Quora Answers Scraper

Give it one or more Quora question URLs and get one flat row per answer: the full answer text, who wrote it, the credential Quora prints under their name, upvotes, comments, views and the date it was written.

No account, no cookies, no login.

What one row contains

Field	Notes
answer_id	Quora's own numeric answer id, as a string. It is an identifier, not a quantity - kept as text so a spreadsheet cannot round it.
question_url	Verbatim the URL you asked for, on every row, so rows from a multi-question run group on it.
question_title	Quora's title for the question that URL resolves to. Quora merges duplicate questions, so an older slug can come back under a newer title - that is Quora's answer, not a mismatch.
answer_text	Plain text. Quora stores prose as a structured document; paragraphs and list items become lines, embedded images contribute nothing rather than a blank line.
answer_url	Permalink to this answer. For a mixed-in answer (see below) this points at the question it was actually written under.
is_direct_answer	False when Quora served this as a relevant answer written under a different question. Read this before treating a row as an answer to your question.
author_name	Null on anonymous answers.
author_credential	Quora's one-line credential - "Aerospace Engineer at British Aircraft Corporation (2014-present)". Null when the author never wrote one.
author_url	Profile link, null for anonymous.
upvote_count, comment_count, view_count	Integers, or null where Quora did not report them.
answer_date	ISO 8601 UTC.

Why the counts are null and not zero

Quora omits the field entirely on some answers rather than sending a zero. A fabricated 0 would read as a measurement - an answer nobody upvoted - and someone would rank on it. null says "Quora did not tell us", which is the truth.

Relevant answers

A Quora question page is not a pure list of answers to that question. Quora mixes in strong answers written under closely related questions, because to a reader they answer the same thing. Those rows are real answers with real authors, but they are not answers to the exact question you asked for.

Two ways to handle it:

• Leave Answer ranking on the default and filter on is_direct_answer afterwards. You get both sets and can decide later.
• Set Answer ranking to Only this question's answers, which sends Quora's own hide_relevant_answers ordering. Measured 2026-08-25: it shifts the mix but does not empty it - a run with it on still returned some rows with is_direct_answer: false. It is a hint to Quora's ranker, not a filter, so is_direct_answer is still the field to trust.

How it fetches

Quora renders a question page from an embedded Relay payload and refills it from /graphql/gql_para_POST, its own persisted-query endpoint. This actor uses that endpoint rather than scraping rendered HTML: it loads the question page once for the CSRF formkey and the numeric question id, then pages the answer list as JSON. Nothing is parsed out of markup, so a cosmetic redesign does not break it.

The connection behind that endpoint is a page list, not an answer list - ad slots, section headers and related-question blocks arrive on it as siblings of the answers. Those are dropped, never counted and never charged for.

Quora reads the TLS handshake, not the User-Agent. Plain HTTP clients get HTTP 403; a real Chrome fingerprint (curl_cffi) got HTTP 200 on every request during testing on 2026-08-25.

Residential proxy is the default and no country is pinned. A datacenter address gets a login wall served with HTTP 200 - a refusal wearing a success code - and this actor treats a page with no formkey in it as exactly that: it rotates to a new exit address and retries, rather than reporting a question with thousands of answers as empty.

What this actor does not do

There is no search-term input. Quora's own search is behind a hard login wall for logged-out visitors: quora.com/search?q=... redirects to a sign-in page and its search query returns null, measured 2026-08-25. Rather than ship a field that silently returns nothing, this actor takes question URLs only.

Pricing

Pay per event. A small actor-start fee, then one charge per answer row. A run that legitimately finds nothing still costs only the start fee.