Pick what matters to you, such as privacy, malware blocking, parental controls, speed, IPv6, or a specific jurisdiction, and the finder narrows 29 global public resolvers to the ones that fit. A full comparison table and research-backed decision notes follow.
Check what matters to you. Transport, DNSSEC, IPv6, jurisdiction and operator type are hard filters. The priorities are scored and ranked.
This measures DNS-over-HTTPS round-trip time from your browser to each DoH-capable resolver, so you can see which is fastest where you actually are. Plain-DNS-only resolvers cannot be tested this way. Results are a relative guide and include TLS and HTTP overhead, so run it a couple of times. Your browser queries each resolver directly, which reveals your IP address to them; nothing is sent anywhere else.
| # | Resolver | Median | Latency | Best |
|---|---|---|---|---|
| Press Run to benchmark from your location. | ||||
Technique inspired by the open-source DNS Speed Test by Silviu Stroe (GPL-3.0); this is an independent implementation. It runs only when this page is served over HTTPS. For resolvers not testable here, that dedicated tool benchmarks a wider DoH set.
Click a column header to sort. Search by name, operator, jurisdiction, or feature. Filter-variant addresses (malware, family, unfiltered) are listed in the Filtering cell.
| Resolver | Jurisdiction | Type | Primary IPs (v4 / v6) | Filtering (default and variants) | DNSSEC | Transports | Logging | ECS |
|---|
Findings from peer-reviewed DNS measurement studies that should shape the trade-offs above.
Encrypted transports (DoH and DoT) add latency per query, yet whole-page load times are often close to plain DNS, and DoH's overhead is small in practice. On lossy or high-latency links, plain Do53 still wins. Performance also varies by provider and region, so the fastest resolver depends on where you are.
Hounsel et al., WWW 2020; Böttger et al., IMC 2019; Chhabra et al., IMC 2021.
The largest end-to-end study of encrypted DNS found queries are far less likely to be intercepted or altered in transit than plain DNS, with only minor overhead. Operator quality varies, though: about 25% of DoT providers in that study served invalid TLS certificates, so favour well-run providers.
Lu et al., IMC 2019.
Whichever provider you choose still sees every domain you look up. If that worries you, prefer no-logging operators, or an oblivious design (ODoH) where a proxy separates your identity from your queries so no single party sees both. Cloudflare and Apple have deployed ODoH.
Schmitt, Edmundson & Feamster, PoPETS 2019; Singanamalla et al., 2021.
Only a validating resolver protects you from spoofed records. Google, Cloudflare and Quad9 all validate, and they handled the first root-key (KSK) rollover without breaking users. If integrity matters, treat DNSSEC validation as a must.
Müller et al., IMC 2019.
EDNS Client Subnet sends part of your IP to CDNs for better geo-routing. Google and OpenDNS send it for sharper CDN mapping; Cloudflare and standard Quad9 leave it off for privacy. Pick based on which you value more.
"A Look at the ECS Behavior of DNS Resolvers", IMC 2019.
The operator's legal home governs what can be compelled or logged, and a handful of providers now carry a large share of the world's recursive traffic. The U.S. NSA has also warned that external resolvers bypass internal DNS filtering and inspection, so weigh control against convenience.
Moura et al., IMC 2020; NSA guidance, 2021.
A 2022 measurement of DoQ found it already beats both DoT and DoH on response time, though about 40% of handshakes were slowed by QUIC's address-validation limit. Where your client and resolver both support it (Quad9, AdGuard, NextDNS, Control D, Mullvad, UncensoredDNS, and the Chinese majors here), DoQ is the encrypted option to prefer.
Kosek et al., PAM 2022.
DNSCrypt predates DoH, DoT, and DoQ (version 2 dates to 2013). It encrypts from the first packet using a resolver's pre-shared public key, so there is no plaintext hostname lookup and no dependency on certificate authorities, and its Anonymized DNS mode (2019) also hides client IPs. Among the resolvers here it is offered by Quad9, OpenDNS, AdGuard, NextDNS, Control D, and Yandex. Reliable usage numbers are scarce, though: population-scale measurements such as APNIC Labs track DoH and DoT but not DNSCrypt, so there is no trustworthy public figure for how many people use it.
DNSCrypt Project; APNIC Labs encrypted-DNS measurement.
Even over DoH, traffic analysis can identify the domains you visit with high accuracy, and the standard EDNS padding does not fully prevent it. If that threat model applies to you, pair encrypted DNS with Tor or an oblivious design rather than relying on padding.
Siby et al., NDSS 2020.
A 2023 study of Extended DNS Errors across major resolvers found they disagreed on diagnostic error reporting in 94% of test cases, with Cloudflare the most precise. Implementation quality and standards compliance differ between providers, which affects troubleshooting and reliability.
Nosyk, Korczyński & Duda, IMC 2023.
Not peer-reviewed, but authoritative and continuously updated. Useful for checking the live state of the resolver ecosystem.
Niche, hobby, community, or country-specific services that are not in the comparison above. Worth knowing about, but check their current status and policies before relying on them. The European entries are catalogued by European Alternatives. Resolvers based in heavily censored or sanctioned regions may enforce local content rules or, conversely, exist mainly to bypass geo-blocks, so treat those with extra care.