What success rate does a residential proxy reach on real websites? On 18 November 2026 we measured QuanticData rotating residential exits on 20 well-known public sites that returned the full page. By plain HTTP request 18 sites did so on 440 of 450 requests on Basic (97.8%) and 431 of 450 on Premium (95.8%); with the right setup, a real browser on a sticky session, Yahoo and Indeed returned it too.
| How the full page came back | Sites | Which |
|---|---|---|
| Plain HTTP request, browser-like TLS client | 18 | BBC News, Bing, Kayak, Kleinanzeigen, LinkedIn, The Guardian, Zillow, Airbnb, Walmart, YouTube, AliExpress, Best Buy, Gumtree, Pinterest, The New York Times, TikTok, Craigslist and CNN |
| With the right setup: a real browser on a sticky session | 2 | Yahoo and Indeed |
Measured by us on 18 November 2026, 15:31 to 15:38 UTC for the HTTP requests and 15:54 to 16:11 UTC for the browser loads, one run. Basic numbers come from a Residential Basic plan and Premium numbers from a Residential Premium plan of the same account, never mixed. A request counts only when it returned HTTP 2xx and the content a visitor would read (a product title, listing cards, search results, the article list). The sites listed here returned the full page in our test; they are not every site there is, and your own targets are worth a test of their own. Per-site figures: residential-success-by-website-2026-11-18.csv.
Success rate by website, Basic and Premium
Residential BasicResidential Premium
Over these 18 sites the median site returned the full page on 100% of requests on Basic and 100% on Premium. Every site in the chart stayed at 80% or more on both lines, most at or near 100%, with one request at a time per site, a new household address on every request and no retries.
All 18 websites, one line each
Full pages out of requests sent, per line, and the median time to first byte of those requests, in seconds. Client: curl_cffi with the Chrome 152 TLS and header profile, 25 requests per site per line.
| Website | Full page, Basic | Full page, Premium | Median TTFB, Basic | Median TTFB, Premium |
|---|---|---|---|---|
| BBC News (News, GB exit) | 25/25 | 25/25 | 0.49 s | 0.52 s |
| Bing Search (Search, US exit) | 25/25 | 25/25 | 1.04 s | 1.13 s |
| Kayak route page (Travel, US exit) | 25/25 | 25/25 | 1.11 s | 1.08 s |
| Kleinanzeigen search (Classifieds, DE exit) | 25/25 | 25/25 | 0.80 s | 0.94 s |
| LinkedIn company page (Social, US exit) | 25/25 | 25/25 | 2.19 s | 2.20 s |
| The Guardian (News, GB exit) | 25/25 | 25/25 | 0.52 s | 0.53 s |
| Zillow city page (Classifieds, US exit) | 25/25 | 25/25 | 1.65 s | 1.51 s |
| Airbnb search (Travel, US exit) | 25/25 | 24/25 | 1.29 s | 1.12 s |
| Walmart search (E-commerce, US exit) | 24/25 | 25/25 | 1.08 s | 1.11 s |
| YouTube channel page (Social, US exit) | 24/25 | 25/25 | 1.14 s | 1.18 s |
| AliExpress search (E-commerce, US exit) | 25/25 | 23/25 | 2.02 s | 2.17 s |
| Best Buy search (E-commerce, US exit) | 25/25 | 23/25 | 1.28 s | 1.63 s |
| Gumtree search (Classifieds, GB exit) | 25/25 | 23/25 | 1.02 s | 1.19 s |
| Pinterest profile page (Social, US exit) | 24/25 | 23/25 | 1.42 s | 1.35 s |
| The New York Times (News, US exit) | 22/25 | 25/25 | 1.15 s | 1.26 s |
| TikTok profile page (Social, US exit) | 23/25 | 24/25 | 2.24 s | 2.36 s |
| Craigslist search (Classifieds, US exit) | 24/25 | 21/25 | 2.25 s | 2.21 s |
| CNN (News, US exit) | 24/25 | 20/25 | 0.88 s | 1.02 s |
Sites that return the full page with the right setup
Two search and job sites build their results in the browser, so a plain HTTP request is not the right tool for them. Loaded in headless Microsoft Edge through a sticky residential session (one household address for the whole page load), they returned the full page on 16 of 16 loads (Yahoo) and 9 of 10 loads (Indeed).
| Website | Full page, Basic | Full page, Premium |
|---|---|---|
| Yahoo Search (Search, US exit) | 8/8 | 8/8 |
| Indeed job search (Classifieds, US exit) | 4/5 | 5/5 |
That is the setup to copy for any site that renders its content in the browser: your own headless browser on a sticky session, as in our test of 53 sites with and without a headless browser, or the Web Unlocker, which runs the browser and the retries for you.
What the residential exit adds
Two things a plain connection does not have: the country the site expects, and a new address on every request. United States exits served the United States sites, the United Kingdom the BBC, The Guardian and Gumtree, Germany Kleinanzeigen; rotation spread the volume so that no site counted it against one IP. TikTok is the clearest case: from our test machine in Italy, without a proxy, it answered with a login page, and through United States residential exits it served the public profile on 47 of 50 requests.
Is Residential Premium better than Basic?
Both lines did the job on these sites. Basic returned the full page on 440 of 450 requests (97.8%) with a median first byte of 1.25 s; Premium on 431 of 450 (95.8%) at 1.27 s. Buy Basic for volume on public pages like these; choose Premium when you need its city and ISP targeting or its pool filters. Both are on residential proxy pricing.
How we measured
- Targets: public pages in six groups (search, e-commerce, travel, classifieds and listings, news, social landing pages), logged out, no forms, no personal data. The URL of each is in the CSV.
- Exits: rotating, a new household address per request, United States for most sites, United Kingdom for the BBC, The Guardian and Gumtree, Germany for Kleinanzeigen. No pool filter on Premium.
- HTTP client: curl_cffi 0.17.0 with the Chrome 152 TLS and header profile, 25 requests per site per line. No JavaScript, no cookies kept between requests, redirects followed, 25-second timeout, no retries.
- Browser: headless Microsoft Edge with a normal user agent, a new sticky session per load, images, media and fonts not downloaded, 5 to 8 loads per line (25 of 26 loads returned the full page).
- Pacing: one request at a time per site with a one-second pause, the two lines alternated within each round, the sites in parallel.
- Full page: HTTP 2xx and the expected content present; the content check for each site is in the CSV column content_check.
- Times: libcurl's start-transfer time (gateway, tunnel to the exit, TLS with the site, server reply). Machine: Windows, home broadband in Italy.
What this test does not show
- One run on one afternoon. Sites change their rules without notice; test your own targets before you plan a crawl.
- Logged-in pages, checkout, search at high volume or any page behind a login: not tested, on purpose.
- Volume. 25 requests per site per line is a sample, not a crawl.
- Other providers. We measured our own two lines and compare nothing else.
- Our content checks are simple text markers. A page that passed them could still lack a field you need.
What to use, site by site
For the 18 sites that returned the full page to a plain request, rotating residential with a browser-like HTTP client is enough: United States residential proxies for the United States sites, the country of the site for the others, every residential country on the same account. Start on the free residential trial, then buy residential proxies on Basic for volume. For sites that build their content in the browser, run a headless browser through a sticky residential session, or hand the page to the Web Unlocker or the Web Scraper API, billed only when the page comes back. Speed per exit country is measured separately in residential proxy speed in 15 countries, and the product page for both lines is QuanticData residential proxies.


