What makes a carrier IP different
A mobile proxy routes your request through a mobile network, so it leaves from an address a carrier assigned to a handset or dongle rather than to a data centre or a home broadband line. The practical consequence is carrier-grade NAT: operators place large numbers of real subscribers behind a small pool of public addresses, so one mobile IP can represent thousands of genuine customers at once.
That is the whole reason mobile capacity survives on targets where everything else fails. Blocking a single mobile address risks cutting off a crowd of real users, so sites are markedly more reluctant to do it than with a datacenter range they can identify by ASN in one lookup. Addresses also rotate naturally as the carrier reassigns them, which means the churn you would otherwise engineer happens on its own.
When mobile is worth the premium — and when it is not
Mobile is the most expensive capacity we sell and the slowest, because it runs over real radio networks. It earns that premium in three situations: hardened social platforms and app APIs that reject other address classes outright; mobile ad verification, where you need to see the placement a phone would actually be served; and testing an experience that genuinely differs on a mobile connection.
Outside those cases it is usually the wrong purchase. If your target does not aggressively fingerprint address class, datacenter proxies are faster and cheaper. If it does fingerprint but is a normal consumer website, residential is the better balance of trust and cost. If you need one address to persist for days against an account, ISP proxies fit better than anything that rotates.
The honest test is empirical: run a few hundred real requests against your actual target on the cheaper option first, and only move up when the success rate tells you to. Choosing a proxy type sets out the decision.
Sessions on a mobile network
Mobile addresses move on their own, which is useful for stateless collection and a problem for anything carrying state. If your workflow logs in, fills a multi-step form or holds a cart, request a sticky session so the address stays put for the length of that flow — most sites bind their session cookie to the IP that created it, and a mid-flow change looks like a hijacked session rather than a returning user.
For independent fetches, let the carrier's natural rotation do the work and keep your request pacing human. Sticky sessions and how rotation works cover both modes.
Why the IP alone will not save you
Mobile gives you the strongest network-layer reputation available, and it is still only one of the signals a modern anti-bot system scores. Your TLS handshake, your HTTP/2 frame settings and your browser environment are all fingerprinted alongside the address, and a pristine carrier IP paired with a standard HTTP library's TLS signature is an obvious contradiction.
Consistency matters as much as quality here. A UK mobile exit, an Accept-Language of en-US and a New York timezone are three answers to the same question that disagree, and that disagreement is itself the signal. The five detection layers explains which ones a proxy can address and which are yours.
The failures we see most
Timeouts and high latency. Expected to a degree — you are routing over a radio network. Reduce concurrency before you assume a fault; past a certain thread count you queue rather than parallelise and everything slows at once.
A challenge page returned with a 200. If you are not validating the response body, you are counting blocks as successes and your metrics are misleading you. Check for a marker you expect on a real page.
Session drops mid-flow. Almost always rotation applied to a stateful workflow. Switch that job to a sticky session.
Our five-step test separates a proxy problem from a target problem quickly, and support is staffed 24/7 if it turns out to be ours.
Acceptable use
Mobile capacity is scarce and shared, so we are stricter about what runs on it. A proxy changes the network path, not the rules: a robots.txt disallow is a refusal whether or not it is enforced, and using our network to circumvent an explicit refusal is outside our terms, as are the targets set out in our acceptable use rules. What we log, and for how long, is documented in our logging policy.
