YouTube Trend Keyword Finder 🎯

Discover related YouTube keywords from one topic or a short seed list. Get ranked keyword ideas, supporting videos, view totals, recent-content shares and direct example links in a CSV-ready table.

What you get

• Localized YouTube autocomplete suggestions.
• Relevant phrases and hashtags appearing in sampled video titles and public description snippets.
• A transparent 0–100 popularity/recency score, not invented monthly search volume.
• Images, example URLs, sample counts and explicit nulls for unavailable metrics.

Input example

{"seedKeywords":["sourdough","ekşi maya","mechanical keyboard"],"maxKeywords":50,"sampleSize":30,"country":"US","language":"en","includeExamples":true}

Use 1–5 seeds, maxKeywords 1–100 per seed (default 50), sampleSize 4–40 (default 30). Country influences search and suggestions. Language influences suggestions; publication-label parsing uses English. includeExamples controls the nested example list, not the flat example URL.

How the score works

Candidates come from suggestions, 2–4-word phrases in titles/description snippets, and visible hashtags. Phrases must include a seed token, or the complete phrase must appear in at least two sampled titles. Repeated generic words alone do not make a phrase relevant. Case, accents, punctuation and hashtag prefixes are normalized. Each sampled video is counted once per keyword.

V = log(1 + summed matching-video views) / log(1 + maximum sampled-video views × sampleCount), capped at 1. F = matching-video count / sampleCount. R = 0.7 × share published within 7 days + 0.3 × share within 30 days, among matching videos with dates. P = 1 − (suggestionPosition − 1) / suggestionCount; 0 if not a suggestion. E = min(1, engagementRate / 10), where engagementRate is 100 × (likes + comments) / views if all components are available. Score = 100 × (40V + 25F + 20R + 10P + 5E) / available weight. R and E weights are omitted when unavailable; the base weight is 75. Scores are rounded to two decimals. Ties use frequency then keyword.

Half the video budget is drawn from relevance results and half from videos filtered to the recent month. This intentionally balances established interest with new content; the score is relative to this sample, not a global trend index. Publication timestamps are estimates reconstructed from relative labels (weeks = 7 days, months = 30, years = 365); examples include publishedAtApproximate and the original label. Borderline 7/30-day shares are approximate.

Output fields

Field	Meaning
seedKeyword, keyword, rank, score	Seed, normalized keyword and per-seed ranking
relation, sources	Discovery method(s): autocomplete, title_ngram, description_ngram, caption_ngram, video_tag or hashtag
frequency, sampleCount	Matching unique videos and all sampled unique videos
totalViews, avgViews	Sum/mean of known views across matching sampled videos
avgLikes, avgComments, engagementRate	Nullable: public search does not normally expose these
recencyScore	0–100 weighted recent-content share
recent7DayShare, recent30DayShare, matchedDateCount	Raw shares and date coverage
growthSignal, growthSignalReason	Null: one snapshot does not prove growth
suggestionPosition	Autocomplete position, or null
topExamples, topExampleUrl, thumbnailUrl	Up to three supporting videos and flat URL/image
platform, scrapedAt, scoreVersion, dataCoverage, id	Platform, observation time, formula version, coverage and stable seed/keyword ID

Limits and honest coverage

sampleSize is a maximum, not a guarantee. Duplicate videos across the two lists are counted once. Autocomplete-only keywords may have frequency 0 and no examples. No search volume or growth history is fabricated. Up to six sampled videos per seed are enriched with available video tags, full descriptions and publication dates. Public caption tracks are attempted when exposed and their first 60 words can contribute candidates. Captions may be unavailable. Channel tags are not fetched. Likes and comment counts are included only when an exact public count is exposed on the sampled watch page. Missing counts are null, never synthetic zeros.

Pricing and free plan

Pay per delivered keyword; no start fee. Free Apify plans have a shared 40-result allowance per Actor per user per UTC day, counted by requested results; all free-plan users also share the daily YouTube service budget. Paid Apify plans are not rate-limited. Input errors finish SUCCEEDED with no results and no charge; see OUTPUT for the reason. Genuine retrieval failures finish FAILED without delivery or keyword charges.

中文说明 🇨🇳

输入一个或多个 YouTube 主题，获得相近关键词、0–100 分排名、样本视频播放量、近期内容比例以及视频链接和图片。

关键词来自自动补全、公开视频标题和搜索结果中的描述片段、可见话题标签。默认每个主题最多 50 个词；视频样本默认 30 个，允许 4–40 个；最多 5 个主题。country 影响搜索与建议，language 影响建议。includeExamples 决定是否输出嵌套示例列表。

评分权重：播放量对数指标 40、出现频率 25、近期比例 20、建议位置 10、可用互动率 5。近期指标为 7 天比例×0.7 + 30 天比例×0.3。缺失近期或互动指标时移除该权重，并按可用权重归一化到 100。此分数仅针对当前样本，不是全站搜索量或真实增长率。

一半样本来自相关结果，一半来自最近一个月筛选结果。发布日期根据相对时间估算，示例标有 publishedAtApproximate。缺失的点赞、评论、互动率用 null 表示；单次采样不能证明增长，因此 growthSignal 为 null。每个主题最多补充六个视频的可用标签、完整描述和发布日期，并尝试公开字幕。不抓取频道标签；公开视频页显示准确点赞或评论数时才使用。字幕可能不可用。

按实际交付关键词计费，不收启动费。免费计划每个用户每天每个 Actor 共用 40 条请求额度，UTC 零点重置；所有免费用户同时共用每日 YouTube 服务预算。付费计划不受这些限制。输入错误以 SUCCEEDED 结束，无结果、无收费；原因请查看 OUTPUT。真正采集故障以 FAILED 结束，不交付、不收关键词费用。

Shared YouTube free-plan daily budget

All free-plan YouTube users share one $1 daily service budget across all YouTube Actors (not $1 per user or Actor), resetting at 00:00 UTC. When exhausted, the run finishes SUCCEEDED with 0 rows and no charge; OUTPUT contains ok:false and errorType:FREE_LIMIT. Any paid Apify plan is unlimited.

YouTube 免费计划共享每日预算

所有 YouTube 免费计划用户、所有 YouTube Actor 共用一个每日 $1 的服务预算（不是每位用户或每个 Actor 各 $1），于 UTC 00:00 重置。预算耗尽时运行以 SUCCEEDED 结束，0 条结果、无收费；OUTPUT 返回 ok:false 和 errorType:FREE_LIMIT。任何付费 Apify 计划均不受此限制。