Domain Keep Rates: Why .com Holds and New TLDs Churn

Domain keep rate is the share of domains that get renewed rather than deleted when their registration comes up for a decision, calculated as renewals divided by renewals plus deletions. It's the single most honest number in the domain business because it ignores marketing and looks only at what registrants actually did. Across 562 measured top-level domains, the median keep rate is 68.7 percent. .com sits above that at 73.3 percent. Most high-volume new extensions sit far below it, some catastrophically so. That gap is the whole story of why some TLDs compound and others churn through millions of throwaway names.

What does "keep rate" actually measure, and what does it leave out?

The definition comes from nTLDData's public methodology, which HostList also documents on its own domain data methodology page. Renewals divided by renewals plus deletions, measured over the trailing twelve reported months, in this case June 2025 to May 2026, pulled from the monthly registry reports filed with ICANN, which typically run about three months behind. Data: nTLDData (ntlddata.com), CC-BY-4.0.

It is not a survival rate. A domain registered for three years only produces one renewal decision, at year three, so multi-year registrations get counted once, not annually. Transfers are excluded, auto-renewals are included, and deletions can lag actual expiry by a month or two because registrars run redemption grace periods. nTLDData only measures a TLD once it has at least 1,000 renewal decisions in the window, which is how 562 TLDs qualify out of the far larger number that technically exist. Ignore any TLD showing a 100 percent keep rate at the top of the raw table, those are almost always brand TLDs auto-renewing their own corporate names, not evidence of genuine registrant behaviour.

What is .com's renewal rate, and why does it dominate every other extension?

.com carries 166,686,934 domains in the zone, a keep rate of 73.3 percent across 134,121,243 renewal decisions, per nTLDData's renewals dataset, and it added 1,120,883 net domains in the most recent 30-day window alone. That single 30-day net figure is larger than the total zone size of most new TLDs. See the live numbers on nTLDData's .com page or HostList's own .com pricing and registrar comparison.

.com's keep rate beats the median of 68.7 percent and edges past the legacy gTLD benchmark of 73.2 percent, itself a high bar set by the other original gTLDs. Scale plus retention is what makes .com structurally different: it's not just that a huge number of people register .com, it's that most of the people who already hold one keep renewing it year after year. That's a compounding base no promotional pricing strategy can replicate.

How does the domain renewal rate compare across other legacy TLDs?

The other original gTLDs mostly track close to .com, some ahead of it. .net posts a keep rate of 76.2 percent across 12,403,676 domains, actually higher than .com. .org holds 69.4 percent across 12,149,073 domains. .biz manages 72.3 percent on a much smaller base of 1,289,862 domains. .info is the outlier among the legacy group at 59.1 percent across 5,191,136 domains, still well ahead of almost every high-volume new extension, but clearly the weakest of the original five.

What is .xyz's renewal rate, and what does it reveal about cheap, high-volume TLDs?

.xyz carries 10,224,648 domains and a keep rate of 16.9 percent. It also added 603,350 net domains in the most recent 30-day window, which tells you volume and retention are two entirely separate stories for this extension. A registry can post enormous headline growth while more than four in five domains that reach a renewal decision get deleted. One registrar, GMO's Onamae.com, holds 54.3 percent of all .xyz registrations, which shows how concentrated the acquisition channel is behind that volume.

Low first-year pricing to drive volume is a legitimate registry strategy, not something to accuse anyone of. But the retention data is what it is. Browse current TLD options and their registries on HostList's registries directory and see the full .xyz breakdown on the .xyz pricing page. For more on how a registry actually operates behind a high-volume string, see HostList's piece on Amazon's registry and its domain extensions.

Which high-volume new gTLDs have the worst renewal rates?

.xyz is not even close to the bottom. Several large extensions show keep rates that mean fewer than one in ten, sometimes fewer than one in fifty, names that reach a decision get renewed. A keep rate under 20 percent means, plainly, that more than four of every five names that reached a renewal decision were deleted. See the table below for the full comparison.

Which new gTLDs actually keep their domains, and what do they have in common?

Look at the top of that table again. .dev sits at 80.5 percent, genuinely higher than .com's 73.3 percent. .app holds 66.1 percent across 1,653,303 domains. .studio manages 69.1 percent, .page 59.4 percent, .agency 52.9 percent, .art 52.3 percent, .cloud 51.9 percent, .blog 48.0 percent, .tech 41.6 percent. Every one of these outperforms the median of 68.7 percent or comes close to it, and most beat the entire high-volume promo-driven cohort by a wide margin.

The pattern is not subtle. These are purpose-led strings that people actually deploy real sites and applications on. .dev and .app are Google Registry strings that require HTTPS by default, which filters out a chunk of the low-effort speculative registration that drags down .xyz or .top. See the full pricing and registrar picture for .dev and .app on HostList.

How do hosting companies and site builders decide which new extensions survive?

This is the part most domain commentary misses entirely. Registrar concentration data from the May 2026 registry reports shows exactly where these extensions get sold. Measured against the names held by each extension's 100 largest registrars, Hostinger holds 17.8 percent of .site, 11.3 percent of .shop, 10.4 percent of .online and 7.4 percent of .store. IONOS holds another 7.4 percent of .store. On the developer-focused side, Cloudflare holds 17.2 percent of .dev and 9.2 percent of .app, while Squarespace holds 14.6 percent of .dev. Compare that to .com, where GoDaddy's 30.9 percent share is large but no single hosting company gets anywhere near dominant control, because .com's buyer base is too broad for any one channel to own.

Nameserver attribution tells the same story from a different angle. Of 161,490,120 domains nTLDData can attribute to a provider by nameserver suffix, with 94,776,485 left unclassified, GoDaddy DNS accounts for 47,686,602, Cloudflare 39,239,235, Google Cloud DNS 10,073,670, Hostinger 7,266,410, Wix 5,669,637, IONOS 4,883,005, Bluehost 1,692,735, HostGator 1,000,852 and Vercel 835,618. See the raw figures on nTLDData's DNS attribution page. On the other side of the ledger, parking services account for a meaningful chunk too: Afternic shows 11,278,437 domains and Namecheap Parking 10,217,986, and given the unclassified pool, that's a lower bound, not the full picture.

Here's the mechanism. A domain sold inside a hosting or site-builder checkout gets a website attached to it on day one. A domain bought purely for a first-year promotional price and then parked or forgotten does not. A bundled or discounted first-year domain is a common offer across the hosting industry, and it's precisely why extensions like .dev and .app, pushed through hosting and cloud platform checkouts attached to a real deployment, hold on to their names, with .dev keeping at a higher rate than .com and .app not far behind, while extensions bought mainly on a first-year price get dropped once the discount ends. HostList has covered the retention mechanics behind this in more detail in Why Hosting Companies Obsess Over Churn.

Will .com always be number one?

Yes, on the evidence, and it's not close. .com's 73.3 percent keep rate on 134,121,243 renewal decisions means its enormous base is largely self-sustaining, and it still added 1,120,883 net domains in a single 30-day window on top of that. No new gTLD comes anywhere near that combination of scale and retention. The extensions that generate headlines for registration volume, .xyz, .top, .shop, are doing it through first-year pricing that a large share of buyers never renew. Volume without keep rate is a leaky bucket, and .com is not a leaky bucket. For a longer view on when waiting for a new extension actually makes sense for a business, see HostList's piece on New TLDs Are Coming: Should Your Business Wait?

What about .ai and .io, the two extensions with no published keep rate?

Both are country-code domains, which sit outside nTLDData's dataset entirely, since it covers generic top-level domains reported through ICANN. There's no comparable published keep rate for either, and anyone quoting one is guessing.

What is verifiable: .ai passed one million registrations at the start of January 2026, and Sherwood News reported it contributed an estimated 70 million US dollars to Anguilla's government revenue the previous year. .ai is also sold with a two-year minimum registration term, which is why HostList's .ai price comparison shows two-year totals rather than a single annual figure, so compare the full term when you shop.

For .io, the UK and Mauritius signed a treaty on 22 May 2025 transferring sovereignty over the Chagos Archipelago, and ICANN published an explanation of how the .io country code would be handled on 14 November 2024. Nobody should predict retirement from that, and HostList won't, but it's worth tracking. See current registrar pricing on the .io page.

My view, and it's an opinion, not a measured fact: .ai and .io will matter more over time precisely because nobody buys them on a promotional whim. They cost more than a typical gTLD, they're bought deliberately, and they're almost always bought to name a specific product or company. That's the opposite of a first-year-promo registration parked and forgotten. A high price paid on purpose is a stronger signal than a low price paid on impulse, even without a published keep rate to prove it.

What this means if you are actually buying a domain

Check the renewal price before the first-year price, every time, on every extension. A free domain with hosting or a heavily discounted first year is a completely normal industry offer, and there's nothing wrong with taking it, but know what you're committing to once that promotional period ends. Second, where you have a genuine choice of extension, prefer one whose keep rate suggests other registrants stick around, that's a real signal about how the string gets used in practice, not just how cheaply it gets sold. Third, use HostList's domain price table to compare registrar pricing across the extension you actually want, rather than defaulting to whichever one is discounted this month.

Frequently asked questions

What counts as a good domain keep rate?

Anything above the median of 68.7 percent across the 562 measured TLDs is strong, and anything close to or above the legacy gTLD benchmark of 73.2 percent puts an extension in the same retention class as .com, .net and .biz. Extensions like .dev at 80.5 percent or .studio at 69.1 percent clear that bar comfortably, while most high-volume promotional TLDs sit well under 20 percent.

Is keep rate the same as a domain survival rate?

No. Keep rate measures the outcome of a renewal decision, renewals divided by renewals plus deletions, over a trailing twelve-month window. It doesn't track an individual domain across its entire lifetime, and a multi-year registration only produces one decision, at expiry, not one every year. It also excludes transfers and includes auto-renewals, so it's a decision-outcome metric, not a survival curve.

Why do so many new gTLDs have such low renewal rates?

Because a large share of registrations on the cheapest, highest-volume extensions are bought on a steep first-year promotional price with no website ever built on them. Once that price reverts to the normal renewal rate, the registrant has no reason to keep paying for a name with nothing behind it. Extensions bought to build an actual product, app or site, like .dev and .app, show far higher keep rates because there's something worth renewing.

Should I register a .ai or .io domain for a startup?

Both are ccTLDs with no published keep rate, so there's no retention data to lean on either way. What you can rely on is the pattern: buyers who choose .ai or .io are almost always naming a specific product deliberately, at a higher price than a typical gTLD, which is a different and generally stronger commitment signal than a discounted first-year gTLD registration. Just budget for .ai's two-year minimum term and compare full-term pricing before you commit.