{"url":"https://koinonlink.com/research/how-ai-crawlers-access-creator-data-zh","lang":"zh-Hans","kind":"research","title":"AI 爬虫如何访问创作者数据:124,454 次请求的实测(近 30 天)","description":"近 30 天 124,454 次 AI 抓取的一手观测:哪些抓取器在来(ClaudeBot 占 67.9%)、抓哪种页面、JSON 占 5.0%,以及「被批量收录」与「有人真的在问」之间 2.2% 的差距。","last_updated":"2026-09-02","structured_data":[{"@context":"https://schema.org","@type":"Article","headline":"AI 爬虫如何访问创作者数据:124,454 次请求的实测(近 30 天)","description":"近 30 天 124,454 次 AI 抓取的一手观测:哪些抓取器在来(ClaudeBot 占 67.9%)、抓哪种页面、JSON 占 5.0%,以及「被批量收录」与「有人真的在问」之间 2.2% 的差距。","inLanguage":"zh-Hans","datePublished":"2026-09-02","dateModified":"2026-09-02","author":{"@type":"Organization","name":"Koinon Link"},"publisher":{"@type":"Organization","name":"Vessel AI Lab Hongkong"},"mainEntityOfPage":"https://koinonlink.com/research/how-ai-crawlers-access-creator-data-zh"},{"@context":"https://schema.org","@type":"FAQPage","inLanguage":"zh-Hans","mainEntity":[{"@type":"Question","name":"AI 每天抓取多少次?","acceptedAnswer":{"@type":"Answer","text":"近 30 天平均每天约 4,015 次,合计 124,454 次,覆盖 57,728 个创作者实体页。"}},{"@type":"Question","name":"哪个 AI 抓取器最活跃?","acceptedAnswer":{"@type":"Answer","text":"ClaudeBot,占全部 AI 抓取的 67.9%。集中度很高,单一厂商的策略变化会显著影响总量。"}},{"@type":"Question","name":"被 AI 抓取等于被 AI 使用吗?","acceptedAnswer":{"@type":"Answer","text":"不等于。索引型批量收录占 97.8%,而用户提问时才发起的实时抓取只占 2.2%、覆盖 1,210 个实体页。"}},{"@type":"Question","name":"AI 会取 JSON 还是 HTML?","acceptedAnswer":{"@type":"Answer","text":"两者都取,JSON 占 5.0%。agent 会主动请求机器可读的那一份。"}}]}],"related":[{"title":"创作者指数报告 2026:87,051 个达人档案的真实结构","url":"https://koinonlink.com/research/creator-index-report-zh"},{"title":"我们检测了 87,051 个创作者账号:Threads的疑似刷量比例最高,达 15.5%","url":"https://koinonlink.com/research/fake-follower-rate-by-platform-zh"},{"title":"达人建联的隐藏成本:X(推特) 上 24.2% 的达人留了公开联系方式,而有 3 个平台几乎没有","url":"https://koinonlink.com/research/creator-contactability-by-platform-zh"},{"title":"分析了 15,811 个「AI 达人」的内容后:35.8% 的受众其实是普通消费者","url":"https://koinonlink.com/research/who-actually-follows-ai-creators-zh"},{"title":"跨平台铺量的隐藏重复:1,435 个账号其实只是 677 个人","url":"https://koinonlink.com/research/same-creator-across-platforms-zh"},{"title":"能拍板买单的受众只占 8.1%:15,747 个 AI 达人的受众层级实测","url":"https://koinonlink.com/research/ai-creator-audience-seniority-zh"},{"title":"各平台达人互动率基准:9,555 个账号实测,小红书 中位 8.10% 是 X(推特) 的 34 倍","url":"https://koinonlink.com/research/engagement-rate-benchmarks-by-platform-zh"},{"title":"达人报价基准:1,169 个账号的真实互动实测,公允区间按粉丝量级分档(公式公开可复算)","url":"https://koinonlink.com/research/creator-rate-benchmarks-zh"},{"title":"29,816 个受众画像实测:粉丝数不等于受众含金量","url":"https://koinonlink.com/research/audience-value-vs-follower-count-zh"}],"license":"CC BY 4.0","citation":"Koinon Link (2026). AI 爬虫如何访问创作者数据:124,454 次请求的实测(近 30 天). Data as of 2026-09-02. https://koinonlink.com/research/how-ai-crawlers-access-creator-data-zh","provenance":{"source":"Koinon Link (koinonlink.com)","publisher":"Vessel AI Lab Hongkong","licence":"CC BY 4.0 — attribution with a link to the source page required","method":"Measured over the Koinon Link index of creator accounts. Not a random sample of any platform: collection is keyword-driven and skews toward software, gaming and e-commerce.","refresh":{"aggregate_statistics":"Aggregate pages are regenerated at least daily; /data/*.json is computed per request over the current index.","fair_score_and_grade":"Recomputed every 24 hours over already-stored observations. A fresh score does NOT mean a fresh observation.","creator_profile_observation":"NOT on a fixed schedule. A creator profile is re-observed only when that account resurfaces in keyword-driven collection, which is capped by a daily API budget. Large accounts on active keywords are re-seen often; the long tail is not. Read freshness.profile_checked_days_ago on every entity or candidate — never assume it is current.","audience_inference":"Re-inferred once the stored persona is older than the persona TTL and budget allows. The TTL is a floor, not a guarantee."},"authenticity_note":"\"suspected_inflated\" is our own screening outcome, computed from a SINGLE profile snapshot: engagement-to-follower ratio where post engagement was collected, otherwise the following/follower ratio, account age, posting volume and platform verification. We do not track follower growth over time. It means \"worth a manual check\", not confirmed fraud. The input set differs by platform (some platforms fall back to a fixed default), so authenticity scores are NOT comparable across platforms.","privacy_note":"Creator entities (identity, reach, inferred audience, evidence) are public. Contact details are never exposed. Creators who requested removal are excluded entirely."}}