Documentation Python quickstart Blog Free tools hello@quanticdata.ioLog in

Kuaishou Proxies: 1,009 Words, Zero Browser

What Kuaishou costs in bytes, measured on 28 September 2026: 341,397 bytes for a public profile over plain HTTP with 1,009 words and 31 video cards, 606,175 bytes for the same profile rendered with no extra words, and 76,153 bytes for a short-video page that needs the browser
What Kuaishou costs in bytes, measured on 28 September 2026: 341,397 bytes for a public profile over plain HTTP with 1,009 words and 31 video cards, 606,175 bytes for the same profile rendered with no extra words, and 76,153 bytes for a short-video page that needs the browser

Kuaishou is the second short-video platform of China, with a public web that almost nobody outside the country has looked at. On 28 September 2026 we fetched it 13 times through residential exits in Singapore and the United States. A public profile answers a plain HTTP client with 341,397 bytes, 1,009 words and 31 video cards from Singapore, and 341,309 bytes and 1,010 words from the United States. Rendered, the same profile is 606,175 bytes, 39.7 seconds and the same 1,009 words. The head is generic on every profile, the video page is the one surface that needs a browser, and robots.txt allows fourteen named crawlers, GPTBot included, and nobody else.

The keyword is empty, which is why this page exists

Type kuaishou proxies and Google suggests privacy policy, pay, pronunciation and platforms. The United States SERP has no proxy vendor on it at all: a South China Morning Post explainer, investor relations, Wikipedia, a CNCF case study about container infrastructure, a Reddit thread, a YouTube registration walkthrough, an AAAI paper whose title happens to contain the word proxy, a Caixin story about data-collection fines, an open issue asking a download tool to support kuaishou.com, and Reuters. Kuaishou api autocompletes to the company's video-generation model, not to the social platform.

So there is nothing to beat and nothing to cite. This page is the first measurement of what kuaishou.com returns to a proxied client outside China, for the reader who wants public creator profiles and video listings for brand research, creator discovery and market monitoring, with the platform's own rule quoted where it exists.

The profile is server-rendered from both countries

We requested kuaishou.com/profile/3x45a2tt9zys2yw, the official account of China Post, as a plain HTTP client with no JavaScript, and then rendered it:

FetchExitStatusBytesMarkdown wordsVideo cardsTime
Profile, plain HTTPSingapore200341,3971,009314.0 s
Profile, plain HTTPUnited States200341,3091,0103134.0 s, two attempts
Profile, renderedSingapore200606,1751,0093139.7 s
Short-video page, plain HTTPSingapore20076,1532002.3 s
Discover hub, plain HTTPSingapore200145,35722503.3 s

The profile body arrives in the HTML: the video titles, the 31 cover images and the account's display text are all in the 341,397-byte response, and the markdown conversion counts 1,009 words and 15,143 characters. The rendered pass returned 1,009 words from 606,175 bytes after 39.7 seconds. That is 1.78 times the bytes and ten times the seconds for zero additional words. On this platform the browser is not harmful, it is merely pointless for profiles.

The audit word count tells a different story, and it is worth explaining because it will trip a naive check. The SEO audit counted 14 words without JavaScript and 14 rendered, from both countries. That count splits on whitespace, and Chinese text has none, so a page full of Chinese content scores as a handful of words. Assert on bytes and characters for CJK pages: 341,397 bytes and 15,143 characters mean content; 76,153 bytes and 113 characters mean a shell. A word-count threshold tuned on English will reject every real page on this site.

The head lies about which page you are on

Every profile we fetched carried the same head: title 快手, the app's name; a meta description that is the app's marketing line; a canonical of https://www.kuaishou.com/profile/ with the account id stripped; an Open Graph URL identical to the canonical; and a JSON-LD block of type BreadcrumbList only. Nothing in the head identifies the account. The pattern we found on Instagram and in follower counts in the head, where the meta description carries the counts, does not apply here: on Kuaishou the head is a template and the body is the data.

Two consequences for a parser. First, the canonical cannot be used to deduplicate or to confirm the fetched account; use the id from the request URL and the display name from the body. Second, a check that classifies pages by title will file every Kuaishou profile as the home page. Classify on the body: the number of video cards is the discriminator, 31 on a real profile and 0 on the hub and the video page.

The video page is the surface that wants the browser

A short-video page, /short-video/ plus the video id, returned 76,153 bytes and 20 words over plain HTTP: a JavaScript shell with the same generic head. The discover hub at /brilliant returned 145,357 bytes and 225 words, mostly navigation and licence links, with no profile or video URLs in it. The sitemap is 1,116 bytes and lists ten corporate URLs, none of them content. So on Kuaishou the profile is the entry point and the only rich server-rendered surface: read profiles over plain HTTP, collect the video ids from their 31 cards, and render only the individual video pages you actually need.

One timing note. From Singapore the plain-HTTP profile took 4.0 seconds; from the United States it took 34.0 seconds over two attempts, and the rendered audit from Singapore needed a second attempt after the first timed out. The bytes were the same from both exits within 88 bytes. Pin Singapore for throughput on this platform; the content does not change, the latency does.

What robots.txt says, and the polarity it has

kuaishou.com/robots.txt is 604 bytes and it is an allow-list, the inverse of what most platforms publish. Fourteen named agents each get Allow: /: Baiduspider, Baiduspider-video, Googlebot, YoudaoBot, two Sogou spiders, 360Spider, bingbot, and then PerplexityBot, Perplexity-User, GPTBot, ChatGPT-User, OAI-SearchBot and deepseekbot. The file closes with:

User-agent: *
Disallow: /

Where TikTok names 25 AI crawlers to keep them out and Naver prints an all-caps notice against AI training, Kuaishou names six AI crawlers to let them in. The rule for everyone else is the same as on Xiaohongshu: unnamed clients are told to stay out of everything. The developer portal at open.kuaishou.com is browser-only, 50,971 bytes and 0 words without JavaScript, so we could not read the platform's terms as a plain client and do not claim to know their wording. What we can say is the usual line for this family: public profiles and video listings are readable as measured; logged-in actions, follows, likes and comments are not something we help automate.

What it costs, and what to stop paying for

Every number came through residential proxies on the Basic line at $0.80/GB, counting a gigabyte as 10^9 bytes:

JobBytes per requestRequests per GBCost per request
Profile with 31 video cards, plain HTTP341,3972,929$0.00027
Profile, rendered606,1751,650$0.00048
Short-video page, plain HTTP (shell)76,15313,131$0.000061
Discover hub, plain HTTP145,3576,880$0.00012

Ten thousand profiles is 3.41 GB over plain HTTP, about $2.73, and 6.06 GB rendered, about $4.85, for the same 1,009 words each. On mobile proxies at $2.30/GB the plain-HTTP profile is $0.00079; nothing we measured examined the network type, so the cheaper line is the one to buy. The 31 video ids per profile are the real yield: at 2,929 profiles per gigabyte that is about 90,800 video references per gigabyte before you render a single one.

The setting that works on Kuaishou

  • Network: residential, Basic line at $0.80/GB. Both exits returned the profile within 88 bytes of each other; the network type is not what this platform examines.
  • Fetch mode: engine: tls, plain HTTP. The profile returns 1,009 markdown words, 15,143 characters and 31 video cards from 341,397 bytes; rendered it is 606,175 bytes and 39.7 seconds for the same 1,009 words.
  • Country: Singapore. The content was identical from the United States, but the plain-HTTP fetch took 34.0 seconds there against 4.0 seconds from Singapore, and both rendered audits needed a second attempt.
  • Parser rule: ignore the head. Title, description and canonical are the same template on every profile, with the id stripped from the canonical; identify the page by the request URL and the body, and classify by video-card count.
  • When the proxy is not enough: there is no Kuaishou collector in our catalogue today. For the individual video page, which is a 76,153-byte shell over plain HTTP, the web scraping API with rendering returns the page at $0.001; for profiles the same API without rendering is $0.0002 and needs no browser.
  • Every account gets $2 of free API usage per month.

For the platform with the opposite shape, where the profile is an empty shell and the counters come from a keyless endpoint, read Douyin; for the one where the exact counters ship inside the plain HTML, Bilibili.

Sources & further reading

FAQ

Quick answers on kuaishou proxies.

Something else? Ask us →

Do I need a proxy to read public Kuaishou profiles?

Not for a handful of requests: all 13 of our fetches on 28 September 2026 returned HTTP 200 from residential exits in Singapore and the United States, with the profile within 88 bytes of the same size from both. You need a pool for volume and an exit near the platform for speed: the plain-HTTP profile took 4.0 seconds from Singapore and 34.0 seconds over two attempts from the United States.

Does rendering a Kuaishou profile add anything?

No. The profile returned 1,009 markdown words and 31 video cards from 341,397 bytes over plain HTTP in 4.0 seconds, and 1,009 words from 606,175 bytes after a 39.7-second render. That is 1.78 times the bytes and ten times the time for zero additional words. Reserve the browser for the short-video page, which is a 76,153-byte, 20-word shell without JavaScript.

Why does every Kuaishou profile have the same title and canonical?

Because the head is a template. Every profile we fetched carried the title 快手, the app marketing line as its description, a canonical of https://www.kuaishou.com/profile/ with the account id removed, and a BreadcrumbList as its only JSON-LD. Nothing in the head names the account, so deduplicate on the request URL and parse the display name and the 31 video cards from the body.

Why does the SEO audit count only 14 words on a Kuaishou page?

Because it counts on whitespace and Chinese has none. The same 341,397-byte response that the audit scores at 14 words converts to 1,009 markdown words and 15,143 characters. For CJK pages assert on bytes and characters: 15,143 characters is a profile, 113 characters is a shell. An English word-count threshold rejects every real page on this site.

Does Kuaishou allow scraping in its robots.txt?

Only for fourteen named crawlers. The 604-byte file grants Allow: / to Baidu, Google, Youdao, Sogou, 360, Bing, PerplexityBot, Perplexity-User, GPTBot, ChatGPT-User, OAI-SearchBot and deepseekbot, and closes with User-agent: * / Disallow: /. It is the inverse of TikTok’s file, which names 25 AI crawlers to exclude them. An unnamed client is told to stay out; the developer portal that would carry the terms is browser-only at 50,971 bytes and 0 words.

What does it cost to collect Kuaishou profiles at scale?

Ten thousand profiles over plain HTTP is 3.41 GB, about $2.73 at $0.80/GB on residential Basic, and each profile carries 31 video cards, so that is roughly 310,000 video references. The same ten thousand profiles rendered is 6.06 GB, about $4.85, for the same words. Counting a gigabyte as 10^9 bytes, a profile is $0.00027 over plain HTTP and $0.00048 rendered.

Read Kuaishou profiles at 2,929 per gigabyte

Residential exits from $0.80/GB return the full profile with 31 video cards in 341,397 bytes over plain HTTP from Singapore in 4.0 seconds, and the web scraping API renders the individual video pages at $0.001 each when you need them. Every account gets $2 of free API usage per month.

Related reading

Social proxiesDouyin Proxies: 2.1 KB a Profile via the API

Nineteen fetches of Douyin on 28 September 2026. A public user page answers a plain HTTP client with 72,914 bytes and zero words from every country, and rendering it worked from Germany in 10.5 seconds, took 65 seconds from Singapore and did not finish from the United States. The endpoint that does the job needs no key: 2,100 bytes of JSON with follower and like counts, identical from all three exits. That is 3,902 times less than the 8,194,019-byte render.

Read →
Social proxiesNaver Proxies: 768 Words Without a Browser

Twenty fetches of Naver on 28 September 2026. The desktop blog is a frame that returns 0 words with or without JavaScript; the same post on the mobile subdomain returns 768 words in 155,493 bytes over plain HTTP, identical from Korea, the United States and Germany. The blog RSS hands over 50 posts in 99,632 bytes, a public cafe serves 8,280 words in a 1990s Korean charset, and the Open API answers a keyless request with 113 bytes and a documented quota of 25,000 calls a day.

Read →
Social proxiesXiaohongshu Proxies: 70 KB a Note, No Browser

Twenty-four fetches of Xiaohongshu on 28 September 2026. A public note requested with the token the feed hands out answers a plain HTTP client with 70,499 bytes, the full note text in the meta description and an Article JSON-LD block; rendered, the same note weighs 2,475,522 bytes. The explore feed is 186,518 bytes of server-rendered cards. The exit country changes the feed, not the note.

Read →