What IP Checks Should You Run Before Using AI? A Checklist
What IP checks should you run before using AI? Before you register or log in to your main account, spend two minutes running through the 7 core checks on this list: IP purity, sharing count, IP geolocation, DNS leaks, WebRTC leaks, IPv6 leaks, and system timezone consistency. Only when all seven are green is your network environment genuinely ready. If any one of them flashes red, fix it or switch lines before you start using it — otherwise you're very likely to get instantly banned or stuck in endless verification.
Plenty of people buy a new line and rush straight off to register an account or bind a card for a membership, only for an "account restricted" popup to appear the moment they hit submit — wasting a legitimate email address and SMS verification costs for nothing. Instead of spending days writing appeal letters afterwards, run through this standardised checklist item by item every time you set up a new environment, and keep risk-control problems outside the door.
The checklist: 7 checks
The risk engines of the major AI platforms judge a visitor on the combined picture of their whole network and device environment. To be thorough, pull out the seven dimensions below and screen them one by one before every formal connection.
The first three are the hard fundamentals of the network exit itself. One is the exit's IP purity, which directly reflects whether this address has ever been used for bulk registration, malicious crawling, or spam. Two is the current sharing count, which tells you whether you've bought a single-household dedicated line or a public pool shared by thousands of people. Three is the exit's IP geolocation, confirming it sits firmly inside a country on the target platform's service whitelist, with all databases in agreement and no cross-border broadcast drift.
Next come three anti-leak checks on the local communication tunnel. Four is DNS leaks — checking whether domain lookups honestly travel out through the overseas proxy tunnel instead of quietly connecting straight to your domestic broadband ISP. Five is WebRTC leaks — checking whether the browser's peer-to-peer interface has punched through the proxy and handed your local network's real LAN or domestic public address to page scripts. Six is IPv6 leaks: many proxy tools only take over IPv4 traffic by default, so when a target site supports dual-stack access, your computer's native IPv6 address connects straight to the target server and bypasses the proxy entirely, instantly exposing your real physical location.
Seven is the consistency of system and browser environment parameters: checking the operating system timezone, the timezone offset in minutes, and the browser's preferred language to make sure they line up exactly with the exit's country — no rookie slip-ups like "exit in Los Angeles, clock running Beijing time". For the underlying principles behind the three core anti-leak checks, see Check These Three Things Before Using Claude: WebRTC, DNS, and IPv6 Leaks.
The pass line for each check
Knowing which seven items to check isn't enough. When a screen full of numbers and English parameters pops up, you need a clear pass mark in mind. The reference table below lays out the pass line for each item plainly — just match your results against it while you test:
| Check | What to Watch | Pass Line (Recommended Standard) | Danger Signal |
|---|---|---|---|
| IP purity | Overall risk score and network type | <20 Pure (for everyday old accounts, relaxed to at most <35 Fairly Pure), and a native residential range | Score falls into <50 Fair, <70 Suspicious, or ≥70 High Risk, or shows as a commercial datacenter range |
| Sharing count | Recent active concurrent users | 1-10 Excellent (for everyday regular use, no higher than 10-100 Good) | Jumps into the 100-500 Poor, 500-5000 Very Poor, or 5000+ Dangerous band |
| IP geolocation | Country/city location and ASN organisation | Lands in a whitelisted supported country, with all databases agreeing on the region and a local consumer broadband provider | Lands in an unsupported restricted region, or has a geographic database conflict from cross-border broadcast |
| DNS leak | Country of the DNS resolver | Resolver country exactly matches the exit IP's country | Resolver shows a mainland China ISP, or a country on a different continent from the exit |
| WebRTC leak | Address returned by the browser's STUN probe | Only shows the overseas proxy exit address, or shows as disabled / not detected | A local broadband real domestic public or private address appears in the candidate list |
| IPv6 leak | Dual-stack network exit connectivity | IPv6 channel disabled, or returns an overseas proxy exit in the same region as IPv4 | The site detects a native mainland China IPv6 address assigned by a local ISP |
| Timezone consistency | System timezone and browser preferred language | System timezone matches the exit city to the second, preferred language is the local mainstream language | Exit in Europe, the US, Japan, or Korea, but the system timezone is still stuck on Asia/Shanghai |
Reading the table, any item that hits a "danger signal" on the right means your environment has a clear risk exposure. That matters most when you're about to register a new account or bind a foreign card for a paid subscription — then you must hold IP purity firmly at the <20 Pure tier and keep concurrent users pinned to the 1-10 Excellent tier, with no lowering the bar to get by. For the details of how each score band reads, see What IP Purity Score Counts as Clean? How to Read Each Tier.
How to run each check
Once you've memorised the standards, actually running the full IP check is very fast — open your browser, walk through the two steps below in order, and the whole thing takes under two minutes.
First, the basic health check on the network exit. After connecting to your proxy line, open the ISPing.me homepage and run an IP check. Wait for the full health report to load, then scan down through four key spots from top to bottom: first check whether the overall score at the top is under 20 (or in the <35 Fairly Pure band); then look at whether the concurrent user count in the details sits in the 1-10 Excellent tier; then confirm the network type is clearly labelled as home broadband and the ASN operator name is a local consumer telecom provider in the target country; for how residential and datacenter IPs differ, see Residential IP vs. Datacenter IP: What's the Difference; and glance at the blacklist status bar to make sure every threat-intelligence database shows zero hits.
Next, the dedicated browser-environment and anti-leak check. Open the dedicated Claude check page, which runs a one-stop cross-check on the local weak spots that high-risk AI platforms love to grab. Once the page has loaded, confirm the four items below one by one:
- Check the DNS server's location: confirm the DNS resolver's country flag and operator shown on the page match the proxy exit you found in the previous step exactly, with no domestic China Telecom, Unicom, or Mobile resolver nodes appearing.
- Check the WebRTC status bar: confirm the probe results show no local LAN range or domestic public address leaking out — it should show the proxy exit or a successful block.
- Check the IPv6 exit status: confirm the current browser isn't connecting to the internet directly through the local native IPv6 channel — it should either show IPv6 as not enabled, or show a location in the same country as the IPv4 exit.
- Compare timezone and language parameters: check the browser timezone name and system language captured on the page, and confirm they match the exit city exactly (for example, if you're on a US West Coast node, the timezone must show
America/Los_Angeles). For how to align timezone and language, see IP, Timezone, and Language: How to Keep Your Account Environment Consistent.
Fix it first, then use it
After running the seven checks, if a few items light up red, don't gamble on "let me just try logging in first" and open the official site anyway. Handle each failed item directly as below, retest until everything turns green, and only then start working.
If the test exposes a local communication leak or a parameter mismatch — say a DNS leak, a WebRTC punch-through, a direct IPv6 connection, or a timezone mismatch — these can all be fixed locally in a few minutes. Open your proxy software's settings panel, force "remote DNS resolution" on and change the local network card's default DNS to 1.1.1.1; in your computer's network adapter properties, uncheck Internet Protocol Version 6 (TCP/IPv6) to shut the IPv6 channel off entirely; install an anti-WebRTC-leak extension in your browser and disable non-proxy UDP traffic; then go into the operating system's time settings and manually set the timezone to the exit city. After those tweaks, refresh the check page and confirm every leak alert is gone. For the detailed fixes, see What Is a DNS Leak? How to Detect It.
If the network exit itself is the problem — say the overall score is as high as 65 (the <70 Suspicious tier), the sharing count is over 500 people, the underlying range shows as a commercial datacenter, or the IP geolocation lands in an unsupported region (for which regions aren't supported, see What to Do When Claude Says Your Region Isn't Supported?) — that's a defect in the underlying line, and no amount of tweaking local computer settings can save it. The fastest move is to disconnect from the current node right away, switch to a dedicated native residential line (for how to verify before switching lines, see How to Verify a Proxy Before You Buy It) that meets the <20 Pure and 1-10 Excellent tiers in your provider's dashboard, and then run the seven-item checklist again on the new line — only log in to your business account once everything is green.
FAQ
Do I need to redo all 7 checks every time I open an AI platform?
No, you don't need to retest before every chat. You must run the full 7-item check when you first buy a line, switch proxy node cities, or reinstall your browser or proxy client. Once you've confirmed the environment is all green and you're fixed on a dedicated static line, just spot-check purity and sharing count every week or two for drift.
My shared proxy node only fails IP purity (60 points, Suspicious) — the other 6 items are all green. Can I still use it?
Not for registering a new account or binding a payment card. A score of 60 means the exit has a clear recent history of heavy sharing or abuse. Even if your local anti-leak setup is solid, the platform firewall will still trigger frequent verification or outright blocks based on the exit's high risk score when the connection is established. Switch to a cleaner line under 35 points.
If I turn off the IPv6 channel on my computer, will it affect my normal use of domestic websites and video streaming?
Not at all. The vast majority of mainstream websites and video platforms at home and abroad fully support the IPv4 stack. Turning off the local IPv6 channel just stops some dual-stack overseas sites from bypassing the proxy and connecting straight to the local network — everyday domestic work, video streaming, and downloads all keep working normally.
My check page shows all 7 items green, but registering still says my environment is abnormal — why?
All green rules out the network exit and browser leaks. If you're still being rejected, the trigger is in the account registration details themselves. Check whether the SMS platform's number range has been overused, whether the registration email is an obscure temporary domain, or whether the credit card's issuing region doesn't match the current exit country.
Notes
- After fixing any leak, clear the old cache before opening the target site: if you accidentally visited the target platform before fixing a DNS or WebRTC leak, the browser's local cache may already hold an abnormal marker — clear the domain's cookies after fixing, then log in.
- Don't turn on your proxy software's "bypass mainland China and common domains" smart-split rule just to chase lower latency: some crude split rules misjudge the third-party verification endpoints the target platform calls (such as human-verification domains) as direct traffic, so the main page goes through the proxy while the verification call goes direct locally — triggering an environment mismatch instantly.
- Every environment in a multi-account setup needs its own run through the checklist: when you create a second or third separate browser profile, don't assume everything is fine just because the host machine is set up — each profile's timezone, language, and anti-leak switches need to be verified separately.
Check these 7 things before using AI
Check your network environment on ISPing.me — confirm the purity, sharing count, and risk signals, and only use it once it passes.
Check nowRelated reading:Check These Three Things Before Using Claude: WebRTC, DNS, and IPv6 LeaksWhat Is IP Purity? How to Read the Risk ScoreWhat Is the IP Sharing Count, and Why Does It Matter?What Is a DNS Leak? How to Detect ItWhat IP Purity Score Counts as Clean? How to Read Each TierWhat to Do If Your IP Geolocation Shows the Wrong PlaceWhat to Do When Detection Results Don't Match Real Usage