
A short video with strong completion and a clear "like" signal increasingly outperforms a long video with heavy drop-off, even if the longer one technically accumulates more total watch minutes. The algorithm is optimizing for whether viewers were genuinely satisfied, not just how long they stayed.
Shorts get ranked on completion rate, loop rate, and swipe-away behavior within the first few seconds. Long-form gets ranked on session duration, retention curves, and click-through rate. Optimizing both the same way wastes effort on the one that doesn't apply.
They help with categorization and can nudge you into a competitor's suggested-video sidebar, but they're not the lever that moves rankings the way titles, thumbnails, and retention do. A lot of dated advice still treats tags as a primary lever; they aren't.
Properly labeled AI-assisted content gets normal distribution. Undisclosed AI content that YouTube detects can face reduced recommendations. This matters increasingly as AI tools get used in editing and production.
Titles and thumbnails that do two jobs simultaneously: signal relevance to the algorithm and earn the click from an actual human scrolling past dozens of other options. Descriptions and chapters structured for both YouTube's search and Google's, since YouTube videos increasingly surface directly in Google search results and AI-generated answers.
Identifying exactly where viewers drop off using your actual analytics data, then restructuring pacing, hooks, and content structure to hold attention longer. Retention curves matter more than almost any other single factor for long-form recommendation.
A genuinely separate approach: hook timing in the first half-second, loop structure so the ending flows back into the opening, captions (most Shorts get watched on mute), and trending audio where it's genuinely relevant, not as a separate afterthought bolted onto long-form strategy.
Branding, channel layout, and playlist architecture. Channels with multiple well-organized playlists consistently see meaningfully more session views than channels without them, since playlists directly extend session duration, a heavily weighted recommendation signal.
VideoObject schema markup and video sitemaps so your content is positioned to surface in Google search carousels and AI-generated overviews, not just within YouTube itself.
A content plan built around genuine audience demand (using YouTube's own search suggestions as real, verified demand signals) and aligned with your broader business goals, not just whatever idea is easiest to film next.
A full review of your current channel and video performance, including where retention drops and what's actually driving (or limiting) recommendations right now.
Identifying real search demand using YouTube's own autocomplete and search data, plus a look at what's working for channels you're actually competing against for the same audience.
Applying fixes to existing videos and channel setup, plus building the strategy for new content going forward, treating Shorts and long-form as distinct workstreams.
Tracking watch time, retention, satisfaction signals, and subscriber growth over time, since YouTube's algorithm shifts in smaller ways continuously and larger ways a few times a year.
Viewer satisfaction, increasingly measured through retention combined with explicit signals like likes and post-watch surveys, has overtaken raw watch time as the primary factor. A shorter video that fully satisfies viewers can outperform a longer one with heavy drop-off.
Yes, genuinely different ones. They run on fully separate ranking algorithms now: Shorts prioritize completion rate, loop behavior, and early engagement, while long-form prioritizes session duration and retention curves. Applying one playbook to both wastes effort on whichever doesn't match.
Less than most YouTube SEO guides suggest. Tags help with categorization and can occasionally influence suggested-video placement, but they're not a dominant ranking lever. Titles, thumbnails, and retention performance matter considerably more.
Typically 4–12 weeks for measurable movement, depending on niche competition, upload consistency, and how quickly retention and engagement improve on optimized content. Channel-level growth compounds gradually rather than appearing overnight.
Yes, as long as AI-generated or significantly AI-altered content is properly disclosed using YouTube's labeling requirements. Disclosed AI content gets normal distribution; undisclosed AI content that YouTube detects can face reduced recommendations.
Increasingly, yes. YouTube videos surface directly in Google search results and AI-generated answers, particularly with proper structured data (VideoObject schema, video sitemaps) and clear chapters or timestamps. Optimizing for YouTube and for broader search visibility overlap more than they used to.
If your YouTube strategy is still built around tags and keyword density, you're optimizing for an algorithm that changed years ago. Let's build one for how YouTube actually ranks content now.