A Country Is Not a Blocklist: Reading Public IP Range Pages on IP99
If you sort IP99's public country directory by IPv4 segments, China comes first. If you sort the same table by address count, the United States does. That split is the point of the page: it is an inventory of how the geo
If you sort IP99's public country directory by IPv4 segments, China comes first. If you sort the same table by address count, the United States does. That split is the point of the page: it is an inventory of how the geolocation database is carved up, not a reason to treat a country as risky or safe.
The index covers 244 countries and territories. Each country page โ for example China, the United States, Singapore, or the Netherlands โ shows IPv4 segment and address counts, a usage-type mix, region and city breakdowns, the ASNs that carry the most segments, and how many proxy-type addresses were newly captured in the last seven days. There is no country-page JSON API (a /en/geo/us.json-style URL returns 404). For a single address, use the free lookup instead:
curl https://ip99.com/v1/ip/8.8.8.8
Figures below are from the pages as of data version 20261121 (the United States page footer was generated 2026-11-21 12:13 UTC). Official docs: when you cite this data, name ip99.com as the source and keep the data version and the statistics window.
More segments are not more addresses
On today's pages:
| Country | IPv4 segments | IP addresses |
|---|---|---|
| China | 20,879,685 | 414,896,966 |
| United States | 1,010,691 | 1,531,299,328 |
| Netherlands | 70,407 | 54,790,678 |
| Singapore | 16,770 | 23,188,449 |
China is listed with about 20.7 times as many IPv4 segments as the United States (20,879,685 รท 1,010,691), but with fewer addresses: 414,896,966 versus 1,531,299,328. Dividing the official counts gives roughly 20 addresses per Chinese segment and about 1,515 addresses per United States segment. The directory is counting how the database is split, not how many people live in a country, and not how much "risk" that country holds.
A segment on these pages is a geolocation-database fragment, not a routing-table entry you can treat as a ban unit. Smaller fragments just mean the inventory is more finely divided.
A usage mix is composition, not a verdict
Each country page publishes a usage-type mix โ residential broadband, datacenter, gateway/NAT egress, mobile, education, CDN, and others. Official page copy: the ratio of residential broadband to datacenter address space is the most direct indicator of what kind of proxy risk a region carries. A high datacenter share means datacenter proxies are more likely; a high residential share means home-broadband proxies and dial-up pools are. That is a statement about composition, not a license to block a country.
Today's published shares:
| Country | Residential broadband | Datacenter | Gateway / NAT egress |
|---|---|---|---|
| China | 36.1% | 25.5% | 28.6% |
| United States | 37.4% | 27.9% | 15.2% |
| Singapore | 34.0% | 31.7% | 26.0% |
| Netherlands | 36.3% | 35.6% | 6.2% |
Singapore and the Netherlands sit closer to an even residential-versus-datacenter split than the United States. Singapore's top ASNs on the same page include AS15169, AS8075, AS32934, AS16509, and AS13335 alongside local operators โ cloud and CDN space that serves users far beyond Singapore. A high datacenter share is a reason to look up the specific exit, not to deny every login with geo.country: "SG".
The same caution applies to cities. California accounts for 15.1% of United States segments; New York City 8.0%. Amsterdam accounts for 28.2% of Netherlands segments. Those concentrations include both access networks and hosting districts. Official API guidance already says not to block on hosting or mobile alone, and not to treat geo as a street address or a person's location.
"Newly captured" is a window, not a stock
The country pages also print a seven-day proxy-type count. In the 7 days to 21 November 2026 the pages report 2,745,728 newly captured proxy-type addresses in China and 349,779 in the United States. Official docs (llms.txt): that figure is how many proxy-type addresses first appeared in the risk library during the window; the same address is counted once. It is not the inventory of all proxy exits, and it is not a push count. Always keep the window when you quote it.
The weekly aggregate of the same feed lives on this week's proxy exits โ totals only, no address lists. Do not turn a large "newly captured" number into "this country is high risk." A fragmented address space and a large access network will surface more first-seen exits without telling you what to do with the next login.
The ASN list is still not a blocklist
Each country page lists the autonomous systems with the most segments. China: AS4134 (8,649,119), AS4837 (4,367,336), AS9808 (910,050). United States: AS7922 (153,879), AS7018 (135,525), AS701 (70,521). Operator names are not published on the country page โ official copy: the machine-readable identity is the ASN. The entries link through to ASN pages when you want the usage mix for a single network.
Addresses under one ASN can share operational traits. That does not mean the right control is to ban the whole ASN. Mixed residential networks and hosting networks need different handling, and official scoring is still per IP: start with evidence_state, then risk.score / risk.level, then the signals. risk.score is the freshness and strength of verifiable evidence, not the probability that someone is an attacker. A missing score is not a clean bill of health.
Look up the address, not the country
The United States inventory includes well-known anycast resolvers. A live lookup today (computed_at 2026-11-21T13:09:09Z):
curl https://ip99.com/v1/ip/8.8.8.8
{
"computed_at": "2026-11-21T13:09:09Z",
"evidence_state": "none",
"geo": {
"city": "Mountain View",
"continent": "NA",
"country": "US",
"district": "",
"lat": 37.38605,
"lon": -122.08385,
"region": "California",
"tz": "America/Los_Angeles"
},
"ip": "8.8.8.8",
"network": { "asn": 15169, "usage_type": "IDC" },
"risk": { "level": "none", "score": 0, "signals": ["hosting"] }
}
Mountain View, California, AS15169, usage_type: "IDC", hosting, score 0 with evidence_state: "none". That is one block's geolocation and a network attribute, not "a United States user," and not a reason to mark the address safe. The same day, 1.1.1.1 (computed_at 2026-11-21T13:09:11Z) geolocated to Brisbane, Australia, AS13335, IDC, hosting, score 0 / none โ so a country inventory can hold infrastructure that is anycast or registered far from most of its users.
If you need a per-address decision, call /v1/ip. Use the country page to understand how space is split; use the lookup to decide whether this exit has current evidence. Step up a check when the signals warrant it; do not auto-ban a country, a city, or an ASN from a directory row.
Try it
curl https://ip99.com/v1/ip/8.8.8.8
curl https://ip99.com/v1/ip/1.1.1.1
curl https://ip99.com/v1/me
Browse the directory at https://ip99.com/en/geo/ and a country page such as https://ip99.com/en/geo/us. Chinese-language provincial pages for China are at https://ip99.com/geo/cn/. ASN mix pages: https://ip99.com/en/asn/. The lookup API is documented at https://ip99.com/api. Anonymous IP lookup is 500/day per client IP (60/min); a free key raises that to 5,000/day. Check remaining quota with GET https://ip99.com/v1/me. Pricing: https://ip99.com/en/pricing.
Do not treat IP99 scores or geo as identical to other ThreatHunter datasets. This article only claims what the public country pages, llms.txt, and the free /v1/ip responses documented on the dates above.
IP99 is a ThreatHunter product. A country page is an inventory of address space, not a verdict on a person. Start at https://ip99.com.
Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes โ full credit and traffic to the original publisher.