The best proxies for web scraping at scale come from Oxylabs and Decodo. Oxylabs' rotating residential pool reached the most distinct IP addresses in Proxyway's April 2026 test, and Decodo came close while selling residential traffic at $4 per GB pay-as-you-go as of October 2026. Byteful had the best residential success rate on real targets in that research, SOAX had the best US success rate in the same round, Webshare suits jobs that mix datacenter and residential IPs, and DataImpulse and Evomi are the budget picks.
A web scraping proxy routes your scraper's requests through another IP address, so the target site sees the proxy instead of your server, as our guide to the role of proxy servers in a network explains. Getting an IP is the easy part; getting pages is not. In Proxyway's 2026 research, the median residential provider succeeded on 99.28% of requests to a test server but on only 74.43% of requests to Amazon, Google and Instagram. That gap, plus what you pay for the failed share, decides your real cost.
Every price below was read from the vendor's own pages on October 3, 2026, and every test figure comes from Proxyway's published tests with its test month; we did not run benchmarks of our own. Nothing here is sponsored, and there are no affiliate links. If you only need free proxies for a small one-off job, our guide to free proxy options for scraping covers that; this page is about paid networks you would trust with production traffic.
Quick Answer: Best Proxies for Web Scraping
How Did We Compare These Web Scraping Proxy Providers?
We compared each paid scraping proxy service on what decides your bill and your success rate, using published sources. We started from the providers named in the review lists ranking on Google in October 2026. We dropped any without a 2026 third-party residential test or with a published success rate below 99%, and ordered the rest by how many distinct IPs that test reached. These were the criteria:
- → Published test results: success rate, response time and unique IPs from Proxyway's 2026 residential tests, with the test month.
- → Results on real targets: Amazon, Google and Instagram, where Proxyway published a figure.
- → Real entry cost: the smallest purchase you can make and the price at around 100 GB, not the lowest "from" price.
- → Billing rules: whether blocked responses count as traffic, when unused traffic expires, and what targeting adds to the price.
- → Limits: concurrent connections, the longest sticky session, and how finely you can target (city, ZIP code, ASN).
- → KYC, restricted targets and refunds: the terms that decide whether you can test cheaply and scrape the sites you need.
- → Sourcing: what each vendor publishes about how its residential IPs are obtained.
Prices, limits and policies come from each vendor's pricing pages, documentation and help center, read on October 3, 2026, in US dollars before tax. Pool sizes and country counts are vendor claims, and where a vendor's own pages contradict each other, we say so.
Test figures come from Proxyway. Its residential test sends 3.6 million requests over 21 days through each provider's global pool to a CDN server and records success, response time and unique IPs. A separate target test ran a US scraper at one request per second against Amazon, Google and Instagram. The data come from its 2026 proxy market research, published June 30, 2026, and from its provider reviews, which each state the test month.
Proxyway discloses that it uses affiliate links, and its own scraping ranking puts Decodo first. We found no comparable 2026 benchmark without a commercial tie: as of October 2026, AIMultiple lists Bright Data, Oxylabs, Decodo and Webshare among paying subscribers to its benchmarking, and ScrapeOps sells a proxy aggregator.
We order by unique IPs because all seven cleared 99% on Proxyway's infrastructure tests, within 0.65 percentage points of each other, and Proxyway publishes quotable real-target figures only for the best provider and the median. Unique IPs show how much of an advertised pool a test actually reached, which matters when you spread a large job across a pool. Evomi's figures come from Proxyway's July 2026 test of its Core pool, the product priced here; its Premium pool, tested in April 2026 with the others, reached 1,668,674 distinct IPs at 99.58%.
What we did not do
The next table puts all seven side by side, with dated prices.
Best Proxies for Web Scraping Compared: Which Provider Fits Which Job?
There is no single best web scraping proxy for every job, so read the table by job, not by row number: the order follows the ranking rule above, while the Best For column says who should buy each one. Unique IPs and success rates are Proxyway's residential figures for 3.6 million requests, with the test month; prices are entry options as of October 2026. Ownership note: Oxylabs has owned Webshare since 2022, and we judged each on its own evidence.
| Provider | Best For | Unique IPs and Success (Test Month) | Entry Price (Oct 2026) | Trial and Refund |
|---|---|---|---|---|
| Oxylabs | Large-scale scraping | 3,241,404 IPs, 99.93% (April 2026) | 5 GB for $30 ($6/GB); no pay-as-you-go listed | Trial for companies on request; 3-day refund |
| Decodo | Best value | 2,853,543 IPs, 99.92% (April 2026) | $4/GB pay-as-you-go; 3 GB for $11.25 | 3-day, 100 MB trial; 14-day refund |
| Byteful | Best success rate on real targets | 1,909,605 IPs, 99.69% (April 2026); best real-target result, 81.23% | 10 GB for $32.50 ($3.25/GB); 1 GB orders allowed, price not listed; never expires | Its pages disagree on a trial; 24-hour refund |
| Webshare | Datacenter and residential mix | 1,745,672 IPs, 99.28% (April 2026) | Residential 1 GB for $3.50 a month; 100 datacenter proxies for $2.99 | Free datacenter plan; 48-hour refund |
| SOAX | US-focused scraping | 1,670,306 IPs, 99.78% (April 2026); best US success in that round, 99.92% | No-fee Sandbox, US traffic at $5/GB; Builder $200 a month | Trial unclear; no refunds |
| DataImpulse | Budget pay-as-you-go | 1,538,768 IPs, 99.74% (April 2026) | 5 GB for $5 ($1/GB), never expires | No trial listed; 7-day refund on the first Intro plan only |
| Evomi | Cheapest per GB at volume | 1,499,867 IPs, 99.91% (Core pool, July 2026) | 100 GB for $49 a month ($0.49/GB); $0.99/GB pay-as-you-go from 15 GB | Trial turns paid unless canceled; 30-day refund if under 10% or 1 GB used |
The table tells you who to shortlist. Which type of proxy you buy from them depends on the sites you scrape.
Which Proxy Type Does Each Scraping Job Need?
Choosing a proxy server for web scraping starts with how hard the target defends itself. Datacenter IPs are the cheapest and fastest, and they work on public pages that do not filter hosting ranges. Rotating residential IPs are the default for retailers, marketplaces and other protected sites. ISP proxies suit multi-step sessions that need one stable address, and mobile IPs are a last resort for the hardest targets.
| Scraping Job | Proxy Type | Why | Usual Billing |
|---|---|---|---|
| Public pages, documentation, sites without bot defenses | Datacenter (rotating or static) | Fastest and cheapest; fails where hosting ranges are filtered | Per IP a month, or per GB |
| Retailers, marketplaces, travel and other protected sites | Rotating residential | IPs belong to home connections, not hosting networks | Per GB |
| Multi-step flows on your own accounts, long sessions | ISP (static residential) | One stable address with a home-ISP label | Per IP a month |
| The hardest targets, mobile-first sites | Mobile (4G/5G) | Carrier IPs are shared by many real users | Per GB |
Datacenter proxies for web scraping are where to start if your targets allow them: Proxyway's April 2026 tests put the datacenter pools of Oxylabs, Decodo, Webshare and Rayobyte between 99.88% and 99.98% success on its test server. Real sites are another matter. In Proxyway's older March 2024 test, Bright Data's datacenter IPs succeeded on 85.14% of Home Depot requests but only 5.83% of Amazon requests, because many large sites filter hosting IP ranges; our site-owner guide to blocking cloud and hosting traffic shows how that works. If your targets accept them, start there: datacenter plans cost a fraction of residential traffic.
Residential proxies for web scraping are what most of this guide ranks, because they are what most protected targets require. Rotating proxies for web scraping switch to a new IP on every request, which suits catalog and search-result crawls. Most residential pools can also hold one IP for a set time, called a sticky session, for pagination or anything that needs the same address across several requests.
ISP proxies for scraping make sense when you need one address for days, such as monitoring your own seller account or a checkout flow you are allowed to test; they cost more per IP but keep the same home-ISP address for as long as you pay. Private proxies for data scraping, usually sold as datacenter or ISP IPs, cost more per address than shared lists because few or no other customers use them. Ask how many customers share each IP before you pay a private-tier price; Webshare, for example, says its private proxies are shared with up to two other users.
Mobile proxies cost more and helped less than you might expect: in Proxyway's 2026 target test, the median mobile pool succeeded on 75.26% of requests, against 74.43% for residential, although the best mobile pool (DataImpulse, 85.31%) beat the best residential one (Byteful, 81.23%).
If your success rate stays low even on residential IPs, check your request rate and the site's terms first, because the IP is often not the problem. Several of these vendors also sell unblockers or scraper APIs that bill only for successful requests. As of October 2026, Decodo's Site Unblocker starts at $0.95 per 1,000 requests and Bright Data's Web Unlocker costs $1.50 per 1,000.
Page weight decides which model is cheaper. At $4 per GB, a 375 KB page costs the same $0.0015 as one successful request at $1.50 per 1,000, so heavy pages and low success rates favor per-request billing. Comparing those products is a separate buying decision that this guide does not rank.
The 7 Best Proxy Services for Web Scraping, Reviewed
These are the best proxies for web scraping we could document, in this order: Oxylabs, Decodo, Byteful, Webshare, SOAX, DataImpulse and Evomi, ranked by the number of distinct IPs Proxyway's 2026 residential test reached. Each review is labeled with the job the provider does best and covers what it sells, where it wins and falls short for scraping, and pricing as of October 2026.
1. Oxylabs: Best for Large-Scale Scraping
Oxylabs sells residential, mobile, datacenter and ISP proxies, plus scraper APIs. Its rotating residential pool can target by continent, country, state, city, ZIP code, coordinates and ASN; sticky sessions last 10 minutes by default and up to 24 hours; concurrent sessions are unlimited; and connections run over HTTP(S), HTTP/3 or SOCKS5. It advertises 175M+ residential IPs in 195 locations.
Where it wins
- The most distinct IPs in Proxyway's April 2026 test: 3,241,404 in 3.6 million requests, with 99.93% success and a 0.60 s average response
- The deepest targeting here, down to ZIP code, coordinates and ASN, with geo-targeting included at no extra charge
- 24/7 support and a dedicated account manager on every residential plan, including the $30 Starter
Where it falls short
- No pay-as-you-go price is published; the smallest plan works out to $6 per GB
- Every customer fills in a KYC form
- Residential IPs cannot reach "Some Google domains", Apple, entertainment, banking, government, streaming, ticketing and mailing sites, and its docs also list gaming, ads and IP-checking sites; some open after a short KYC
- Refunds only within 3 days of the first purchase and below 20% of traffic used
Pricing as of October 2026: Starter 5 GB for $30 ($6/GB), Basic 20 GB for $100 ($5/GB), Advanced 125 GB for $500 ($4/GB) and Corporate 1 TB for $2,500 ($2.50/GB), with VAT where it applies. Its residential product page says plans start at $99 for 11 GB, which does not match the pricing-page plans listed here.
Verdict: the strongest choice when you need a very large spread of IPs, fine targeting and support at any hour, and can live with its KYC and its restricted-target list. Skip it if your targets include the Google domains it restricts or you want to avoid KYC.
2. Decodo: Best Value for Most Scraping Projects
Decodo, the name Smartproxy has used since April 22, 2025, sells rotating residential traffic by the gigabyte with continent, country, state, city, ZIP code (US only) and ASN targeting. Its docs allow sticky sessions from 1 to 1,440 minutes, and connections run over HTTP(S) or SOCKS5. Its site lists 115M+ IPs in 195+ locations.
Where it wins
- 2,853,543 distinct IPs, 99.92% success and a 0.53 s average in Proxyway's April 2026 test, plus the most distinct US IPs of any provider: 961,218 in 1.2 million US requests
- Pay-as-you-go at $4 per GB, bought 1 GB at a time and usable for 12 months or until you buy a monthly subscription
- A 14-day money-back window on a first purchase used below 20%
- The clearest billing rules of the seven, and quick-start code in six languages
Where it falls short
- Target responses such as 403, 404 and 503 count toward paid traffic, and so does the TLS handshake
- Unused plan traffic is lost at renewal unless you upgrade
- The 3-day trial needs a card, becomes a paid plan unless you cancel, and removes the refund right
- Mail, streaming and business sites need ID verification first
- Its pricing page says sticky sessions last up to 30 minutes; its docs say 24 hours
Pricing as of October 2026 (plus VAT): pay-as-you-go $4/GB; monthly plans from 3 GB for $11.25 ($3.75/GB) to 100 GB for $275 ($2.75/GB); enterprise rates from $2.50/GB at 250 GB down to $2/GB at 1 TB.
Verdict: the best balance of measured results, flexible entry and transparent billing for most teams. Budget for blocked pages: because Decodo bills them, a low success rate on a hard target raises your cost per page. Skip it if you need ZIP targeting outside the US.
3. Byteful: Best Success Rate on Real Targets
Byteful, which traded as Ping Proxies until an April 2026 rebrand, sells residential bandwidth that never expires, with country, region, city, ISP and ASN targeting (it lists more than 10,000 networks and ASNs), sticky sessions of up to 24 hours and unlimited concurrent sessions over HTTP or SOCKS5. It advertises 35M+ residential IPs in 195+ countries.
Where it wins
- The best residential success rate on real targets in Proxyway's 2026 research: 81.23% across Amazon, Google and Instagram, against a 74.43% median
- The fastest residential pool in that research, at 0.41 s, with 1,909,605 distinct IPs and 99.69% success (April 2026)
- Bandwidth never expires, even if you cancel or pause
Where it falls short
- Refunds only within 24 hours and below 2 GB or 25% of the allocation, whichever is lower
- No IP whitelisting on residential (IP authentication is for static proxies only), and Proxyway says its sourcing is described "in rather vague terms"
- Its pricing page says there is no free trial, while its residential page advertises a 1 GB trial that its trial page ties to KYC
- Its support hours are listed inconsistently (European weekday daytime in its FAQ, 5 AM to 7 PM EST on its support page), with slower 24/7 cover outside them
- Its blocklist includes select e-commerce platforms, decided case by case, and the full list is not published
Pricing as of October 2026: $3.25 per GB on small orders (10 GB for $32.50), $3 at 25 GB and $2.75 at 100 GB ($275); its homepage advertises residential traffic from $1.75 per GB without naming the volume.
Verdict: the pick when your targets are well defended and success rate matters more than the last cent per gigabyte, provided your sites are not on its restricted list. Test inside the first day, because the refund window closes after 24 hours, and skip it if you need IP whitelisting on residential.
4. Webshare: Best for Mixing Datacenter and Residential IPs
Webshare, owned by Oxylabs since September 2022, is best known for cheap datacenter proxies and also sells rotating residential bandwidth on monthly plans. Its FAQ lists country, state, city, ZIP and ASN targeting for residential, connections run over HTTP or SOCKS5, and you can log in with a username and password or an authorized IP. It claims 80M+ residential IPs.
Where it wins
- Cheap entry for easy targets: 100 shared datacenter proxies for $2.99 a month with 250 GB, and its datacenter IPs scored 99.93% in Proxyway's April 2026 test
- An ongoing free datacenter plan (10 proxies, 1 GB a month) for testing your scraper
- Residential from 1 GB for $3.50 a month, with 1,745,672 distinct IPs and 99.28% success in April 2026
- No KYC unless suspicious activity is detected
Where it falls short
- No pay-as-you-go: residential bandwidth is monthly and does not roll over
- Refunds within 48 hours and below 0.1 GB of residential use
- Paid plans allow 500 concurrent connections; above that you get 429 errors unless you buy an add-on
- No published sticky-session length, and Proxyway notes "limited location and sessions settings" on residential
- The lowest infrastructure success rate of the seven (99.28%, April 2026), and human agents work 24/5 behind a 24/7 chatbot, according to Proxyway
Pricing as of October 2026: residential 1 GB for $3.50, 10 GB for $27.50, 100 GB for $225 ($2.25/GB) and 1,000 GB for $1,500 ($1.50/GB) a month, shown as discounts from higher list prices; shared datacenter from 100 proxies for $2.99 a month.
Verdict: a sensible first stop when most of your targets accept datacenter IPs and only some need residential ones. The 500-connection default is the first limit a large crawler will hit, and banking, government, streaming, ticketing and mailing sites are restricted on its residential network. Skip it if you need pay-as-you-go residential traffic.
5. SOAX: Best for US-Focused Scraping
SOAX sells residential and mobile traffic from one prepaid credit balance, priced by the country tier of the exit IP; US traffic sits in the most expensive tier. Every plan includes country, region, city and ISP targeting, and sessions can rotate per request, stay sticky for up to 60 minutes, or rotate on a timer from 10 seconds to a week. It advertises 155M+ IPs.
Where it wins
- The best US success rate in Proxyway's April 2026 research round, 99.92%, and 99.78% globally with 1,670,306 distinct IPs (April 2026)
- City and ISP targeting on every plan, including the no-fee Sandbox
- Per-request, sticky and timed rotation in one product
Where it falls short
- Payments are non-refundable, and its KYC policy (effective June 1, 2026) requires identity verification before any purchase, although its help center lists ID checks only for crypto payments and restricted sites or ports
- No classic pay-as-you-go: Sandbox top-ups buy US traffic at $5 per GB and expire after 60 days
- ASN and ZIP targeting start on the $1,500-a-month Scale plan
- Only ports 80 and 443 unless you pass ID verification and request others, and 4,000 connections per gateway on Sandbox
- No exit IPs in Texas, which SOAX attributes to the state's regulations
- Its product pages advertise a three-day, 400 MB trial for $1.99 that its pricing page does not list
- Paid plans carry a "24/7" support tier, but its SLA promises round-the-clock response only for critical (P0) issues, and Proxyway says its support "has been nerfed hard"
Pricing as of October 2026 (plus VAT): Sandbox has no fee, with top-ups at $5/GB for US traffic, $4 for tier 2 and $2 for tier 3 countries; Builder costs $200 a month at $3/GB for the US, Team $500 at $2.20 and Scale $1,500 at $1.50.
Verdict: worth it for steady US workloads on a monthly budget. With no refunds, a small Sandbox top-up, after identity verification, is the lowest-risk way to test it. If you need a proxy service for web scraping in the US, also compare Decodo, which returned the most distinct US IPs in Proxyway's 2026 research: 961,218 in 1.2 million requests. Skip SOAX if you need refunds, Texas exits or ASN targeting below $1,500 a month.
6. DataImpulse: Best Budget Pay-as-You-Go
DataImpulse sells residential traffic at $1 per GB that never expires. Country targeting is included; state, city, ZIP and ASN targeting is billed at double the rate. Sticky sessions last up to 120 minutes, each plan allows 2,000 threads, and connections run over HTTP(S) or SOCKS5. It also sells mobile traffic at $2 per GB.
Where it wins
- A cheap pay-as-you-go start: 5 GB for $5 ($1 per GB), and the traffic never expires
- 99.74% success with 1,538,768 distinct IPs in Proxyway's April 2026 test, and the fastest residential responses on real targets (2.58 s)
- Its mobile pool had the best real-target success in Proxyway's 2026 research: 85.31% across Amazon, Google and Instagram
Where it falls short
- Proxyway found that "Nearly 15% of the US proxies came from servers", not homes
- State, city, ZIP and ASN targeting doubles the per-GB price
- A 2,000-thread cap per plan; raising it requires KYC
- The 7-day refund covers only a first Intro plan and never crypto payments
- On October 3, 2026, Trustpilot showed its rating as unavailable "due to a breach of our guidelines"
Pricing as of October 2026: residential at $1/GB (Intro 5 GB for $5, Basic 50 GB for $50) and $0.80/GB on the 1 TB Advanced plan ($800); Premium Residential $5/GB; mobile $2/GB; datacenter $0.50/GB.
Verdict: the cheapest way to start and good value for country-level crawls. Check a sample of the US IPs it gives you before you rely on them as residential, and skip it if you need city or ZIP targeting at the $1 rate.
7. Evomi: Cheapest per GB at Volume
Evomi, a Swiss provider, sells Core residential traffic on monthly plans or pay-as-you-go, plus a Premium pool for targets the Core pool cannot reach. Targeting covers country, region and city at no extra cost, with ZIP code, ASN and ISP available through "expert settings" that raise GB use; sticky sessions run 30 minutes by default and up to 24 hours, and it sets no concurrency limit. It lists 54M+ residential IPs in 150+ countries.
Where it wins
- The lowest per-GB price here: 100 GB for $49 a month ($0.49/GB), or $0.99/GB pay-as-you-go from 15 GB with no expiry
- 99.91% success and a 0.70 s average in Proxyway's July 2026 test of its Core pool, with 99.98% success and a 0.30 s average in the US
- No limits on sessions or concurrency
Where it falls short
- The smallest measured pool on the list: Proxyway's July 2026 test of the Core pool reached 1,499,867 distinct IPs globally and 75,253 in the US
- ZIP, ASN and other expert filters multiply GB use, up to 15 times
- Trials turn into paid plans unless you cancel, and the Core residential trial lasts only one day
- Subscription GB does not roll over, and the 30-day refund applies only to a first purchase with less than 10% of the allowance or 1 GB used, whichever is smaller
- No IP whitelisting or sub-users (Proxyway), and on October 3, 2026, Trustpilot showed its rating as unavailable for a guidelines breach
Pricing as of October 2026: Core residential 100 GB for $49 a month with extra GB at $0.49, or pay-as-you-go at $0.99/GB from 15 GB ($0.79 from 1,000 GB); Premium residential $3.98/GB pay-as-you-go.
Verdict: the cheapest gigabytes here for broad, country-level crawls. Skip it if your crawl depends on ZIP or ASN filters, because the multipliers erase the price advantage.
Several more providers came up in our research. Prices, test figures and other details for them are as of October 2026.
Also considered, and why they are not above
- IPRoyal: 96.11% success in Proxyway's April 2026 residential test, below our 99% bar, and $7.35/GB for 1 GB pay-as-you-go, though its traffic never expires and sticky sessions can run up to 7 days. Trustpilot showed its rating as unavailable for a guidelines breach.
- Rayobyte: 98.02% residential success (April 2026) and the cheapest rotating datacenter traffic we found at $0.30/GB; residential costs $3.50/GB for 1 to 49 GB, but it says refunds are generally not provided for bandwidth products.
- Infatica: 97.57% success in Proxyway's April 2026 residential test, $4/GB pay-as-you-go, and mandatory ID-and-selfie KYC for every customer.
- Bright Data: not tested by Proxyway since March 2024. Residential proxies are for verified companies only, it blocks or bypasses requests to google.com, youtube.com and bing.com on its regular proxies and points users to its SERP API or Web Unlocker API, the list price is $8/GB pay-as-you-go, and its contract calls fees non-refundable.
- Massive: the most recent Proxyway test we found, as of October 2026, dates from April 2025; monthly plans run from $80 for 10 GB to $490 for 100 GB, unused GB rolls over, and sticky sessions last up to 12 minutes.
- ProxyEmpire and Proxyrack: both took part in Proxyway's April 2026 residential test, at 98.17% and 96.73% success, below our 99% bar.
- NetNut: excluded. The FBI seized hundreds of NetNut domains in July 2026 in an action tied to the Popa botnet, as KrebsOnSecurity reported.
What Do Web Scraping Proxies Really Cost?
Scraper proxies are billed per gigabyte (residential, mobile and rotating datacenter), per IP per month (static datacenter and ISP), or per successful request (unblockers and scraper APIs). As of October 2026, entry-level residential prices among the seven range from $0.49 per GB on Evomi's 100 GB plan to $6 per GB on Oxylabs' 5 GB plan. The number that matters, though, is the cost per page you keep, which depends on page weight, success rate and what the provider counts as billable traffic.
By early 2026 the price cuts had stalled. Proxyway's 2026 market research says residential rates "contracted by up to 75%" between 2023 and 2025, and that "In early 2026, it looks like the situation with residential proxies has stabilized and even started to revert."
Here is the math with round numbers. Say each product page transfers 1 MB through the proxy, headers included; that is an assumption for the example, so measure your own pages first.
-x http://USER:PASS@proxy.example.com:8000 \
-w "body %{size_download} B, headers %{size_header} B\n" \
https://www.example.com/product/123
# TLS handshake bytes are not included; some providers bill them
If your scraper drives a headless browser, measure in the browser instead, because images, scripts and fonts also pass through the proxy unless you block them.
Ten thousand pages are then about 10 GB, or $40 at Decodo's $4-per-GB pay-as-you-go rate if every request succeeds. At 74.43% success, the median Proxyway measured on Amazon, Google and Instagram, you need about 13,435 requests for 10,000 good pages. Decodo counts blocked responses as traffic, so if a blocked response weighed as much as a real page, the same 10,000 pages would cost about $53.74. Blocked responses are usually smaller, so the real figure falls between the two.
| Provider | Blocked Responses Billed? (Oct 2026) | Unused Traffic | Targeting Surcharge | Refund |
|---|---|---|---|---|
| Oxylabs | Not published | Not published | None; geo-targeting included | 3 days, below 20% used |
| Decodo | Yes: 403, 404, 503 and other non-200 responses, plus the TLS handshake | Pay-as-you-go lasts 12 months or until you subscribe; plan GB lost at renewal | None listed | 14 days, first purchase, below 20% used |
| Byteful | Not published | Never expires | None listed | 24 hours, below 2 GB or 25% used |
| Webshare | Not published | Monthly; no rollover | None listed | 48 hours, below 0.1 GB used |
| SOAX | Routed requests count, failed connections do not; blocked responses not addressed | Credits expire after 60 days | ASN and ZIP need the Scale plan | None |
| DataImpulse | Not published | Never expires | 2x for state, city, ZIP and ASN | 7 days, first Intro plan only |
| Evomi | Not published | Pay-as-you-go never expires; plans do not roll over | ZIP, ASN and other expert filters, up to 15x | 30 days, first purchase, below 10% or 1 GB used |
"Not published" means we could not find the rule on the vendor's own pages on October 3, 2026. Ask before you buy, because on a hard target the blocked share can be a quarter of your traffic or more.
Which Scraping Proxies Are Cheapest?
If you want cheap proxies for scraping, separate the cheapest gigabyte from the cheapest test. As of October 2026, the lowest residential rate here is Evomi's 100 GB plan at $0.49/GB; on pay-as-you-go, Evomi charges $0.99/GB from 15 GB and DataImpulse $1/GB from 5 GB, falling to $0.79 and $0.80 per GB at 1 TB. For datacenter traffic, Rayobyte's rotating pool at $0.30/GB and Webshare's 100 shared proxies for $2.99 a month are the cheapest entries we found. City- or ZIP-level targeting raises both budget picks: DataImpulse doubles its rate, and Evomi's expert filters multiply usage by up to 15 times.
The smallest real tests cost $4 for 1 GB at Decodo and $5 for 5 GB at DataImpulse; Byteful accepts 1 GB orders without listing the price, and Webshare's free plan covers datacenter tests. A useful test can use more traffic than the refund rules allow, since Webshare refunds only below 0.1 GB of residential use and Decodo below 20%, so treat the first purchase as spent. The limits in the next section can matter more than the price.
What Limits Do Scraping Proxy Providers Set?
The limits that stop a scraper are rarely in the headline: concurrency caps, sticky-session length, targeting depth, KYC and the list of sites a provider will not let you reach. As of October 2026, Webshare caps paid plans at 500 concurrent connections by default, DataImpulse allows 2,000 threads per plan, and Oxylabs, Byteful and Evomi set no concurrency cap. Those caps are generous for most jobs: concurrent connections roughly equal requests per second times response time, so 10 pages a second at the 3.21-second median Proxyway measured on real targets needs about 32.
| Provider | Concurrency (Oct 2026) | Longest Sticky Session | Targeting Depth |
|---|---|---|---|
| Oxylabs | Unlimited concurrent sessions | 24 hours | Continent to ZIP code, coordinates, ASN |
| Decodo | Not published | 24 hours in docs; 30 minutes on pricing page | Country to ZIP (US only); ASN cannot be combined with continent, state, city or ZIP |
| Byteful | Unlimited concurrent sessions | 24 hours | Country, region, city, ISP, ASN |
| Webshare | 500 by default; more as a paid add-on | Not published | Country, state, city, ZIP, ASN |
| SOAX | 4,000 per gateway on Sandbox; requests per second vary by package | 60 minutes; timed rotation up to a week | Country, region, city, ISP; ASN and ZIP from Scale |
| DataImpulse | 2,000 threads per plan | 120 minutes | Country included; state, city, ZIP, ASN at 2x |
| Evomi | No limit | 24 hours | Country, region, city; ZIP, ASN and ISP via expert settings |
A long sticky session is a maximum, not a promise: Oxylabs says a session may end sooner because its residential pool is dynamic, and Decodo says the IP can rotate early when the device behind it goes offline. Check authentication too: a username and password work from any server, while IP whitelisting breaks when your scraper's own address changes, and Byteful's docs and Proxyway's Evomi review say neither offers it on residential plans. KYC and restricted sites matter even more, because they decide whether you can scrape your targets at all.
| Provider | KYC (Oct 2026) | Restricted on Residential by Default |
|---|---|---|
| Oxylabs | KYC form for every customer | Apple, entertainment, banking, some Google domains, government, gaming, streaming, ticketing, mailing, ads, IP-checking sites; some open after a short KYC |
| Decodo | Automated checks at sign-up; further ID checks when needed, including to open restricted categories | Banking and crypto, government, entertainment, app stores, ticketing, gaming, mailing, streaming, business, telecom; Google is reachable |
| Byteful | Risk-based; not every customer | Government, banking, payment processors, mail, adult and select e-commerce sites, plus direct-IP requests since September 7, 2026; the full list is not published |
| Webshare | Only if suspicious activity is detected | Entertainment, banking, government, streaming, ticketing, mailing; some open after KYC |
| SOAX | KYC policy requires ID before purchase; help center lists ID checks for crypto and restricted sites | Financial services, government, certain news sites; any port other than 80 and 443 |
| DataImpulse | When use is unclear, and to raise the thread cap | Government, banking and payment sites, adult content and harmful domains, plus specific domains listed in its docs |
| Evomi | Not by default | Banks, government websites "and more" |
If search results are part of your job, read these lists closely. Oxylabs restricts "Some Google domains" on residential proxies, while Decodo says Google is reachable on rotating residential. Outside our seven, Bright Data blocks or bypasses requests to google.com, youtube.com and bing.com on its regular proxies and points users to its SERP API or Web Unlocker API instead.
Support hours matter when a large crawl fails overnight. As of October 2026, Oxylabs, Decodo, DataImpulse and Evomi advertise 24/7 support, while Byteful's main hours are weekday daytime, Proxyway says Webshare's human agents work 24/5 behind a chatbot, and SOAX's SLA promises round-the-clock response only for critical issues.
Acceptable-use policies are part of the price too. IPRoyal's policy, updated September 14, 2026, forbids circumventing "robots.txt directives" and "technical rate limits"; Rayobyte's bans crawling against a site's robots.txt or at an unreasonable rate; and Oxylabs' allows only publicly available data unless the site has given permission. A provider can close an account that breaks these rules, whatever the plan cost. Rules govern how you scrape; the next question is where the IPs come from.
Where Do Residential Scraping IPs Come From?
Residential proxy IPs belong to real people's devices, usually enrolled through apps that bundle a proxy SDK and pay the app's developer or the user. That supply chain has come under heavy pressure: Google disrupted the IPIDEA network in January 2026, the FBI warned about SDK enrollment in March 2026, and the FBI seized NetNut's domains in July 2026. Before you route business traffic through a pool, ask how its owner gets consent.
Google's threat intelligence team described the IPIDEA takedown on January 28, 2026: in one seven-day period it saw more than 550 threat groups using IPIDEA exit nodes, and it named 13 proxy and VPN brands that it found were controlled by the actors behind IPIDEA. It added that ethical-sourcing claims "are often incorrect or overstated" and must be "backed by transparent, auditable proof of user consent." On March 12, 2026, the FBI published a public warning about residential proxy networks, noting that proxy services pay app developers "for each person who downloads the application."
Research since then points the same way. In June 2026, the IP-intelligence firm Spur found residential-proxy SDKs in 2,058 of 6,038 LG and Samsung smart-TV apps it examined; Bright Data, Bright Data Ltd and Bright SDK accounted for 367 of those apps as publishers, and Honeygain UAB, which Spur describes as an Oxylabs subsidiary, appeared as the publisher of another 16. LG and Samsung then moved to stop such SDKs on their platforms. In July 2026, the FBI seized hundreds of NetNut domains in an action tied to the Popa botnet, which KrebsOnSecurity reported covers at least two million devices.
Here is what our seven publish. Oxylabs is a founding member of the Ethical Web Data Collection Initiative (EWDCI), and Decodo is a co-founder holding its "Certified" designation; Byteful, DataImpulse and Evomi are also listed as members. Membership commits a provider to shared principles, and we found no public evidence that it involves an audit.
Webshare says its pool comes from "carefully selected partners", and SOAX says its IPs are "voluntarily shared by individuals" but publishes no SDK or payment details. As Proxyway's 2026 research puts it, "simply claiming that something is the case doesn't make it real." Ask any provider these questions:
- → Do users see a clear consent screen, and can they opt out at any time?
- → Are users or app developers paid, and how?
- → Which SDKs or partner apps supply the pool?
- → Has anyone independent audited those claims, and can you read the report?
The answers tell you whose IPs you are renting. The next step is to check the IPs you actually receive.
How Can You Verify a Web Scraping Proxy Yourself?
Before you buy proxies for scraping in bulk, start with the smallest plan, collect a sample of the IPs you actually receive, and check four things: whether they sit on hosting ranges, which networks they come from, whether they carry abuse history, and whether they land where you targeted. Then measure your own success rate and cost per page on your real targets before you commit to a monthly plan.
To collect the sample, send a few hundred requests through the gateway to an IP-echo address; Oxylabs and SOAX both include one in their quick-start guides. Save the exit IPs to a file, then work through these checks.
1 Look for server IPs in a residential pool
2 See which networks the IPs come from
3 Check abuse history
4 Confirm the location
5 Measure success and cost per page
If a trial returns many 403 or 429 responses, our guide to why websites block an IP address covers the usual causes, from request rate to shared-IP reputation. Before you run that trial, check your plan against seven common buying mistakes.
What Mistakes Do Buyers Make With Scraping Proxies?
The costliest are pricing a job on 100% success, trusting advertised pool sizes and buying before reading the restricted-site list. Each one hides a cost that only shows up once you scrape at volume. These are the ones the published data and the providers' own terms point to, as of October 2026.
- Trusting advertised pool sizes. In Proxyway's 2026 research, advertised residential pools had a median of 54 million IPs and Oxylabs claimed the largest, 175 million; the most distinct IPs Proxyway actually reached in 3.6 million requests was 3,241,404, also Oxylabs. Judge a pool by what a test reaches, not by the banner.
- Reading uptime as scraping success. The same research measured a much lower success rate on real targets than on a test server, and Proxyway added: "Our scraper had issues with Google in general – it has become a tough nut to crack."
- Paying residential prices for easy targets. If a site accepts datacenter IPs, a residential gigabyte costs many times more for the same page.
- Ignoring what gets billed. Blocked responses, targeting multipliers and expiring traffic can change your cost per page more than the headline rate does.
- Buying before reading the restricted list. A plan is worthless if your main target is on it, and several providers restrict search, ticketing, banking or mail sites by default.
- Posing as Googlebot or another known crawler. Sites can check a crawler's IP with a reverse DNS lookup, as our guide to verifying Googlebot and AI crawlers shows, so a borrowed user agent marks you as fake rather than hiding you.
- Expecting a new IP to fix a blocked scraper. Sites judge request rate and behavior as well as the address. Honor the Retry-After header on 429 responses, cap concurrent requests per domain, cache pages you already have, respect robots.txt, or use an official API where one exists.
The costliest mistake: pricing a job on 100% success
With those traps in mind, here is the short version.
The Short Version
For large-scale work, Oxylabs offers the best proxies for web scraping that we could document: the most distinct IPs in Proxyway's April 2026 test, the deepest targeting and 24/7 support. The trade-offs are KYC for every customer and a restricted-target list. Decodo is the best value for most teams, with $4 pay-as-you-go traffic as of October 2026, a 14-day refund window and the clearest billing rules, though it bills blocked responses. Byteful had the best success rate on real targets, Webshare suits jobs that are mostly datacenter-friendly, SOAX suits steady US work, and DataImpulse and Evomi are the cheapest ways to buy gigabytes.
Whichever you choose, buy the smallest plan first and measure cost per page on your own targets before you commit. If your targets accept hosting IPs, our datacenter proxy rankings show where to cut the bill. If you need one stable address for long sessions instead of a rotating pool, our comparison of static ISP proxy providers is the better fit.
Prices and terms here are as of October 3, 2026, so confirm them on the vendor's page before you pay. Then set one trial proxy in your browser and run the proxy checker below to see the IP, headers and anonymity level a website receives. Some providers, Oxylabs among them, block IP-checking sites on residential proxies, so use the provider's own IP-echo address if the page does not load.