IPure · Data report · 2026 Q4

IPure IP Purity Report 2026 Q4
English summary

Name: IPure IP Purity Report 2026 Q4 — English summary
Creator: IPure
License: https://creativecommons.org/licenses/by/4.0/

We aggregated 56,688 real IP checks to answer a question people usually argue about from experience: which kinds of IPs — mobile, residential, business, data center, native or broadcast — are actually cleaner?

Sample: 56,688 checks
Snapshot: 2026-10-23 13:02 UTC+8
Coverage: IPv4 / IPv6
License: CC BY 4.0

Full report (Chinese, with every table and method note): https://ipure.dev/reports/2026-q4

This is an English summary, not a full translation. When citing numbers, use the snapshot date above.

Three conclusions

The distribution is barbell-shaped, not bell-shaped. 39.9% of checks fall in the clean end, 36.5% in the risk end, and 23.7% in between — far fewer in the middle than a bell curve would predict.
IP type explains most of the gap. Mobile averages 96, residential broadband 91, business broadband 91, data center 41. 89% of data center IPs are flagged as proxy or VPN exits by at least one source.
Native vs broadcast matters, but it is a diagnostic signal, not a cause. Native IPs average 74, broadcast IPs 35; most broadcast ranges are data center or proxy ranges, so the low score is not caused by “broadcasting” itself.

Read the sample before the numbers. These are IPs people voluntarily checked on IPure. They skew heavily toward data center and proxy exits (51.7% of the database). The report answers “among the IPs people test, which categories are cleaner”, not “what does the internet look like”.

1. Distribution: clean or risky, rarely in between

Purity band	Checks	Share
Pristine (95+)	19,687	35%
Clean (85–94)	2,927	5%
Neutral (70–84)	3,697	7%
Suspicious (50–69)	9,711	17%
Risky (25–49)	14,992	26%
Dangerous (<25)	5,674	10%

Risk signals stack: a proxy exit often also has a blocklist hit, high sharing and a data center range. That is why “partially clean” is less common than either end.

2. IP type: mobile and residential vs data center

IP type	Checks	Avg purity	Flagged proxy / VPN
Mobile	5,983	96	5%
Residential broadband	4,056	91	26%
Business broadband	16,991	91	11%
Data center	29,303	41	89%

The proxy flag rate for residential broadband (26%) is the highest of the four. A “residential” label is not a guarantee of trust: residential proxy services are large, and some of those IPs are resold exits.

3. Native vs broadcast: 74 vs 35

Native type	Checks	Avg purity
Native	43,614	74
Broadcast	13,024	35
Unknown small sample	50	82

Native means the registration country matches where the IP is used; broadcast means it does not. The gap is real, but most broadcast ranges are data center or proxy ranges, so broadcast is a useful diagnostic signal — not the cause of the low score. A native IP can still be on a blocklist; a broadcast IP can still be clean.

4. Providers: sample counts are not market share

Provider	Checks	Avg purity	50+	Blacklist / abuse	Native IP
AWS	4,411	32	7%	6%	23%
Tencent Cloud	1,600	37	50%	40%	96%
BandwagonHost	485	40	5%	1%	6%
Vultr	566	39	27%	3%	50%
Linode	876	29	15%	22%	43%

Providers are aggregated by ASN. A provider with more tested IPs is not necessarily worse; people test the machines they use for proxies, and that behavior shapes the sample. Full provider table (Chinese): ipure.dev/providers. English summary: ipure.dev/en/providers.

5. What this report cannot say

It is not a representative sample of the internet. Data center and proxy IPs are heavily over-represented; the database reflects what people bring to a checker.
Provider sample counts are not market share. They reflect which machines people tested, not how many IPs a provider has.
Correlation is not causation. More-tested providers tend to have lower averages, but the data cannot separate provider quality from how those IPs are used.
No platform’s internal risk data is visible. A score is a risk signal at the IP layer; it is not a guarantee that an account will be accepted or blocked.

Citation block

Source: IPure IP Purity Report, Q4 2026 (n=56,688 real IP checks, snapshot 2026-10-23 13:02 UTC+8). English summary: https://ipure.dev/en/reports/2026-q4 — full report (Chinese): https://ipure.dev/reports/2026-q4. Samples are self-selected and over-represent data center and proxy IPs. Licensed CC BY 4.0.

Check your own IP: run curl -L ipure.dev, or open ipure.dev.