The short answer
- 1 Novada SERP API plus cities Visit site
- 2 ProxySale.com Budget SERP API Visit site
- 3 Oxylabs.io Parsed SERP endpoint Visit site
17 best proxies for SEO and rank tracking, compared
| # | Provider | Fit | From | Readers | Protocols | Links |
|---|---|---|---|---|---|---|
| 1 | 4.8/5 | $0.50 / GB | 88% | HTTP(S) | ||
|
SERP API at $0.9/1k results plus 100M+ residential IPs with city-level precision and residential pricing from $1.3/GB. Good for this job
Watch out
|
||||||
| 2 | 4.6/5 | $0.06 / GB | 100% | HTTP, HTTPS, SOCKS5 | ||
|
SERP API from $0.75/1K results plus 86M+ residential IPs with country/state/city targeting and residential pricing from $0.5/GB. Good for this job
Watch out
|
||||||
| 3 | 4.5/5 | $0.40 / GB | 100% | — | ||
|
Web API /Search at $1/1K searches sits alongside 175M+ residential IPs in 195 locations, so rank runs can use a parsed SERP endpoint. Good for this job
Watch out
|
||||||
| 4 | 4.4/5 | $0.35 / GB | 90% | HTTP, HTTPS, SOCKS5, HTTP/2 | ||
|
Residential targeting reaches city, ZIP and ASN, with unlimited concurrent connections and rollover GB that never expire. Good for this job
Watch out
|
||||||
| 5 | 4.4/5 | $0.08 / GB | 100% | HTTP(S), SOCKS5 | ||
|
City, ZIP and ASN targeting across 115M+ residential IPs, with 99.92% residential success and $2.75/GB at 100GB. Good for this job
Watch out
|
||||||
| 6 | 4.3/5 | $0.02 / month | 100% | — | ||
|
Rotating residential targets city, ZIP and ASN; 80M+ IPs and thread limits of 500-3,000 suit scheduled keyword sweeps. Good for this job
Watch out
|
||||||
| 7 | 4.2/5 | $2.20 / GB | 88% | SOCKS5, HTTP, HTTPS | ||
|
Country, city and ISP targeting with ZIP claimed, unlimited concurrent sessions and 95% clean IPs for search-result pages. Good for this job
Watch out
|
||||||
| 8 | 4.0/5 | $0.90 / GB | 86% | HTTP, HTTPS, SOCKS5 | ||
|
City-level targeting and unlimited concurrent sessions, with third-party country success rates of 96.8%-98.7%. Good for this job
Watch out
|
||||||
| 9 | 3.9/5 | $2 / GB | 88% | SOCKS4, SOCKS5, HTTP, HTTPS | ||
|
Country, city, ZIP and ASN targeting with unlimited concurrent connections and 90% ban protection claimed for Google. Good for this job
Watch out
|
||||||
| 10 | 3.8/5 | $0.63 / GB | 86% | SOCKS5, HTTP, HTTPS | ||
|
Country, city and ASN targeting at $1.00/GB rotating traffic, with a 95% clean IP rate claimed. Good for this job
Watch out
|
||||||
| 11 | 3.7/5 | $2 / GB | 80% | HTTPS, SOCKS5 | ||
|
Unlimited concurrent requests and 7-day sticky sessions, but 38+ countries is a short list for multi-city rank tracking. Good for this job
Watch out
|
||||||
| 12 | 3.6/5 | — | 81% | SOCKS5, HTTP, HTTPS | ||
|
Country and city targeting across 80M+ residential IPs, though no free trial or refund leaves pool quality untested before purchase. Good for this job
Watch out
|
||||||
| 13 | 3.5/5 | $2 | 82% | HTTP/S | ||
|
No concurrency limit and unlimited threads suit daily sweeps, but geography stops at country or US state level. Good for this job
Watch out
|
||||||
| 14 | 3.3/5 | $0.30 / GB | 96% | HTTP, HTTPS, SOCKS5 | ||
|
Residential at $0.30/GB with 220+ locations is cheap per gigabyte, but the record names no city-level targeting. Good for this job
Watch out
|
||||||
| 15 | 3.2/5 | $5 / proxy | 85% | — | ||
|
Per-proxy billing from $5 and 150+ locations, but no city targeting or per-GB option for many small rank checks. Good for this job
Watch out
|
||||||
| 16 | 2.5/5 | ₽490 / day | 83% | SOCKS5, HTTP | ||
|
Mobile proxies billed per day with unlimited traffic, not per gigabyte, and no city targeting stated for rank checks. Good for this job
Watch out
|
||||||
| 17 | 1.9/5 | $1 | 82% | — | ||
|
Shared and dedicated static proxies with a $10 minimum and no city targeting or SERP endpoint for keyword tracking. Good for this job
Watch out
|
||||||
Ordered by how well each provider fits this job, from the figures the providers publish, our review scores and reader votes — see our methodology.
Why rank tracking needs proxies
A rank is not a property of a keyword. It is a property of a keyword, a location, a device and a moment. Someone searching emergency plumber from a phone in Bristol and someone searching the same words from a laptop in Edinburgh can get different first pages, and neither of them is seeing "the" ranking. If you want to know what the Bristol searcher sees, the request has to leave from an address the search engine treats as Bristol.
That is the entire job. For a handful of keywords you can get away with a VPN and a private window, switching exits by hand. It stops working for two reasons. The first is volume: 400 keywords across 8 cities once a day is 3,200 requests before you have checked a single mobile result, and a consumer connection does not produce that pattern. The second is that a single address making that pattern gets challenged, and once it is challenged it stays challenged. You then get the "unusual traffic" interstitial, or a captcha, or — worse — an HTTP 200 carrying a degraded page with no organic results on it.
There is a third reason that people forget until it bites. Without a proxy, the address absorbing all of that is your office IP. Losing it means your team cannot search normally for a day, and you find out at the worst possible moment.
Why your own browser is the worst rank checker
A private window is not a neutral observer. It clears cookies; it does not clear your address, your device, your screen size, your fonts or your language settings, and it does not clear the personalisation attached to a signed-in session elsewhere on the machine. A logged-in Chrome profile browsing the same query twice a week is the least representative sample you can build.
Manual checking also produces a number with no provenance. There is no timestamp, no city, no device class and no stored response, so when a rank moves you cannot tell whether the ranking changed or the measurement did. Two people checking the same keyword on the same morning will disagree and neither will be able to prove anything.
One more confusion worth naming. Search Console's average position is not a rank. It is an average across queries, locations and devices, over impressions you actually received, and it only exists for queries where you appeared. It is a useful signal and a bad substitute for "position 4 in Manchester on mobile for this exact phrase". Track both if you like, but do not plot them on the same axis.
Which type of IP survives a search engine
Search engines are the most aggressive target most trackers hit, because classifying a request as automated is a solved problem for them and the cost of a false positive is low. What gets classified is not just the address.
- Datacentre addresses. Their ranges are published and easy to label. Light volumes against Google can work; sustained sweeps are usually challenged quickly. Against Bing and smaller engines they last considerably longer.
- Residential addresses. Real consumer connections, sold by the gigabyte or by the port. They look like households, which is the point, but they are shared, so a neighbour scraping at volume burns the address for you and the failure appears as a missing rank rather than an error.
- Static residential / ISP addresses. A residential-looking address that stays put. Useful when you want the same identity for a whole keyword set, less useful when an address with history gets flagged and you cannot rotate off it.
- Mobile (4G/5G) addresses. Carriers put thousands of real subscribers behind the same address, so blocking one outright costs the search engine a lot of legitimate traffic. These survive longest and are the slowest and most expensive. City targeting is also vaguer, because the carrier gateway decides where you exit, not you.
Address type matters less than the whole request. A residential IP paired with a TLS fingerprint that says Python, an Accept-Language that contradicts the exit country and no cookies at all is more conspicuous than a clean datacentre address behaving like a browser. Decide whether you need sticky sessions (same address for a few minutes, a few dozen queries) or per-request rotation, and prefer sticky for keyword sweeps: one address doing exactly one search and then vanishing is itself a pattern.
Whatever you buy, verify the exit before you trust it. Check the city field, not the country field — many pools sell "UK" and only have London. What is my IP does this in a browser; the proxy checker does it across a batch.
How to set up localised rank tracking
Decide the unit of geography before you buy anything. Country is cheap and easy. City is where local SEO work actually lives. Postcode-level is mostly marketing: almost no pool reliably exits from a specific postcode, so a postcode campaign is a city campaign with a finer label on the report. Be honest about that in the deliverable.
Then, in order:
- Pick the city and verify the exit.
curl -x http://user:pass@gateway:8080 https://ipinfo.io/jsonand read the city field. If the pool cannot give you Leeds, no amount of parameter tuning will. - Set language and country.
hlsets interface language,glsets country. Neither moves you to a city.gl=uksent from a Frankfurt address is still a Frankfurt request with a British interface. - Use the location parameter, carefully. Google accepts a
uuleparameter carrying an encoded canonical location name. It is undocumented, the encoding has changed over time, and it is best-effort. Verify it worked by checking that the local pack in the response names businesses in the right city. - Match headers to the location.
Accept-Language: en-GB,en;q=0.9alongside a UK exit, a current desktop or mobile user agent, and the same values on every run. A UK exit sendingen-USis a contradiction a classifier can read. - Fix the device profile and record it. Mobile and desktop results differ enough that mixing them makes a trend line meaningless. Either pick one or run them as two separate series.
- Store the raw response with the timestamp. Parsers break and layouts change. Re-fetching a fortnight of history gives you today's rankings labelled as last week's.
- Handle the consent interstitial. In the UK and EU a fresh session may be served a consent page with no organic results. A parser that does not recognise it records a rank of zero.
A working request looks roughly like this:
curl -x http://user:pass@gateway:8080 -H 'Accept-Language: en-GB,en;q=0.9' -A 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0 Safari/537.36' 'https://www.google.com/search?q=emergency+plumber&gl=uk&hl=en&num=20'
Cross-check the series against Search Console for the same period. If your tracked positions and your impression-weighted average position diverge wildly, the series is wrong, not Search Console.
SERP APIs versus raw proxies
With a raw proxy you own the request: the headers, the retries, the consent handling, the parser, the failure classification. You pay less per fetch and more in engineering, and you can see exactly why a request failed. That last point matters more than it sounds — a raw proxy lets you distinguish a block from a genuine drop out of the top 100.
A SERP endpoint hands you parsed JSON for a query, a location and a device. The provider absorbs the blocking, the retry logic and the parser maintenance, and charges more per search for it. You give up three things: visibility into failures, control over the device fingerprint, and any location that is not on their list. Their "Manchester" may be a data centre with a Manchester geo flag, which is fine until a client asks why the local pack is empty.
A hybrid is common and sensible: the endpoint for the bulk of a large sweep, raw proxies for the twenty keywords that decide the account and for spot checks when the two disagree. Some providers sell both from one account, which removes most of the integration work. Individual provider write-ups sit in the reviews section.
What this costs per keyword
The billable unit is a keyword-location-device triple, and it is worth costing one before you commit to a plan. Take 800 keywords across 6 cities once a day: 4,800 fetches per sweep, roughly 144,000 a month. Add a mobile series and double it. That is the number you take to a price list.
For a byte-priced pool, a search results page is a few hundred kilobytes of HTML uncompressed, and gzip cuts that by roughly three quarters if your client sends Accept-Encoding. Planning at 150–250 KB transferred per fetch puts a 4,800-fetch sweep near a gigabyte and a month near 30 GB; your figures will move with result count and compression, so measure one sweep rather than trusting an estimate. This is why per-gigabyte pricing suits rank tracking: you need a lot of bytes and a modest number of concurrent sessions, not a large stable set of identities. Per-IP pricing charges you for the thing you need least.
For an endpoint priced per search, multiply 144,000 by the per-thousand rate and compare that directly with the byte estimate plus your own engineering time. Be realistic about the second half: a parser, a retry queue, a failure classifier and someone to maintain them is a real cost, and it is usually the number that decides the argument.
The savings are in the schedule, not the vendor. Track daily the keywords that carry revenue, weekly the long tail, and the top five cities by revenue rather than all forty. Cache within the day — a rank at 09:00 is still that rank at 09:05. Public lists such as our free proxy list are fine for learning the plumbing and seeing what a 429 looks like; they are not a production source, because the addresses are public, shared and mostly already burnt, and a plaintext HTTP proxy can read and alter what passes through it.
Mistakes that corrupt a rank report
Most bad rank reports are not caused by bad proxies. They are caused by a report that cannot tell the difference between four different things.
- Recording a block as a missing rank. A captcha page returns HTTP 200 with no organic results, the parser finds nothing and writes "not in top 100", and the chart shows a cliff that never happened. Classify "failed" separately from "not found" and exclude failures from the average.
- Silent location drift. A pool falls back to another city when the requested one is short. Your trend line is now noise with a plausible shape. Log the actual exit city with every fetch.
- Changing the request between runs. A new user agent, a different
Accept-Languageor a switched device class mid-series makes every comparison invalid. - One sample per keyword per day. Rankings are noisy near the top. Take three fetches and report the median or the mode, rather than whichever one you saw first.
- Ignoring SERP features. AI overviews, local packs, shopping units and People Also Ask push organic results down the page. A move from 3 to 5 may be layout, not ranking. Record which features were present.
- Inconsistent timing. Rankings move during the day. A sweep that starts at 06:00 on Monday and 14:00 on Tuesday is measuring the clock.
- Cookies leaking between requests. A session reused across cities carries the previous city's preferences into the next SERP.
- Trusting unchallenged geo claims. If nobody has checked the exit city for a week, assume it has drifted.
- Throwing away the raw response. Without it, every parser fix requires a re-fetch, and every re-fetch backdates today's data.
The test of a rank report is whether a single number in it can be traced back to a stored response, an exit city, a device and a timestamp. If it cannot, it is a chart of your infrastructure, not of your rankings.
What to look for
- City and postcode targeting
- A rank is a local fact. Tracking from the wrong city returns a result nobody in your market sees.
- A SERP endpoint, or just proxies
- Some providers sell a parsed search endpoint alongside the pool, which removes the part of rank tracking that breaks most often.
- How well the pool survives search engines
- Search engines are the most aggressive target most trackers hit, and a burnt address shows up as a missing rank, not an error.
- What a tracked keyword costs
- Rank tracking is many small requests, so per-gigabyte pricing is kinder here than per-IP pricing.
- Concurrency for scheduled runs
- A daily sweep of thousands of keywords is a burst, and a capped pool turns a burst into a queue.
Every provider, in order
1. Novada — SERP API plus cities
4.8 / 5 for this job · from $0.50 / GB
Novada sits first because it is the only provider here selling a parsed SERP endpoint and city-level targeting from the same account, which removes the two things that break a rank-tracking run most often. The record puts the residential pool at 100M+ IPs across 195+ countries with city-level precision, and the SERP API is listed against Google, Bing, Yahoo and more. Because the scraping products draw on a single credit balance, a tracker can pull results from the SERP endpoint and top up the same balance for the page fetches behind the proxy pool, rather than running two vendors and reconciling two invoices.
Published country counts are the useful part of the pool claim: 8,426,658 IPs in the United States, 2,838,522 in the UK, 3,167,312 in Brazil, 1,645,341 in Canada and 1,257,655 in France. A local rank is a local fact, so those figures say more about whether a city can be served than the headline total does. Residential sessions run 1–120 minutes, and rotating ISP holds 360-minute sticky sessions, which suits a keyword sweep where one address fetches several positions from the same locale. Mobile inventory is separated out at 2M+ IPs and 120+ carriers. Residential starts at $1.3/GB, rotating datacenter at $0.50/GB, rotating ISP at $0.65/GB, and the SERP API at $0.9/1k res, which is the billing shape rank tracking wants. There is a $10 credit with no card required and a 10GB new-user residential pack for $10.
The catch is the success-rate claim: 99.99% is quoted overall but 99% for rotating datacenter, so the headline does not cleanly cover every pool a large sweep might touch. Unlimited Proxies, IPv6 Proxies and Datasets have no published price and go through an account manager. The free trial covers the Scraper API only; the general offer is a $10 credit, which is a different thing.
Key features
- SERP API covering Google, Bing and Yahoo
- City-level precision across 195+ countries
- 100M+ residential IPs, with country counts published for the US, UK, Brazil, Canada and France
- Residential sessions of 1–120 minutes; 360-minute sticky sessions on rotating ISP
- SERP API from $0.9/1k res on a credit balance shared across scraping products
- $10 credit with no card, plus a 10GB new-user residential pack for $10
Best suited for: teams that want parsed search results and city-targeted proxies billed from one credit balance
2. ProxySale.com — Budget SERP API
4.6 / 5 for this job · from $0.06 / GB
Proxy-Sale takes second because it pairs a cheap SERP endpoint with city-level targeting on a pool the provider puts at 86M+ IPs across 200+ countries. For rank tracking the arithmetic is simple: each keyword is a request, a daily sweep is thousands of them, and a per-result SERP price of $0.75/1K results keeps the recurring cost down. The Web Scraper API comes in lower still at $0.39/1K results and bills only for successfully delivered results, which matters when a failed fetch is a missing rank rather than an error you can see.
On the proxy side, residential targeting covers country, state and city, which is the level a local rank needs, and the residential product carries HTTP, HTTPS and SOCKS5 on the same service. The provider quotes 99.5% average success rate and response times under 0.6s for residential, alongside 99.9% uptime. Billing is usage-based per gigabyte rather than per IP, which is the friendlier shape for many small requests. Static IPs are available at $5/IP residential and $3/IP data centre for jobs that need a fixed address. The Unlimited Residential plan removes IP and traffic caps with random and sticky rotation, but it is billed at $69.68 per day, so it only makes sense on unusually heavy runs.
The awkward parts are worth stating. Uptime, success rate and response time are provider-declared with no third-party audit shown, and the service is not available in mainland China by the provider's own policy. The 30-day zero-setup trial covers the Web Scraper API, not the proxy network, so you cannot test the pool itself for free.
Key features
- SERP API from $0.75/1K results
- 86M+ residential IPs across 200+ countries
- Country, state and city targeting on residential proxies
- HTTP, HTTPS and SOCKS5 supported
- Web Scraper API billed only on successfully delivered results, with a 30-day trial
- 99.5% residential success rate and sub-0.6s response, provider-declared
Best suited for: budget-conscious rank tracking where a per-result SERP endpoint and city targeting outweigh audited performance figures
3. Oxylabs.io — Parsed SERP endpoint
4.5 / 5 for this job · from $0.40 / GB
Oxylabs earns third on the strength of its parsed search endpoint and, more unusually, published unit rates at every tier. The Web API /Search is listed at $1/1K searches on every plan from the $5 welcome credits up to enterprise, so a keyword budget can be modelled before anyone talks to sales. That consistency is rare, and for rank tracking — where cost scales with keyword count times check frequency — it is the difference between a plan you can cost and one you guess at. The pool is quoted at 177M+ IPs, 175M+ of them residential, across 195 locations.
The same transparency applies to proxies: residential falls from $6/GB on Starter to $5/GB on Basic, $4/GB on Flexible and $2.5/GB at enterprise, and datacenter runs from $0.6/GB down to $0.40/GB. One subscription carries the whole catalogue, which helps if a tracker needs a parsed endpoint for the SERP and raw residential traffic for landing-page checks. Compliance is the other reason it ranks here: ISO/IEC 27001:2022 and SOC 2 Type II are held, GDPR and EU AI Act alignment are stated, and Technology E&O and cyber insurance apply across products. For agencies reporting to clients on how data was gathered, that paperwork is often the deciding factor.
The catch is entry cost. The only free route is $5 in welcome credits, which is under a gigabyte of residential traffic at Starter rates, and the cheapest paid plan is $30/month. Enterprise starts at $2,500/month through sales. VAT may apply on Starter, Basic and Flexible, so the headline rate is not always the invoice, and Bitcoin is not accepted. Certification scope is per product, not company-wide: ISO/IEC 27001:2022 and SOC 2 Type II each name specific products rather than the entire catalogue.
Key features
- Web API /Search at $1/1K searches across every tier
- 175M+ residential IPs within 177M+ total, across 195 locations
- Residential from $6/GB on Starter to $2.5/GB at enterprise
- Every product and tier priced publicly before contact
- ISO/IEC 27001:2022 and SOC 2 Type II, each covering named products
- $5 welcome credits; cheapest paid plan $30/month
Best suited for: agencies and in-house teams that need to cost a keyword sweep precisely and justify the data source internally
4. ProxyEmpire — Rollover city targeting
4.4 / 5 for this job · from $0.35 / GB
ProxyEmpire is fourth and it wins its place on two things a scheduled rank tracker feels immediately: rollover bandwidth and genuinely granular targeting. Unused gigabytes roll over and never expire on GB-based plans, so a month where you test heavily and a month where you barely run do not produce the same waste. Residential targeting reaches country, region, city, ISP, ASN, OS/TCP and ZIP, which is about as close to a local rank as a proxy can get you without being in the room. The pool is quoted at 30M+ residential IPs across 170+ countries.
Pricing is prepaid with no contracts and no monthly minimums, and the entry point is small: $3.50 for a 1GB pay-as-you-go gigabyte, falling to $1.50/GB on the 1000GB tier. Code EMPIRE50 takes 50% off rotating residential and rotating mobile on eligible GB-based plans. Unlimited sessions and concurrent connections are offered for approved use cases, which matters when a daily sweep is a burst rather than a trickle and a capped pool would turn it into a queue. There is also a $1.97 trial with 150MB that does not auto-convert into a subscription, plus a public API with an endpoint generator.
The awkward parts are real. The cheapest 150MB trial excludes static ISP and datacenter and covers unlimited residential only on request. Unlimited bandwidth is billed per day or week with no rollover, from $200/day, and EMPIRE50 does not apply to it. Coverage is uneven: 62 datacenter countries and 24 static ISP countries against 170+ for residential and mobile, and dedicated mobile reaches 5+ countries and 50+ locations. The site also quotes its Trustpilot score as both 4.6 and 4.5 from 100+ reviews, which is the sort of detail that makes you check everything else twice.
Key features
- 30M+ residential IPs across 170+ countries
- Rollover bandwidth; unused GB never expire on GB-based plans
- Residential targeting by country, region, city, ISP, ASN, OS/TCP and ZIP
- Pay-as-you-go from $3.50/GB; 1000GB tier at $1.50/GB
- $1.97 trial with 150MB that does not auto-convert
- No contracts, no monthly minimums, all plans prepaid
Best suited for: teams running irregular or seasonal sweeps who want unused bandwidth to survive the quiet months
5. Decodo (ex Smartproxy) — City and ZIP targeting
4.4 / 5 for this job · from $0.08 / GB
Smartproxy, now trading as Decodo, takes fifth on targeting depth rather than price. The residential pool is quoted at 115M+ IPs across 195+ locations, with targeting down to city, ZIP code and ASN rather than country alone. For rank tracking that is the whole argument: a city-level rank checked through a national pool is a number nobody in that market sees, and ZIP-level targeting gets closer to how local search actually resolves. Country-level counts back it up, including 16M+ IPs in the United States, 5M+ in the UK, 5.6M in Germany, 11M in India and 4.6M in France.
Sessions rotate per request or hold one address for up to 30 minutes, which is long enough to fetch a full page of results from one locale before moving on. HTTP(S) and SOCKS5 are both supported, so older SOCKS5-only clients still work. Pricing is bandwidth-based and falls with volume: $3.75/GB at 3GB down to $2.75/GB at 100GB, with pay-as-you-go at $4/GB and enterprise at $2.50/GB from 250GB and $2/GB at 1TB. There is a Web Data API from $0.15/1K requests for teams that would rather not parse the SERP themselves, and a 3-day 100MB proxy trial plus a starter plan with $5 in credits and no card. The provider quotes 99.92% success rate and sub-0.5s average response on residential.
The catch sits in the refund rules. The 3-day trial and the 14-day money-back option are mutually exclusive — using the trial removes refund eligibility — and refunds are limited to a first purchase, must be requested within 14 calendar days, and are capped at under 20% of a residential plan's traffic. Pay-as-you-go at $4/GB is wallet-only, sold one gigabyte at a time, and cannot be combined with a monthly subscription. High-Speed Proxies carry no published price.
Key features
- 115M+ residential IPs across 195+ locations
- City, ZIP code and ASN-level targeting
- Sticky sessions holding one IP for up to 30 minutes
- HTTP(S) and SOCKS5 support
- Residential from $3.75/GB at 3GB to $2.75/GB at 100GB
- Web Data API from $0.15/1K requests; 3-day 100MB proxy trial
Best suited for: trackers that need ZIP-level granularity and sticky sessions long enough to pull a full results page
6. Webshare.io — Low-cost city targeting
4.3 / 5 for this job · from $0.02 / month
Webshare closes the list at six, and it is here for one reason: it is the cheapest way to put a lot of addresses behind a rank-tracking script. Per-IP pricing falls from $0.0299 to $0.0179 as volume grows, rotating residential starts at $3.50/month, and there is a permanent free plan with 10 proxies and no credit card. If you are checking a few hundred keywords across a handful of cities, that entry cost is close to nothing, and the ability to test before buying is not a marketing line here — it is the product.
Targeting is better than the price suggests. Rotating residential filters by country, state, city, ZIP code and ASN, which is what a local rank needs, and the pool is quoted at 80M+ IPs. Both HTTP and SOCKS5 endpoints are available on the same credentials, IP authorisation is offered as an alternative to username and password, and there is a Chrome extension for anyone who would rather not script it. Bandwidth tiers run from 250GB to unlimited with thread limits from 500 to 3,000, so a scheduled sweep has room to burst instead of queueing.
The awkward parts should be read before buying. The unlimited bandwidth claim carries an asterisk pointing to a Fair Usage Policy, and that policy is the actual limit. The provider's own pages give both 195 countries and 50+ locations, so the real footprint is hard to pin down, and the static residential list names Comcast and AT&T and then says "& more", so you cannot check a specific carrier before paying. The 99.7% success rate and 99.97% uptime are the provider's own figures with no independent verification alongside them. Datacenter depth — 400K+ IPs — is the strongest part of the network; residential is the part to watch.
Key features
- Permanent free plan with 10 proxies, no card required
- Per-IP pricing from $0.0299 down to $0.0179 at volume
- Rotating residential from $3.50/month
- Country, state, city, ZIP code and ASN targeting on rotating residential
- HTTP and SOCKS5 on the same credentials; IP authorisation available
- Bandwidth from 250GB to unlimited, threads from 500 to 3,000
Best suited for: small teams and solo SEOs who want city-level checks at the lowest possible entry cost and can live with unaudited performance claims
7. NodeMaven — Clean-IP city targeting
4.2 / 5 for this job · from $2.20 / GB
NodeMaven sits seventh because the thing it does best for rank tracking — a clean address in a specific city — is also the thing it charges most for. The record lists 30M+ residential IPs and 190+ countries with targeting down to ZIP level, plus city and ISP filters. For a local rank grid that is the correct shape of product: a rank pulled from the wrong city is a different rank.
The figure that matters on search engines is the clean-IP rate: the provider quotes 95%, with a residential fraud score under 5/100 and ISP under 10/100. Sticky sessions run up to 24 hours with unlimited concurrent sessions, so a scheduled sweep does not queue behind a session cap, and traffic rolls over between plans. Speeds are quoted under 0.2s for residential and under 0.1s for static ISP.
The catch is cost and evidence. Billing is per gigabyte from $2.20/GB, which is kind to many small requests but adds up across thousands of keywords, and there is no free trial — the $3.50, 750 MB entry point is a paid demo. Smaller locations are reported to be uneven in depth, exactly where a city-level grid is thinnest, and refunds are pro-rated rather than full. What the record documents is a RESTful API and SDKs for pulling proxies; a parsed SERP endpoint is not part of the offer as listed, so plan on doing the parsing yourself.
Key features
- Residential from $2.20/GB, with traffic rollover on monthly and pay-as-you-go plans
- Targeting by country, city and ISP, with ZIP level claimed by the provider
- 95% clean IP rate; residential fraud score under 5/100
- Sticky sessions up to 24 hours with unlimited concurrent sessions
- Paid 750 MB demo at $3.50; no free trial
- RESTful API and SDKs for all platforms
Best suited for: city-level rank grids where a clean residential address matters more than the per-gigabyte price
8. FlyProxy — Unlimited concurrent city runs
4.0 / 5 for this job · from $0.90 / GB
FlyProxy takes eighth on price at volume rather than on polish. Residential runs $2.50/GB at 20 GB, $2.40/GB at 50 GB and $0.90/GB at the 10,240 GB tier, which is the shape of pricing a rank tracker wants: many small requests, billed by the gigabyte, with no per-IP charge attached to the addresses they land on. Unlimited concurrent sessions are stated by both the provider and the third-party listing, so a nightly keyword burst goes out as a burst rather than a queue.
Targeting reaches state and city in the third-party listing, against country and city in the provider's own wording. Third-party country tests put success between 96.8% in Mexico and 98.7% in Germany, with average response times from 0.52s to 1.00s and an IP lifetime of around 24 hours. On search engines that spread is the whole story: the soft end shows up as a missing rank rather than an error, and you cannot tell a genuine absence from a burnt address without re-running the check.
The weaknesses are structural. There are no mobile proxies, so the address type that survives search engines longest is not on the menu. Refunds are not offered once a product is sold, stated by the provider itself, and the free-trial position disagrees between sources — a 500M offer through sales against a third-party listing of no trial. Pool size (100M+ provider, 69M third-party), the protocol list and dashboard availability also disagree between pages. Committing to volume is the only way the price works, and nothing is refundable if it does not.
Key features
- Residential $0.90/GB at the 10,240 GB tier; $2.50/GB at 20 GB
- Unlimited concurrent sessions, stated by both the provider and the third-party listing
- 195+ countries and regions; state and city targeting in the third-party listing
- Third-party country success from 96.8% in Mexico to 98.7% in Germany
- No mobile proxies in the line-up
- No refunds once products are sold, per the provider
Best suited for: high-volume scheduled keyword sweeps where per-gigabyte cost and unlimited concurrency outweigh the absence of mobile addresses
9. GonzoProxy — ZIP and ASN targeting
3.9 / 5 for this job · from $2 / GB
GonzoProxy sits ninth because its targeting is finer than most and its price at small volumes is not. The filter set covers country, city, state, continent, ZIP code and ASN across a 20M+ pool in 190+ countries. ZIP and ASN matter here: a ZIP gets you closer to a real local result than a city centroid, and ASN lets you see a result the way one ISP's customers see it.
Billing is per gigabyte with no expiry, and unused traffic rolls into the next plan, which suits a tracker sending a few hundred kilobytes per keyword. Connections are unlimited on every plan, sticky sessions are claimed up to 72 hours, and there is no KYC — the provider says access is live in about 30 seconds. The provider also claims 90% ban protection for Facebook, Google and TikTok, ISO/IEC 27001:2017 certification, and country success rates of its own from 96.8% in Indonesia up to 99.3% in Germany.
The catches are the ones that bite a rank tracker. The ladder starts at $6.5/GB for 1–24 GB and only reaches $2/GB at 5000+ GB, so a small sweep is the expensive kind. Mobile proxies cover Poland and Ukraine only. The dashboard lacks detailed per-request logs, which is precisely the record you want when a keyword comes back empty. The demo costs $1.99 for 0.3 GB and a separate trial row is $4.50 for 1 GB, and two published plan ladders quote different per-gigabyte rates for the same volumes. The speed and success figures are the provider's own and are not independently verified on its site.
Key features
- Targeting by country, city, state, continent, ZIP code and ASN
- 20M+ IP pool across 190 countries
- Unlimited concurrent connections on every plan; sticky IP claimed up to 72 hours
- Traffic never expires and rolls over to the next plan
- Residential at $6.5/GB for 1–24 GB, down to $2/GB at 5000+ GB
- Mobile proxies cover Poland and Ukraine only
Best suited for: ZIP- and ISP-level rank checks where the grid has to match a real local result and the monthly volume reaches the lower pricing tiers
10. Floppydata — Low-cost city rotation
3.8 / 5 for this job · from $0.63 / GB
Floppydata is tenth because it is the cheapest route onto this list for rotating through cities, and the least proven. Rotating traffic starts at $1.00/GB across residential, mobile and datacenter, with open tiers of 20% off at 96 GB and 30% off at 120 GB; the $0.63/GB headline belongs to the datacenter bundle at 120 GB, not to residential. Targeting covers country, city, ASN and IP type across 190+ locations, which is the combination a rank grid needs.
For many small requests, per-gigabyte billing is the right meter, and a free 100 MB trial for each of residential, datacenter and mobile lets you check that a city actually resolves before committing. There is a client API at api.floppydata.net authenticated with an X-Api-Key header, so runs can be scripted, and the provider claims a 95% clean IP rate alongside SOC 2, CCPA and GDPR compliance.
The awkward parts: the company was founded in April 2024, so the pool has a short track record, and the 99.99% uptime and sub-0.3-second response times are the provider's own claims. The 100 MB trial is a connectivity check, not a load test. There is no phone support and the documentation is thin on custom rotation logic, which is where a scheduled sweep usually needs help. The datacenter tier is the wrong address type for search engines however cheap it is — those ranges are recognised quickly, and the saving shows up as missing ranks.
Key features
- Rotating traffic from $1.00/GB across residential, mobile and datacenter
- Country, city, ASN and IP-type targeting over 190+ locations
- 100 MB free trial for each of residential, datacenter and mobile
- 20% off at 96 GB and 30% off at 120 GB; datacenter at $0.63/GB in the 120 GB bundle
- Client API at api.floppydata.net with X-Api-Key authentication
- Founded April 2024; 99.99% uptime and sub-0.3s response are provider claims
Best suited for: budget-limited city rotation where a free per-type trial and per-gigabyte billing matter more than a long operating record
11. AnyIP.io — Long sticky sessions
3.7 / 5 for this job · from $2 / GB
AnyIP.io is eleventh because its strongest feature for this job is not the one rank tracking usually asks for. Sticky sessions run up to seven days, with Smart Sessions holding an address inside the same carrier and area, plus unlimited rotation and unlimited concurrent requests. That is built for continuity — a logged-in session, a checkout, a multi-step flow — rather than for a sweep that wants a fresh city every few seconds.
The plan structure is simple: residential and mobile on one plan at one price, metered only by bandwidth, with filtering by country, ISP, regions and precise GPS coordinates. Unused data rolls over, there is no contract, and orders are processed automatically. The provider quotes 100M+ residential and mobile IPs, a request success rate of about 98.6% and an average response time of about 0.6s.
Two things work against it here. The country list is 38+, which is short for multi-market rank tracking, and the geo filters stop at regions and coordinates rather than naming cities, so building a city grid takes more work than with a provider that sells city targeting directly. The pricing also hides the rate: $2/GB is the Enterprise rate on $1,000 for 500 GB, while Micro is $25 for 5 GB with extra traffic at $5/GB and Basic is $80 for 20 GB with extra at $4/GB. There is no free trial — Micro is the trial and it is paid up front — and crypto payments fall outside the 14-day money-back guarantee, which covers cards only.
Key features
- Sticky sessions up to 7 days; unlimited rotation and unlimited concurrent requests
- Residential and mobile on one plan, metered by bandwidth only
- Filtering by country, ISP, regions and precise GPS coordinates
- $2/GB is the Enterprise rate; Micro is $25 for 5 GB with extra at $5/GB
- 38+ countries; unused data rolls over month to month
- No free trial; the 14-day money-back guarantee covers card purchases only
Best suited for: long-running regional rank checks that need one address held for hours rather than a city grid swept fresh each run
12. Nsocks — City targeting, no trial
3.6 / 5 for this job
Nsocks is last because buying it is a commitment made without evidence. City targeting is real — country and city per the provider, country, city, ISP and ASN per third parties, across 190+ country and city-level regions — and the pool is quoted at 80M+ real-user residential IPs. But there is no free trial and the provider states it does not refund products sold, so the first payment is also the test.
What you get for it: 99.95% uptime claimed, SOCKS5, HTTP and HTTPS, nine interface languages, a dashboard with traffic statistics, 24/7 live chat and email, and integrations with most major anti-detect browsers alongside Playwright, Selenium and an API. Rotating residential is sold in tiers from 10 GB to 500 GB; static residential and datacenter are per IP with unlimited bandwidth, capped at 20 Mbps.
The problems are the ones a rank tracker notices first. The provider's own home page quotes two pool sizes, 80M+ and 90M+, which is not a rounding difference anyone can check from outside. The record carries no published per-gigabyte figure for the residential tiers, so there is no way to work out what a keyword costs before you are inside the dashboard. The company was founded on 26 August 2024, so the pool has a short history of its own, and search engines are the hardest target to keep an address alive against. City targeting at a low price is genuinely useful; buying it blind, with no trial and no refunds, is the part to weigh first.
Key features
- Country and city targeting; country, city, ISP and ASN per third parties
- 80M+ real-user residential IPs claimed, while the home page also states 90M+
- No free trial, and the provider does not refund products sold
- Residential sold in tiers from 10 GB to 500 GB, with no per-gigabyte figure published
- Static residential and datacenter sold per IP with unlimited bandwidth, capped at 20 Mbps
- Founded 26 August 2024; 99.95% uptime claimed
Best suited for: teams that already know the market they need and will accept a paid first commitment in exchange for cheap city-level targeting
13. Live Proxies — Unlimited SEO threads
3.5 / 5 for this job · from $2
Live Proxies sits at this position because of one line in its record: no concurrency limit, with unlimited threads and unlimited requests as stated by the provider. A daily rank sweep is a burst — thousands of keywords fired at once and then nothing until tomorrow — and a pool with a thread cap turns that burst into a queue. The provider also says IPs are privately allocated and shared only with you, which is the other half of the problem for search work: an address nobody else has already burnt on a search engine.
Rotating residential carries 60-minute sticky sessions and rotating mobile 30 minutes, so a keyword check can stay on one address while it walks through result pages. Metered and unmetered billing are both offered, and unused bandwidth rolls over to the standby plan, which helps when a tracking month slips. Dashboard access, API access and IP whitelisting are included; SOCKS5 is on the B2B tier only. Activation is quoted at about ten minutes, with network activation up to twenty.
The catch is geography, the one thing rank tracking cannot fudge. The record describes coverage as available countries or state-level options with no count given, and state IPs are US only. If your market is a city or a postcode outside the US, that tells you nothing about whether you can hit it. Pricing is also muddled: the home page shows $45.00, $65.00 and $55.00 per plan while the pricing page says B2C plans start from $70, and static residential is marked permanently unavailable. The free trial goes to custom plan users after an agreement, with B2C buyers pointed at the smallest paid plan instead.
Key features
- No concurrency limit; unlimited threads and unlimited requests, per the provider
- Private IP allocation, with addresses shared only with you
- 60-minute sticky sessions on rotating residential, 30 minutes on rotating mobile
- Metered and unmetered billing; unused bandwidth rolls over
- Coverage given as available countries or state-level options, with no count
- SOCKS5 on the B2B tier only
Best suited for: burst keyword sweeps where the concurrency cap rather than the pool is the bottleneck
14. Proxy-Seller — Large pool per GB
3.3 / 5 for this job · from $0.30 / GB
Proxy-Seller earns this slot on price shape. Residential starts at $0.30/GB, and rank tracking is many small requests rather than a few large downloads, so billing that follows traffic instead of address count suits the job. The declared pool is 47M+ rotating residential IPs across 220+ locations, which is the spread you want when a keyword has to be checked from the right country rather than from a datacentre in Frankfurt.
HTTP, HTTPS and SOCKS5 are supported with a RESTful API and IP whitelisting, and the provider states no monthly lock-ins — worth something when a tracking project runs for six weeks and stops. The compliance paperwork is unusually complete for the price band: ISO/IEC 27001:2022 and ISO 9001:2015 certification, GDPR, CCPA and ePrivacy compliance, and DPA/SCC readiness. That matters when the proxy spend has to clear a procurement review before anyone can run a report.
Two things to weigh. Mobile coverage reaches 20+ countries against 220+ residential locations, so if the plan was to lean on mobile addresses for the engines that fight hardest, the footprint is far thinner than the headline suggests. And the uptime, success-rate and response-time figures are the provider's own, with no independent test cited, so treat them as claims rather than measurements. The record lists no parsed search endpoint, which leaves SERP parsing to you: the part of a rank tracker that breaks most often stays your responsibility.
Key features
- 47M+ rotating residential IPs across 220+ locations, as declared
- Residential from $0.30/GB
- HTTP, HTTPS and SOCKS5, with RESTful API and IP whitelisting
- No monthly lock-ins stated
- Mobile coverage reaches 20+ countries
- ISO/IEC 27001:2022 and ISO 9001:2015 certification stated
Best suited for: per-gigabyte keyword tracking at volume, where the bill follows traffic rather than addresses
15. ASocks.com — Per-proxy budget pick
3.2 / 5 for this job · from $5 / proxy
ASocks.com is the budget option per address, not per gigabyte. Residential and corporate proxies are $5 per proxy per month and mobile is $15, with volume discounts of 5% from 10 proxies, 10% from 50 and 15% from 100. For a tracked keyword set small enough to sit on a fixed number of addresses, that is a predictable monthly figure instead of a meter running on every fetch.
The declared footprint is 150+ locations and 7M+ IP addresses, and the provider runs its own infrastructure, which it links to cleaner addresses and fewer bans. That is the claim that matters most here, because a burnt address does not announce itself: it comes back as a missing rank. Support is 24/7 with a stated average response under 10 minutes, which helps when an overnight sweep stalls at three in the morning.
The catch is that none of it is audited. The 99.7% connection success rate, the 150+ locations, the 7M+ IP addresses and the 100,000+ clients are all provider-reported, and the top-G2 rating is the provider's own summary. Per-proxy billing also gives you a fixed set of addresses rather than a pool you draw from, so a large scheduled sweep spread across a small address count is exactly the pattern search engines are good at spotting. Mobile costs three times the residential rate, so leaning on mobile for the harder engines is a real budget decision rather than a footnote.
Key features
- $5 per proxy/month for residential and corporate; $15 for mobile
- Volume discounts: 5% from 10 proxies, 10% from 50, 15% from 100
- 150+ locations and 7M+ IP addresses, as claimed by the provider
- Own infrastructure, which the provider links to cleaner addresses
- 24/7 support with a stated average response under 10 minutes
Best suited for: small tracked keyword sets on a fixed monthly budget, where cost per address matters more than cost per gigabyte
16. MobileProxy.Space — Mobile-only rental
2.5 / 5 for this job · from ₽490 / day
MobileProxy.Space takes the mobile-only route, and that is the argument: mobile addresses are the ones that tend to survive an engine that has already seen a thousand datacentre requests from the same subnet. Mobile plans bill by rental period with unlimited traffic rather than by the gigabyte, the minimum rental is one day, and rotation can be triggered on a timer or by hitting an API link. The provider quotes up to 250 simultaneous connections per mobile proxy, enough to keep a modest sweep moving.
The dashboard includes a proxy checker, port checks, a speed test and a browser proxy manager extension, and the API is joined by an MCP server for Claude Code, Cursor, Python, Node.js, PHP and cURL. Dead proxies are replaced automatically according to the provider, and unused time is refundable in its wording. Declared performance is modest: 348 ms average ping, 3.24 seconds response time and 10.00 Mbps download, with 4G described at about 30 Mb/s and up to about 4 Mb/s when the device is remote.
The catch is twofold. Third-party write-ups flag the prices as relatively high, and rank tracking is a volume job, so the daily rate compounds across a month of sweeps. And the numbers do not agree with each other: third-party records and the provider's own figures differ on pool size, country count, operator count and server count; the trial is given as 2 hours in one record and 3 by the provider; and some customers have reported location problems, laggy proxies and late-night maintenance. For city-level targeting, that disagreement is the thing to resolve before buying.
Key features
- Mobile plans billed per rental period with unlimited traffic
- Minimum rental of one day
- Rotation on a timer or via an API link
- Up to 250 simultaneous connections per mobile proxy, per the provider
- Dashboard with proxy checker, port checks and speed test
- 30+ countries in third-party records; 53 per the provider
Best suited for: short, high-value keyword sets where a mobile address is likelier to return a rank than a datacentre one
17. ChangeMyIP.com — Shared static testing
1.9 / 5 for this job · from $1
ChangeMyIP.com sits last because its cheap tiers are shared, and a shared address is close to the worst thing you can hand a rank tracker. The provider says outright that bandwidth is shared, that multiple users sit on the same IP and that performance can suffer on busy servers. When other people have already queried a search engine from that address, the block lands on you, and it arrives as a missing rank rather than an error you can see.
What it is good for is the plumbing. Shared HTTP and SOCKS5 are $1.00, dedicated HTTP is $1.80, SOCKS5 $2.20, and the protocol range is wider than most: HTTP, SOCKS5, ShadowSocks, Trojan, DNS proxy and Wireguard VPN, on 1Gb/s ports with unlimited bandwidth and no contract. If you are building the client that talks to a tracking tool and want to point it at something cheap before renting a metered pool, a dollar a proxy is a fair way to find out whether your parser works.
The catches add up. A $10 minimum order applies even when you want one $1.00 shared proxy. Torrents and paypal.com are blocked at proxy level on every product. The 30-day money-back guarantee does not apply once the service has been used, and refunds are limited to bugs or slow performance. The fully anonymous and undetectable wording is unverified marketing with no published independent test behind it. Nothing in this record is built for a scheduled sweep of thousands of keywords.
Key features
- Shared HTTP and SOCKS5 at $1.00; dedicated HTTP at $1.80
- Unlimited bandwidth on the plans it sells
- 1Gb/s ports, per the provider
- $10 minimum order
- Multiple users on the same IP on shared tiers
- Torrents and paypal.com blocked at proxy level
Best suited for: testing a rank-tracking client against a cheap static IP before committing to a metered pool
SEO and rank tracking proxies: questions we get
How many proxies do I need for rank tracking?
It is a concurrency question, not a count question. A sweep of 5,000 requests spread over an hour needs perhaps 20 to 40 simultaneous sessions drawing on a few hundred addresses so none of them carries too much. Buying 5,000 static addresses for 5,000 keywords is the wrong shape entirely. Ask a provider for concurrent sessions and requests per minute, not pool size.
Can I check Google rankings with free proxies?
You can, and it is a reasonable way to learn how the plumbing fits together. But public list entries are shared with thousands of people, most are dead or already burnt, and the inconsistency in your results will be impossible to separate from a genuine ranking change. Use them to test a script, not to produce a report.
Does using a proxy actually change the results I see?
Yes, because the exit location is one of the inputs behind the result page. But a proxy does not override personalisation: a signed-in session still carries history and preferences that no exit address removes. Run tracked checks from a clean session with no cookies and no account.
What is a SERP proxy?
The term covers two products. One is an ordinary proxy pool sold for collecting search results, where you write the request and the parser. The other is a parsed search endpoint, where a provider takes a query, location and device and returns structured JSON. The second costs more per search and removes the part that breaks most often.
Do I need residential or mobile proxies for SEO?
It depends on the target and the volume. Datacentre addresses are fine for light checking and for engines that block less aggressively. When Google is the target and the volume is real, residential addresses are the usual choice, and mobile survives longest because carriers put many real subscribers behind one address. Mobile is also the slowest and most expensive option.
Why do I keep getting a captcha or an 'unusual traffic' page?
Usually a rate problem on that address, often combined with a request that does not look like a browser: mismatched language headers, an unusual TLS fingerprint, no cookies. Retrying immediately from the same address makes it worse and extends the block. Rotate, slow down, and fix the headers before increasing volume.
How often should I track rankings?
Daily for the keywords that carry revenue, weekly for the long tail. More frequent collection rarely changes a decision and multiplies the bill, and it also amplifies noise: single-sample daily data looks volatile even when nothing has moved. Keep the time of day fixed either way.
Why do two rank tracking tools give different positions?
They are measuring different things. Location granularity, device class, sample size, whether SERP features are counted, how personalisation is suppressed and what time of day the sweep ran all move the number. Before assuming one tool is wrong, compare the exit city, the device and the timestamp for the same keyword.
Can I track mobile rankings through a proxy?
Yes, but mobile results are a separate series rather than a setting. Send a current mobile user agent, keep it fixed, and store it with the data; mixing mobile and desktop fetches into one trend line produces movement that is entirely an artefact of the device class. Some providers expose mobile device profiles you can select instead of setting them yourself.
Is scraping Google search results legal?
Scraping public search pages is not automatically unlawful, but it is almost always a breach of the search engine's terms of service, which is a contract question rather than a criminal one. In the United States, the hiQ v. LinkedIn line of cases made it hard to treat scraping of public pages as unauthorised access; in the UK and EU, database rights, computer misuse legislation and data protection law are the statutes that matter if you store personal data. Treat it as a commercial risk you accept knowingly, get advice for anything beyond internal rank monitoring, and remember that robots.txt is evidence of intent, not the boundary of legality. Buying results through a licensed endpoint moves the terms question to the provider.
Is Search Console's average position the same as a tracked rank?
No. Average position is an average across queries, locations and devices, calculated only from impressions you actually received, and it does not exist for queries where you never appeared. A tracked rank is one keyword, one city, one device, one moment. Both are useful, but they answer different questions and should not share an axis.
Other jobs people buy proxies for
- Web scraping
- Ad verification
- Price monitoring
- Multi-account management
- Antidetect browsers
Some links on this page are affiliate links: if you buy through them we may earn a commission, at no cost to you. It does not change the order above — see our monetisation policy.