Proxy Geo-Targeting

Proxy geo-targeting is choosing the geographic location of the exit IP that carries your request, so the target site sees traffic arriving from a place you selected. Every residential and mobile proxy vendor exposes it as a parameter on the connection string or the API call, and the values available define a granularity ladder: country, then state or region, then city, then ASN or mobile carrier for a specific network operator.

That ladder is also the price ladder. Country targeting is universal and effectively free. State and region are common. City is where vendors start gating access behind a higher tier, and ASN or carrier sits at the top with the thinnest supply behind it. SOAX advertises targeting "by region, city, or ASN" across "195+ geo locations" from a pool it describes as 155+ million IPs (soax.com/residential-proxies, read 2026-07-27), which is representative of how the category presents itself.

The mapping from address to place is the vendor's own, built from what it knows about its peers. Your target resolves the same address against its own geolocation database, and the two do not have to agree. A peer the vendor files under Lyon may read as Paris to the site you are scraping, or simply as France. Disagreement grows as granularity gets finer, and that is the same point at which the pool thins out. A country offers millions of candidate exits. A mid-sized city might offer a few hundred, and rotating through a few hundred addresses on a defended target burns them quickly.

Before paying for city exits, check whether the data actually varies by city. Plenty of localized pricing and inventory is set at country or state level, and plenty of region-gated content and locale-specific search output responds to an Accept-Language header plus a country-level exit. Where city really is required, verify the claim instead of trusting it: pull a hundred exits for that city, resolve each one against an independent IP geolocation database you control, and measure what fraction lands where the vendor said it would. Run that test before signing a committed-volume contract, not after.