Nextdoor is the neighbourhood network behind an address check, and it is also a platform that publishes more to a logged-out client than its reputation suggests. On 28 September 2026 we fetched it nineteen times through residential exits in the United States, Canada and the United Kingdom. A public city page returned 2,404 words to a plain HTTP client, byte-for-byte the same from all three countries: resident counts, six neighbour conversations with their neighbourhood and verification year, 1,147 groups, six marketplace listings with prices and 44 events. Rendering the same URL cost 1,046,614 bytes and returned 322 fewer words. The browser is the wrong tool here, and the numbers say so.
On this keyword, "proxy" means shareholders and census forms
Search for Nextdoor proxies and Google cannot complete the phrase: the autocomplete box is empty for the head term, for "proxies for nextdoor" and for "nextdoor scraper". The results page explains why. Rank one is the SEC filings page of Nextdoor Inc., because a proxy statement is what a listed company mails before its annual meeting. Rank three is a Reddit thread from census enumerators about proxy respondents. In between sits one vendor page selling per-IP packages "for nextdoor.com", and further down a forum thread whose only technical advice is to use "proxies tied to the GEO, though as Nextdoor is hardcore local".
That last line is the one claim worth testing, and we tested it. The rest of this post is about reading what Nextdoor serves to anyone, for local market research, safety and public-agency monitoring, marketplace price tracking and neighbourhood-level demographics. It is not about the feed behind the address verification, and the last section says where that line is.
A city page is 2,404 words of structured local data without JavaScript
We audited nextdoor.com/city/san-francisco--ca/ from three countries, each time as a pure HTTP client and as a full browser.
| Exit | Plain HTTP words | Rendered words | Plain HTTP bytes | h1, plain / rendered |
|---|---|---|---|---|
| United States | 2,404 | 2,082 | 676,650 | San Francisco / San Francisco, California |
| Canada | 2,404 | 2,082 | 676,641 | San Francisco / San Francisco, California |
| United Kingdom | 2,404 | 2,202 | 676,643 | San Francisco / San Francisco, California |
Three exits, one page, nine bytes of variance. The response carried the same x-nextdoor-langpref: en-us header from all three countries and never redirected a British exit to the .co.uk site. "Hardcore local" describes the product, not the server: the public pages are the same everywhere, which means geo targeting on Nextdoor buys nothing for reading, and a pool can be spread across countries for throughput without changing a single byte of output.
What those 2,404 words contain is the point. In one plain response:
- Demographics with a stated source: 851,036 residents from US Census data, plus Nextdoor’s own scores for safety (50), affordability (63), friendliness (80) and family (85), 39% homeowners, average age 39, average income $137K.
- Six neighbour conversations, each with initials, neighbourhood, age of the post and the year the author’s address was verified ("Verified in 2014"), including two lost-pet posts with the PawBoost case numbers.
- Groups: 1,147 near the city, with three named and their member counts (317, 177, 152).
- Six marketplace listings with price and neighbourhood ($2,500 · East Marina, $1,500 · East Marina, two free items, two recently sold), and a "1,000+ listings" count.
- Events: three named with attendee counts, out of 44.
- A FAQ block in plain text, a list of roughly 230 neighbourhoods, a business category index, and a BreadcrumbList in JSON-LD.
The neighbourhood page one level down is the same shape at 626,406 bytes: /neighborhood/westportal--san-francisco--ca/ returned 3,779 residents, a friendliness score of 91, 81% homeowners, six conversations, three groups with member counts, a coastal flood advisory, marketplace items with prices, and a "Fave Awards" list of local businesses with their vote counts, led by a trattoria at 714. That block is the closest thing to a public business ranking on the platform, and it is server-rendered.
Rendering costs 1.55 times the bytes and loses 322 words
The rendered pass is the one to strike out of the budget. From the US it moved 1,046,614 bytes in 9.8 seconds and the audit counted 2,082 words against 2,404 without JavaScript. The h1 changed from "San Francisco" to "San Francisco, California", which is the one thing the browser improved, and nothing else appeared that the plain response did not already carry. A full text extraction of the rendered DOM reached 5,560 words, but the extra is the neighbourhood and category link lists expanded, not new data.
This is the pattern we keep finding on platforms that server-render for search engines: the HTML is the product, the JavaScript is the app shell for members, and a logged-out browser gets the shell on top of the HTML and pays for both. For the platform in this series where the opposite was true, where the render was the only way to read anything, see TikTok.
robots.txt leaves two paths open and closes everything else
nextdoor.com/robots.txt is 2,990 bytes and it is written by name. Nine crawlers (Googlebot, bingbot, Slurp, msnbot, DuckDuckBot, ia_archiver, Teoma, SemrushBot, Seobility) share ten path exclusions: join, invitations, unsubscribe endpoints, password reset and /businesses/. Twitterbot is allowed onto /pages/, /events/, /agency/ and the city feed for link previews. GPTBot is refused everywhere. Then the rule for everyone else:
User-agent: *
Disallow: /
Allow: /link_preview_image/
Allow: /for_sale_and_free/Two things follow. An unnamed client is invited to exactly two places: link-preview images and the marketplace. Everything else we measured above is published for search engines and served to anyone, but the file does not say it is for you. Read it as the platform’s statement of intent, and read the city and neighbourhood pages the way a search engine does: slowly, for research and monitoring, not as a feed. There is no sitemap line in the file at all; the neighbourhood list on each city page is the index.
One absence is worth recording. The city page lists dozens of business categories, but not one /pages/ link appears in its HTML (we extracted every href and found zero). Business pages exist, Twitterbot is allowed onto them, and the public site does not link to them from the pages we could reach. If business pages are your target, the route is not the public HTML.
The Display API is the route to posts by keyword and radius
Nextdoor publishes a developer portal, and it is more useful than the SERP suggests. The Display API documentation describes a Search endpoint that takes latitude, longitude, radius, category and keywords and returns public "anyone" posts from the last 30 days with title, description, photos, comment and reaction counts and the neighbourhood name; active marketplace listings with price, location and photos; upcoming events and business pages, the last two marked as coming soon. A second endpoint returns the 100 most engaging posts per city, and a third the posts and contact details of over 5,500 verified public agencies.
Access is by application form, reviewed case by case, and the page says plainly that approval is not guaranteed; the separate Ads API is for advertising partners and offers no research exceptions. There is no endpoint that answers without a token, so we could not weigh a response. But for the two jobs people actually want from this platform, keyword monitoring around a point and marketplace tracking, that API is the only path with permission attached, and it is where a proxied crawl of city pages should hand over.
What a gigabyte buys on Nextdoor
Nothing in nineteen fetches looked at the address. Every response was a 200 from a rotating residential exit on the first attempt, from three countries, with identical bodies. The cost driver is bytes, and the arithmetic below uses residential proxies on the Basic line at $0.80/GB, counting a gigabyte as 10^9 bytes.
| Request | Bytes | Requests per GB | Cost per request |
|---|---|---|---|
| City page, plain HTTP | 676,650 | 1,478 | $0.00054 |
| City page, rendered | 1,046,614 | 955 | $0.00084 |
| Neighbourhood page, plain HTTP | 626,406 | 1,596 | $0.00050 |
Put a real job through it. Reading every one of roughly 230 San Francisco neighbourhood pages once a day for a month is about 6,900 requests and 4.3 GB, or $3.46. Twenty thousand neighbourhood pages, which is a national sweep of larger cities, move 12.5 GB for about $10. The rendered route costs 55% more per page and returns less, so the saving is not in the proxy line but in the browser you do not run. If you are sizing a pool before you buy, our note on how much proxy data you need does the same arithmetic in the other direction.
The setting that works on Nextdoor
- Network: residential proxies, Basic line, $0.80/GB, rotating. Nineteen fetches from three countries met no address-level obstacle on the public pages; the pool is for throughput, not for trust. Mobile at $2.30/GB and ISP at $2.50/IP per month buy nothing we could measure here.
- Fetch mode:
engine: tls, plain HTTP, never rendered. The city page returns 2,404 words in 676,650 bytes; the rendered page returns 2,082 words in 1,046,614 bytes. - Country: none required. The same URL returned 676,650, 676,641 and 676,643 bytes and the same 2,404 words from the United States, Canada and the United Kingdom, with the same
en-uslanguage header. Spread the pool for throughput; pin a country only if you are checking how a Nextdoor ad renders in that market. - When the proxy is not enough: there is no Nextdoor collector in our catalogue, and the public pages do not need one. The web scraping API without rendering returns each city or neighbourhood page as clean Markdown or JSON from $0.0002 per page, retries and rotation included, and failed requests are never billed. For posts by keyword and radius, marketplace listings and business pages, the route is Nextdoor’s own Display API, by application. For the business directory itself, Google Maps places returns the same local businesses with address, phone and category per delivered result, without the address check.
- Free tier: every account gets $2 of free API usage per month, which is 10,000 neighbourhood pages through the scraping API before you pay anything.
Read the city and neighbourhood pages over plain HTTP, take the demographics, groups, listings and events from the HTML, hand keyword monitoring to the Display API, and stay out of the feed behind the address verification: it is not public, and no proxy makes it so. That is the whole configuration. For the two platforms in this series where a public page also carried everything a logged-out client needs, see Facebook, where the Marketplace behaved the same way, and Reddit.