Which DNS blocks ads? A test of public DNS services
We tested 12 DNS services against 1,880 ad and tracker addresses that 429 popular websites in 23 countries actually loaded. Cloudflare 1.1.1.1, Google Public DNS, Quad9 and OpenDNS blocked 0.1% or less of them, and their family versions did not block ads either. AdGuard DNS blocked 87.1%, Mullvad 66.3%, EZOBLOCKER 96.5%, Control D 97.8% and dnsforge.de 98.5%. The score weighs what each service blocks against the addresses sites need that it blocks by mistake.
Results, sorted by score
12 DNS services, 1,880 ad and tracker addresses from 429 websites in 23 countries, measured October 7, 2026. EZOBLOCKER is measured the same way, with its default categories (Security, Trackers, Ads).
| DNS service | Score | Catch | Precision | Ad requests blocked | Wrong blocks | Unlisted trackers blocked |
|---|---|---|---|---|---|---|
| 1EZOBLOCKER | 98.2 | 96.5% | 99.8% | 95.0% | 3 | 22 |
| 2Control D (ads & tracking) | 98.0 | 97.8% | 98.2% | 93.3% | 34 | 34 |
| 3dnsforge.de | 97.6 | 98.5% | 96.8% | 98.4% | 62 | 82 |
| 4AdGuard DNSIn the app | 93.0 | 87.1% | 99.8% | 88.3% | 4 | 13 |
| 5AdGuard DNS FamilyIn the app | 92.6 | 87.1% | 98.8% | 88.3% | 20 | 15 |
| 6Mullvad DNS (adblock) | 79.6 | 66.3% | 99.7% | 80.7% | 4 | 12 |
| 7Cloudflare for FamiliesIn the app | 0.3 | 0.2% | – | 0.1% | 2 | 1 |
| 8OpenDNS FamilyShieldIn the app | 0.2 | 0.1% | – | 0.1% | 0 | 0 |
| 9Cloudflare 1.1.1.1In the app | 0.1 | 0.1% | – | 0.1% | 0 | 0 |
| 10Google Public DNSIn the app | 0.1 | 0.1% | – | 0.1% | 0 | 0 |
| 11Quad9In the app | 0.1 | 0.1% | – | 0.1% | 0 | 0 |
| 12OpenDNSIn the app | 0.1 | 0.1% | – | 0.1% | 0 | 0 |
Catch: share of ad and tracker addresses blocked. Precision: share of blocked addresses that are ads or trackers (not shown for services that block almost nothing). Ad requests blocked: share of the requests sites made to ad and tracker addresses. Wrong blocks: addresses sites need that were blocked. Unlisted trackers blocked: addresses on no list that we identified as ads or trackers. In the app: selectable in EZOBLOCKER under Advanced, DNS provider.
Quick answers
-
Does Cloudflare 1.1.1.1 block ads?
No. In our test Cloudflare 1.1.1.1 blocked 0.1% of 1,880 ad and tracker addresses. It is a fast, private resolver, not a filter. Google Public DNS (0.1%), Quad9 (0.1%) and OpenDNS (0.1%) behave the same way; Quad9 blocks malware, not ads.
-
Does AdGuard DNS (94.140.14.14) block ads?
Yes. AdGuard DNS blocked 87.1% of the ad and tracker addresses in our test, and 4 addresses that sites need. Its family version (94.140.14.15) blocked the same share of ads and also blocks adult sites.
-
Do family DNS services like 1.1.1.3 block ads?
Mostly no. Cloudflare for Families (1.1.1.3) blocked 0.2% and OpenDNS FamilyShield 0.1% of the ad and tracker addresses; they block adult and harmful sites. AdGuard DNS Family is the exception among the family services we tested: it blocked 87.1%.
-
Which DNS blocks the most ads?
In this test the highest shares were dnsforge.de (98.5%), Control D's ads and tracking filter (97.8%) and EZOBLOCKER (96.5%). They also blocked different numbers of addresses that sites need: dnsforge.de 62, Control D 34, EZOBLOCKER 3. The score in the table weighs both.
How the score works
- Catch: the share of the 1,880 ad and tracker addresses a service blocks. An address counts as ad or tracker when at least two of 5 public lists contain it.
- Precision: of the addresses a service blocks, the share that are ads or trackers. Blocking an address a site needs to work (its images, video, comments, cookie consent window or checkout) lowers precision.
- Score: the balance of the two, the F1 score, out of 100. A service that blocks nothing and one that blocks everything both score low; a high score needs a high catch and few wrong blocks.
- Not counted either way: addresses on only one list, and addresses on no list that turned out to be trackers the lists miss. The last column shows those.
What each service blocked by mistake
238 addresses were on no list but blocked by at least one service. We sorted them by hand: 102 are needed by the site (images, video, comments, cookie consent windows, checkout), the rest are trackers the lists miss or unclear. Below are the needed ones each service blocked, most used first.
-
EZOBLOCKER3 wrong blocks
imgserve.rambler.ru920201.rufio.hls.live-video.netimgserve-uat.rambler.ru
-
Control D (ads & tracking)34 wrong blocks
sdk.privacy-center.orgcookie-cdn.cookiepro.comstatic.vidgyor.comcdn.opencmp.netbuy.tinypass.comprivacy-cs.mail.ru
-
dnsforge.de62 wrong blocks
cookie-cdn.cookiepro.comapi-2-0.spot.imcdnjs.atresmedia.comcdn.iubenda.combuy.tinypass.comcdnstatic.detik.com
-
AdGuard DNS4 wrong blocks
imgserve.rambler.ru920201.rufio.hls.live-video.netagmipnyy.captain.vpplayer.netimgserve-uat.rambler.ru
-
AdGuard DNS Family20 wrong blocks
yandex.ruwww.pomponik.plstatic.feber.sewww.vagalume.com.bryandex.comfeber.se
-
Mullvad DNS (adblock)4 wrong blocks
imgserve.rambler.ru920201.rufio.hls.live-video.netstatic.addtoany.comimgserve-uat.rambler.ru
-
Cloudflare for Families2 wrong blocks
st.vkvideo.rum.vidio.com
A blocked address does not always mean a visibly broken page; it means a part the site loads did not arrive. Family services block adult sites on purpose, and some of their wrong blocks are such sites.
Method
- Sites: the same 429 popular news, sports and entertainment websites in 23 countries as our speed test. Each was opened in Chrome for 20 seconds with scrolling and unfiltered DNS, and every address it contacted was recorded: 4,787 addresses.
- Labels: an address counts as ad or tracker when at least two of these 5 public lists contain it: HaGeZi Multi PRO, AdGuard DNS filter, EasyList + EasyPrivacy, Peter Lowe's list, OISD big. Requiring two lists keeps any single list, and any service built on it, from setting the answer.
- Queries: on October 7, 2026 every address was sent to every service over DNS over HTTPS. It counts as blocked when the service answers 0.0.0.0, a block page address, "does not exist" or a "filtered" marker while unfiltered DNS returns a real address. Addresses that no longer exist were left out.
- Wrong blocks: the 238 addresses on no list that at least one service blocked were sorted into "needed by the site" and "ad or tracker", first automatically, then checked one by one by hand.
- EZOBLOCKER: measured through its own servers with a test device and the default categories, the same setup a new user gets.
Raw data
Every number on this page comes from these files. The address table lists every address with the lists that contain it and the answer of every service, so anyone can recount the results.
-
dns-test-results.csvOne row per service: score, catch, precision, wrong blocks. -
dns-test-results.jsonThe same results with the test summary and list licenses. -
dns-test-hosts.csvEvery address: lists that contain it, group, requests, and 1 or 0 for each service.
Data license: CC BY 4.0. The raw measurement, scripts and copies of the lists are kept unchanged for each run.
Questions people ask
Is Google DNS (8.8.8.8) an ad blocker?
No. Google Public DNS answers every address and blocked 0.1% of the ad and tracker addresses in our test. To block ads with DNS you need a filtering service.
Can I use these DNS services in EZOBLOCKER?
Yes. Under Advanced, then DNS provider, you can pick Cloudflare, Cloudflare for Families, Google, OpenDNS, OpenDNS FamilyShield, Quad9, AdGuard, AdGuard Family or NextDNS, or enter a custom address such as Mullvad, Control D or dnsforge.de. While another provider is selected, EZOBLOCKER's own categories are off.
Does a DNS ad blocker break websites?
Sometimes. DNS only sees addresses, so when a service blocks an address a page needs, that part of the page does not load. The Wrong blocks column counts this for each service, and the examples are listed on this page.
Can DNS block YouTube ads?
No. Video ads inside YouTube come from the same addresses as the videos, so no DNS service in this test can block them without blocking the videos.
Is this test repeated?
The date at the top is the date of the measurement. Each new run is published with its own date, raw data and copies of the lists, and the old runs stay available.