A public Quora question page gives up its head over plain HTTP from a German or Indian residential exit: the title, the canonical URL, the answer count and a 300-character opening of the top answer, 19 to 22 words in 233,743 to 250,265 bytes. The full body, 2,254 words on the page we measured, arrives only when a browser renders it from a US exit, for 1,246,131 bytes and 18.7 seconds. On Quora the setting is two settings: a German or Indian exit over plain HTTP for the head, a US exit rendered for the text.
The search results for this keyword are Quora threads about proxy vendors and vendor pages about running several Quora accounts. That second intent is not what this article is for, and we say so below in one line. What it is for is the question nobody on the first page answers: what does a question page actually return to a proxied client, by country, with and without JavaScript? We measured it on 28 September 2026.
Plain HTTP returns the head from Germany and India, and 6 words from the US
We audited one public question, quora.com/What-is-a-proxy-server, which Quora redirects to its canonical slug, through residential exits in three countries.
| Exit country | Plain HTTP | Rendered in a browser |
|---|---|---|
| Germany | 19 words: title, description, canonical, Open Graph; 250,265 bytes | 44 words, no canonical |
| India | 22 words: title, description, canonical; 233,743 bytes | 43 words, no canonical |
| United States | 6 words; 6,003 bytes | 2,254 words, title, canonical, h1; 1,246,131 bytes |
Look at what the German plain response contains, because it is more than it sounds. The title is the question. The canonical is the slug Quora considers primary, which is how you deduplicate the same question reached under three URLs. The meta description opens with Answer (1 of 48) and then the first 300 characters of the top-ranked answer. The Open Graph tags carry the same title and URL. For a monitoring job that needs to know which questions exist, how many answers they have and what the top answer says first, that is the record, and it costs a quarter of a megabyte with no browser.
The body is not in that response. Quora's question page is a React application: the 250 KB is a shell, the preload headers list eleven webpack bundles, and the answer text arrives after they run. The 19 to 22 words the audit counts are the head and a footer, and that is all plain HTTP will ever return on this platform, from any country.
The body needs a browser, and the browser needs a US exit
Rendered from the US, the same page returned 2,254 words: the h1, every answer the logged-out view exposes, author bios and credentials, and the canonical. The rendered DOM declares 64 Answers at the top, where the description said 48, so the two counters measure different things (the description counts ranked answers at the time it was generated) and you should store both. The render took 18.7 seconds and 1,246,131 bytes, five times the plain fetch.
Rendered from Germany and India, the same page returned 43 to 44 words with no canonical, the audit's pass from India taking 49 seconds to say so. That is the inversion that makes Quora different from every other platform in this series: the plain fetch works from Europe and Asia and not from the US, and the rendered fetch works from the US and not from Europe and Asia. Neither mode works everywhere. You pin two countries, one per mode.
Two more facts from the rendered pass. First, the logged-out rendered page carries meta robots noindex,nofollow, while the German plain response carries noarchive only: Quora tells search engines not to index what an anonymous browser sees. Second, the logged-out view ends where the sign-in prompt begins. Of the 64 answers the page declares, the render contains the ones Quora shows before that prompt and a single further line asking to continue; we did not count how many answers that is, only that the total text is 2,254 words. There is no logged-out pagination, so the depth of one question is the depth of one render.
What it costs, counting a gigabyte as one billion bytes
| How you fetch it | On-wire bytes | What you get | Per GB | Cost at $0.80/GB |
|---|---|---|---|---|
| Head, plain HTTP, German exit | 250,265 | title, canonical, answer count, 300 chars of top answer | ~3,996 pages | $0.00020 |
| Head, plain HTTP, Indian exit | 233,743 | same fields | ~4,278 pages | $0.00019 |
| Body, rendered, US exit | 1,246,131 | 2,254 words, h1, all logged-out answers, 18.7 s | ~802 pages | $0.00100 |
| Rendered, DE or IN exit | n/a | 43 to 44 words, no canonical | n/a | wasted |
Read the first and third rows as a pipeline, not as alternatives. Discover and classify questions with the head fetch at $0.20 per thousand pages; render only the questions whose answer count or top-answer opening says they matter, at $1.00 per thousand. A job that tracks 50,000 questions weekly and renders the 5,000 that changed costs about $10 in head fetches and $5 in renders per week. Rendering all 50,000 would cost $50 and take, at 18.7 seconds each on one lane, nearly eleven days.
The web scraping API prices the same two modes at $0.0002 per page plain and $0.001 rendered, which is the same arithmetic without running a browser pool. Ask it for the plain mode from a German exit and the rendered mode from a US one; asking for the rendered mode from Germany is the wasted row in the table.
Which proxy type Quora needs, and which countries
Residential is the network that produced every successful pass above, from $0.80/GB. We did not measure anything cheaper, and the geography of the results says the edge in front of Quora is making decisions per session that a datacenter range would not improve. Mobile at $2.30/GB buys nothing here: the rendered US pass succeeded on a residential exit on the first attempt.
Country is the whole decision on this platform, so here it is as a rule. Head reads: Germany or India, where plain HTTP returned title, canonical and description in 250,265 and 233,743 bytes. Body reads: the United States, where the render returned 2,254 words. Never the other way round: a US plain fetch returned 6 words and a German or Indian render returned 43 to 44. Keep the two lanes on sticky sessions so a question's head and body are fetched from a consistent pair of exits.
Rate limits: Quora documents none for anonymous reads, and nothing in nine passes throttled us. The Crawl-delay is absent from robots.txt. Hold one request per second per exit per lane and add exits for concurrency.
What Quora's own files permit, in one line each
The robots.txt is 18,849 bytes and is the clearest statement on this platform. It opens with a notice that all crawlers are prohibited from using Quora content to train AI models without a contractual licence. For an unnamed client, User-agent: *, it allows six paths (the home page, about, press, login, signup and one campaign path) and then Disallow: /: every question page is off limits to a generic bot. Googlebot is allowed question pages but not */answer/ or */unanswered/*. Eighteen crawlers are excluded from the whole site by name, among them GPTBot, ChatGPT-User, OAI-SearchBot, ClaudeBot, Claude-User, PerplexityBot, Bytespider, Scrapy and ia_archiver; the file explains the archive exclusion at length, as protection for authors who later anonymise answers.
The Terms of Service page itself returned the same 6 KB response as a question page to our plain client, so we cite it without quoting it; Quora's own answers on the subject say automated reads and bot actions are not allowed and that accounts and IP addresses engaged in them are removed. That is the plain line: reading Quora at volume as an unnamed client is disallowed by its robots.txt and its terms, and no proxy changes that. Running several Quora accounts through proxies, which is what the vendor pages on this keyword sell, is account abuse and we will not help with it.
What this measurement is properly for: checking how your brand's questions and answers render from the countries your customers search from, monitoring a known list of public questions where your product is discussed, and answering the honest engineering question of what a logged-out visitor sees before you decide whether a job is worth doing at all. Do not automate answers, votes or follows, and do not read anything behind the sign-in prompt.
The setting that works on Quora
- Network: residential, Basic at $0.80/GB. Every successful pass in nine came from a residential exit; mobile adds nothing measured.
- Fetch mode: two lanes. engine: tls for the head: 19 to 22 words with title, canonical, answer count and the top-answer opening in 233,743 to 250,265 bytes. Rendered for the body: 2,254 words in 1,246,131 bytes and 18.7 seconds.
- Country to pin: Germany or India for the plain lane, the United States for the rendered lane. Swap them and the numbers become 6 words and 43 to 44 words.
- What to store: both answer counts, 48 from the description and 64 from the DOM; they diverge and both are real.
- When the proxy is not enough: the web scraping API runs both lanes for $0.0002 and $0.001 per page; for discovering which Quora questions rank for a topic without touching Quora, the Google search collector returns question URLs, titles and snippets as rows.
- Free tier: every account gets $2 of free API usage per month, which covers about 10,000 head fetches or 2,000 rendered questions.