Video Search

We work across YouTube, TikTok and Bilibili. Each platform has a distinct recommendation engine, a distinct audience composition, and distinct ranking signals — and we treat them distinctly rather than applying a single template. YouTube is a long-form search engine with a strong suggestion algorithm; TikTok is a short-form interest graph where virality and watch-time completion rates interact in ways that are not fully transparent; Bilibili is a Chinese video platform with a highly engaged ACG-origin community and a growing professional content vertical that rewards series-format content.

On YouTube, our work starts with a channel-level keyword and topic audit. We use Ahrefs's video SERP data and YouTube autocomplete to map the queries your channel could realistically rank for, stratified by volume, competition and intent alignment. The output is a topic cluster map — a set of interconnected subjects where one anchor video can accrue authority that lifts related content. We then produce a metadata playbook: title formula, description structure, chapter timestamp conventions, card and end-screen placement, and tag strategy. Small, consistent metadata improvements compound faster on YouTube than almost any other SEO channel because the platform's own search and browse systems amplify them immediately.

Thumbnail quality is the primary driver of click-through rate from both search and suggested feed. We run A/B tests on thumbnail concepts — not design opinions, but structured experiments with data from YouTube Studio's impression CTR panel. A consistent thumbnail language builds channel-level recognition that lifts CTR across all videos over time, not just the ones being tested.

Retention curve analysis is where the content feedback loop closes. YouTube's analytics show per-segment audience retention with relative performance benchmarks. We read those curves diagnostically: where do viewers drop off, what does a healthy vs. unhealthy hook look like for this audience, and how does the retention shape correlate with the algorithm surfacing the video in browse feed vs. suggested. We bring these readings back to the production team as specific, actionable briefs — not general advice about "engaging content."

For TikTok, the work is different. Short-form search on TikTok is growing — the platform reported that over 40% of Gen-Z users use TikTok as a search engine for restaurant and product discovery. We build metadata systems for TikTok that work for both the interest graph (hashtag strategy, audio selection, posting cadence) and the emerging search layer (keyword seeding in captions, text overlays timed to important terms).

Bilibili requires understanding its community structure. The platform's quality rating (弹幕 density, coins, favourites) matters as much as view count for surface placement. We build content series designed for Bilibili's archival audience — people who binge a creator's library — and programme launch cadences around Bilibili's own seasonal promotion cycles. For brands targeting younger Chinese consumers, Bilibili is underused relative to its reach in that demographic.

We measure video SEO through platform analytics — YouTube Studio, TikTok Creator Center, Bilibili Studio — aggregated in a Looker Studio report that tracks impressions, CTR, average view duration, and subscriber conversion rate. For brands that have Google Search Console access, we also track video rich result appearances in web SERPs, which are often overlooked as a traffic source.