How to Test Proxy Quality: A Simple Four-Step Method

To test proxy quality properly, check four things in order: does it connect, is it really in the location you chose, does it return the right content from your own target pages, and does it stay that way over time. Most people stop after the first check — “it connected, so it works” — and only discover later that half their data came from the wrong country. This guide walks through all four steps with a script you can run today.

If you are comparing several providers, this method pairs with our checklist on how to choose a proxy provider.

Why “it connects” is not enough

Imagine hiring a courier and only checking that they picked up the parcel. You also need to know it arrived at the right address, undamaged, and that they do it reliably every day. Testing a proxy is the same:

Step Question What goes wrong if you skip it
1. Connection Does the request get through? You cannot start at all
2. Location Is the exit really where it says? You collect another country’s data
3. Content Is the page complete and correct? “Successful” requests return check pages or partial content
4. Stability Does it hold up over hours and days? Works in the trial, fails in production

Before you start: pick your test pages

The most important decision is what you test against. Use your own real target pages — 20 to 50 of them — not a generic speed-test site. A proxy that loads a test site perfectly can still behave differently on the pages you actually care about.

Pick a mix: a few simple pages, a few search or listing pages, and a few detail pages. They often behave differently.

Also write down, for each page, one thing you expect to see if the content is correct — for example the currency symbol “€” for a German store, or a product name. You will use this in step 3.

Step 1: Does it connect?

Send a request through the proxy to a simple page that returns your outgoing address. If you get an address back, the connection and login work.

curl -x http://user:pass@proxy-host:port https://api.ipify.org

If this fails, check the basics first: host, port, username, password, and whether you need http:// or socks5h:// (see HTTP vs SOCKS5). An error code 407 means the login was refused.

Step 2: Is the location real?

Do not rely on a single IP lookup website. Different location databases often disagree, especially for recently assigned addresses. Use two checks instead:

  1. Look the address up in two or three different databases. If they all agree, good. If they disagree, step 3 decides.
  2. Check what your target site thinks. Open a page that shows location-dependent information — currency, language, delivery country. That is the location that actually matters for your task.

Also make sure your own settings match the location: browser language and time zone. A German exit with a browser asking for Chinese pages can produce a page no real German visitor would see.

Step 3: Is the content right?

This is the step that separates a useful proxy from a merely working one. For each test page, check:

  • Status code — 200 is a start, not proof.
  • Your expected marker — is the “€” or product name you noted actually in the page?
  • Page size — a page much smaller than usual is often a check page or an error page with a 200 status.
  • A look by eye at a few saved pages — nothing beats opening them.

Step 4: Does it stay stable?

A single test run is a snapshot. Run the same test:

  • At different times of day — morning, afternoon, evening in the target market.
  • On at least two or three different days.
  • At the speed you will really use — a gentle test does not tell you how things go at your real request rate (within each site’s stated limits).

For a static proxy, also confirm the address is the same each time. For a rotating proxy, confirm it does change, and that a sticky session holds as long as promised.

A test script you can run

This Python script runs steps 1–3 against your list of pages and saves the results to a spreadsheet file. Put your target URLs and expected markers in a file called targets.csv with two columns: url,marker.

import csv, time, requests

PROXY = "http://user:pass@proxy-host:port"
proxies = {"http": PROXY, "https": PROXY}

exit_ip = requests.get("https://api.ipify.org", proxies=proxies, timeout=30).text
print("Exit address:", exit_ip)

with open("targets.csv") as f_in, open("results.csv", "w", newline="") as f_out:
    w = csv.writer(f_out)
    w.writerow(["time", "url", "status", "kb", "seconds", "marker_found"])
    for row in csv.DictReader(f_in):
        start = time.time()
        try:
            r = requests.get(row["url"], proxies=proxies, timeout=30)
            found = row["marker"] in r.text
            w.writerow([time.strftime("%H:%M"), row["url"], r.status_code,
                        round(len(r.content) / 1024), round(time.time() - start, 1), found])
        except Exception as e:
            w.writerow([time.strftime("%H:%M"), row["url"], "error",
                        0, round(time.time() - start, 1), False])
        time.sleep(2)  # keep a polite pace

Run it a few times across different days (step 4), keeping each results.csv under a new name.

How to read the results

The number that matters most is how many pages came back with the marker found — that is your share of useful results. Then look at:

  • Errors vs wrong content: errors are visible; wrong content is silent and more dangerous.
  • Which pages fail: if only search pages fail, that is a different problem from everything failing.
  • Time of day: results that drop at certain hours tell you when to schedule work.
  • Speed spread: a few very slow pages can matter more than the average.

When comparing two proxies, compare useful results per unit of cost, not just the raw success share. Our guide on proxy pricing models explains how to work out cost for your usage.

Common testing mistakes

  • Testing on a speed-test site instead of your own targets.
  • Counting status 200 as success without checking content.
  • Testing once and treating it as the long-term result.
  • Comparing providers on different days or with different pages — test them side by side.
  • Trusting one IP lookup site for location.
  • Mismatched browser language or time zone for the chosen location.

What a test cannot tell you

  • It cannot predict the future perfectly. Websites change; re-test occasionally, especially if results drift.
  • Your result applies to your targets. It does not transfer to other sites — and nor does any provider’s published success figure, including ours.
  • A good result does not change a site’s rules. Test only on publicly visible pages and your own assets, at a reasonable pace.

If you are testing YiluProxy, the same script works with all five lines — rotating residential, static residential, rotating datacenter, static datacenter and mobile — over HTTP or SOCKS5. Running the cheaper datacenter line through it first often shows whether you need residential at all.

FAQ

Q1. How many pages should I test?

A. Usually 20 to 50 real target pages is enough to see a pattern, as long as they cover the different page types you will use.

Q2. How long should a proxy test run?

A. Spread it over at least two or three days and several times of day. One run is only a snapshot.

Q3. Why do IP location sites disagree?

A. They use different databases, updated at different times. Your target site’s own view — currency, language, delivery country — is the check that matters.

Q4. What if two proxies score the same?

A. Compare cost per useful result, speed spread, and how support handled a real question during the trial.


Test first, then buy

Run the four steps on your own target pages. New users can apply for a free trial of rotating residential, mobile or datacenter traffic, or shared datacenter addresses, through our client.

Start a free trial

(This article describes a general method for testing proxies. It contains no pool-size, success-rate or performance figures, as those vary by provider, target and time. Written October 2026.)

Scroll to Top