Historical evidence note: The seven source snapshots and extraction outputs have not been reproduced for this update. Record totals and tables below are retained as reported historical measurements. Coverage changes prevent deriving an Internet adoption rate, a country ranking or a count of unique IPv6 hosts from these figures alone.
The original analysis examined seven Project Sonar FDNS snapshots from November 2017 through April 2020. Reported AAAA-record counts rose from 23.5 million to 219.7 million, a 9.35x change in this corpus. That is a change in observed records, not a like-for-like measure of global deployment.
How I Measured This
The data comes from Rapid7's Project Sonar, which conducted internet-wide forward DNS scans and published the results as open datasets. Sonar queried known hostnames and recorded the responses, including AAAA records, the DNS record type that maps a hostname to an IPv6 address. Where an A record points to an IPv4 address like 203.0.113.50, a AAAA record points to an IPv6 address like 2001:0db8:85a3::8a2e:0370:7334.
I processed 7 snapshots between November 2017 and April 2020, extracting every AAAA record, parsing the associated domain to identify its TLD, and aggregating counts by TLD and IPv6 prefix. The 2017-11-10 snapshot serves as my baseline, with each subsequent snapshot measured against it.
A few important caveats. Project Sonar's scanning methodology evolved over time, the jump between the November 2018 and April 2020 snapshots partly reflects expanded scan coverage, not just organic AAAA record growth. The 2017 snapshot is the cleanest for TLD-level comparison because the methodology was most consistent relative to the number of hostnames queried. I'll use 2017 data for TLD breakdowns and the full timeline for trend analysis.
Also, a AAAA record existing doesn't mean the IPv6 address is actively serving traffic. Some records are created automatically by DNS providers, some point to infrastructure that may or may not be reachable over IPv6. What AAAA records do represent is IPv6 readiness at the DNS layer, someone (or something) decided this hostname should be reachable over IPv6.
The Growth Trajectory
Here's the raw timeline across all 7 snapshots:
| Date | Total AAAA Records | Growth vs Previous |
|---|---|---|
| 2017-11-10 | 23,475,569 | , |
| 2018-01-27 | 26,623,767 | +13.4% |
| 2018-04-13 | 27,518,751 | +3.4% |
| 2018-06-22 | 47,146,819 | +71.3% |
| 2018-09-08 | 54,873,748 | +16.4% |
| 2018-11-24 | 79,716,110 | +45.3% |
| 2020-04-25 | 219,682,526 | +175.6% |
The standout is the June 2018 jump, a 71.3% increase in just two months. Something systemic changed between April and June 2018, and I'll get to what that was in a moment.
The reported 9.35x change combines observations collected with changing scan coverage. Even the 3–13% movements between early snapshots require comparable target lists and extraction rules before being attributed to new IPv6 deployment.
For comparison, IPv4 A records in the same Sonar dataset were relatively flat over this period. The DNS namespace wasn't growing 9x, IPv6 records were being added to existing hostnames that previously only had A records.
Which TLDs Lead the Transition
The 2017 baseline snapshot gives the cleanest TLD-level picture. Here are the top 15 TLDs by AAAA record count:
| Rank | TLD | AAAA Records | Share of Total |
|---|---|---|---|
| 1 | .com | 10,715,245 | 45.6% |
| 2 | .de | 2,400,114 | 10.2% |
| 3 | .net | 2,011,168 | 8.6% |
| 4 | .org | 919,371 | 3.9% |
| 5 | .ru | 834,744 | 3.6% |
| 6 | .info | 538,226 | 2.3% |
| 7 | .nl | 351,894 | 1.5% |
| 8 | .uk | 209,457 | 0.9% |
| 9 | .pl | 195,488 | 0.8% |
| 10 | .cz | 171,482 | 0.7% |
| 11 | .eu | 166,770 | 0.7% |
| 12 | .fr | 131,497 | 0.6% |
| 13 | .io | 108,271 | 0.5% |
| 14 | .ie | 104,377 | 0.4% |
| 15 | .au | 102,360 | 0.4% |
The .com dominance at 45.6% is expected, it's by far the largest TLD. What's revealing is what comes next.
The .de row accounts for 10.2% of AAAA records in the reported baseline table. Its 2.4 million records exceed the .net row's 8.6% share in that corpus. Without consistent domain denominators, this is not the proportion of German domains using IPv6 or a ranking of country adoption.
The Czech Republic (.cz) and Ireland (.ie) are punching well above their weight. Neither is a large ccTLD by registration count, yet both crack the top 15 for AAAA records globally. The Czech Republic's strong showing traces to CZ.NIC, the .cz registry operator, which has been an aggressive IPv6 advocate. They not only enabled IPv6 on their own infrastructure early but actively encouraged registrars to support it. Ireland's .ie presence likely reflects the concentration of major tech companies (many of which are IPv6-forward) hosting under the Irish ccTLD.
The presence of .io at position 13 is interesting for a different reason. The .io TLD is disproportionately used by tech companies and developer-focused products, exactly the demographic most likely to deploy IPv6 early.
You can inspect AAAA records for any domain using the DNS Inspector tool, just select the AAAA record type to see whether a domain has IPv6 addresses configured.
The Cloudflare Effect
The original provider attribution reported 1,176,091 Cloudflare-associated AAAA records in April 2020, compared with a near-zero baseline. The provider-matching rules and input outputs need to be recovered before that classification can be independently checked.
That figure does not explain the full increase to 219.7 million records or establish the cause of the June 2018 71.3% jump. Provider automation is a possible contributor, but this dataset does not isolate it from changes in coverage or other deployments.
Regional Patterns
The IPv6 prefix distribution from the 2017 data reveals clear regional patterns:
| IPv6 Prefix | Records | Region |
|---|---|---|
| 2001::/16 | 108,522 | Global (common early allocation) |
| 2a00::/12 | 17,457 | RIPE NCC (Europe, Middle East, Central Asia) |
| 2600::/12 | 3,300 | ARIN (North America) |
| 2400::/12 | 988 | APNIC (Asia-Pacific) |
Within the three /12 blocks sampled here, the RIPE block held 5.3x as many AAAA records as the ARIN block and 17.7x as many as the APNIC block. That is a comparison of record counts inside one selected /12 per registry, not of prefixes and not of each registry's full allocation. IANA has assigned each RIR further ranges, including 2003::/18 to RIPE NCC and 2610::/23 and 2620::/23 to ARIN, so these figures indicate direction rather than total regional deployment. Neither this partial-prefix comparison nor the TLD corpus share establishes a regional adoption leaderboard.
Why Europe? Several reinforcing factors are at play.
RIR exhaustion has several stages. APNIC's chronology places its final-/8 phase in April 2011, before RIPE NCC's corresponding 2012 transition. Exhaustion of general allocation and later depletion of a final pool are different events. Neither chronology establishes what caused changes in this DNS corpus.
European registries were early IPv6 advocates. CZ.NIC (.cz), DENIC (.de), SIDN (.nl), and IE Domain Registry (.ie) all deployed IPv6 on their registry nameservers and actively promoted IPv6 compatibility among their registrar communities. When the registry infrastructure speaks IPv6, it creates a natural pull for downstream domains to follow.
Government mandates in Europe favored IPv6. Multiple European governments issued guidance or mandates for IPv6 deployment in government networks during the 2012-2016 period. While these mandates didn't directly affect commercial domains, they created a pool of IPv6-capable infrastructure and expertise that spilled over into the broader hosting ecosystem.
The relatively low APNIC number in 2017 is worth noting because Asia-Pacific, particularly India and Japan, has since become one of the fastest-growing IPv6 regions. India's massive mobile carrier Reliance Jio launched as an IPv6-only network in 2016, and by 2020 India had one of the highest IPv6 traffic percentages in the world. This mobile-driven adoption may not show up as strongly in AAAA records for traditional domain names, since mobile apps often use different resolution paths than browser-based domain lookups.
What's Driving Adoption
Provider defaults, address availability and application requirements can affect deployment decisions. The snapshot comparison does not measure their individual contribution. A causal claim would require stable scan coverage, dated provider changes and a reproducible attribution method.
The email-authentication research raises a similar distinction between records present in a corpus and policies adopted by a stable set of domains.
The Dual-Stack Reality
One pattern that's clear from the Sonar data is that the IPv6 transition is almost entirely additive. The overwhelming majority of hostnames with AAAA records also have A records. Pure IPv6-only deployments: where a hostname has a AAAA record but no corresponding A record, remain vanishingly rare in the public DNS.
This is the dual-stack approach in practice: keep IPv4 working for compatibility while adding IPv6 as an additional path. From a DNS perspective, it means a hostname resolves to both an IPv4 and an IPv6 address, and the client's operating system and network stack decide which to use (typically preferring IPv6 when available, per Happy Eyeballs / RFC 8305).
The dual-stack strategy is pragmatic but creates a long tail problem. As long as every IPv6-capable hostname also has an IPv4 fallback, there's no forcing function to complete the transition. IPv4 never becomes unreachable; it just becomes the slower, more congested, more expensive path. The transition completes through economics and friction rather than through a hard cutoff.
A dual-stack hostname publishes both A and AAAA records, but the corpus does not establish that every observed name is dual-stack. The 219.7 million AAAA records are not necessarily 219.7 million unique hostnames. Deduplication and a join to corresponding A records are required.
What This Means for DNS Operators
If you're managing DNS infrastructure, the AAAA growth trajectory has practical implications.
Zone file sizes are growing. Every AAAA record added to a zone increases the zone file size. AAAA RDATA occupies 16 octets on the wire versus 4 for an A record. In zone files the textual form is variable, since RFC 5952 requires leading zeros to be suppressed and :: to be used to its maximum extent, so addresses are normally written compressed rather than fully expanded. For large zones with millions of records, the growth in AAAA records measurably increases zone transfer times, memory requirements for authoritative servers, and the time to propagate changes.
DNS response sizes increase. When a client ends up holding both A and AAAA records for a name, typically by issuing separate A and AAAA queries as dual-stack clients do, the combined response data is larger. Note that ANY queries are no longer a reliable way to retrieve both: RFC 8482 permits a responder to return just one RRset, a subset, or a synthesized HINFO instead. This matters for UDP-based DNS because larger responses are more likely to exceed the 512-byte UDP limit, triggering TCP fallback or EDNS0 negotiation. The practical effect is slightly higher resolver load and marginally increased latency for dual-stack domains.
AAAA monitoring is now essential. If your domain has AAAA records, even if you didn't add them intentionally (your CDN or hosting provider may have). You need to monitor them. An AAAA record pointing to a decommissioned IPv6 address creates a partial outage that only affects IPv6-capable clients, which is an increasingly large subset of users. Stale records that resolve to deprovisioned infrastructure carry a security risk too, the same class of problem behind stale DNS records and subdomain takeover. These are notoriously difficult to debug because the site works fine for IPv4 clients, and most monitoring tools default to IPv4.
IPv6 glue records matter. If you run authoritative nameservers, their AAAA glue records in the parent zone are increasingly important. Some resolvers now prefer IPv6 when contacting authoritative servers, especially in regions with strong IPv6 deployment (Europe, parts of Asia). If your nameserver has a stale or missing AAAA glue record, you may see intermittent resolution failures that are difficult to diagnose.
The historical corpus shows substantial growth in observed AAAA records. It cannot establish that the same growth continued after April 2020 or that all new cloud resources receive IPv6 by default. For an operational check, verify the chosen provider, network and service configuration directly.