Geo-targeting and IP-based results

Geo-targeting is the set of signals a search engine uses to decide which country or region your page serves. The query enters Google's pipeline, the pipeline maps the requester to a locale, and the SERP assembles from that locale's index slice.

Three signal classes dominate. The ccTLD on your root domain. The IP address of your server. The hreflang, language tags, and structured data you ship with each URL. Google's own documentation ranks ccTLD as the strongest, hreflang next, server IP third (https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites).

The user signal arrives last but matters most in practice. Google also weighs the requester's IP, their Google account locale, and prior session behavior. The composite determines which version of your site surfaces for any given query, per Google's multi-regional site documentation (https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites).

IP-based geolocation is a noisy signal. A request from a corporate VPN in Singapore can register as San Francisco. A mobile user on a roaming carrier in Tokyo can resolve as their home billing region. The address type, not the city, drives the answer.

GeoIP databases differ. MaxMind, IP2Location, and Google's internal map all disagree on 8-15% of routable IPs at the city level, per ongoing 2025 benchmark work referenced in the technical SEO community (https://www.technicalseo.com/tools/). A 95% accuracy claim is a marketing line, not a measurement.

CDNs further erode IP locality. Your site's origin sits in Frankfurt, your edge nodes sit in 250 cities, and Googlebot receives the edge response. The IP Google sees is the edge node, not the origin. The server location signal weakens each time you add a POP.

CDN operators know this. Cloudflare, Fastly, and Akamai publish network ranges so crawlers can identify edge IPs. Google's crawler adjusts its interpretation. The localization hit from a multi-region CDN is small, not zero. Audit it quarterly.

A ccTLD still anchors geo-targeting. example.de signals Germany harder than example.com/de/, even with hreflang in place. The signal pre-dates the query. Google's 2025 documentation confirms ccTLD as the strongest single cue, ahead of hreflang and server IP (https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites).

Subfolders carry the second-strongest signal. example.com/de/ consolidates authority into one domain, simplifies backlink accounting, and keeps operational complexity low. The cost is a softer geographic cue than a ccTLD. Most multinationals chose subfolders by 2025, per ccTLD trend analysis (https://www.namesilo.com/blog/en/seo/do-country-code-domains-still-drive-local-seo-success).

Hreflang is the binding layer between locale variants. Each regional page returns tags pointing to every other regional version, with reciprocal links, in valid language-region codes. Broken hreflang chains silently demote every linked variant. Audit before you scale.

Google started redirecting its own country domains to google.com in April 2025. Google.sg, google.fr, google.de all now resolve to the single google.com surface. The change covers Google's interface only, not how third-party sites rank (https://www.heroesofdigital.com/seo/google-is-phasing-out-cctlds-seo-impact/).

The redirect signals internal maturity, not external rule change. Google confirmed that site-level ccTLDs, hreflang, and Search Console geo-targeting still operate as before. Your .de domain still gets the German geo signal. Nothing about your own ranking logic moved.

Localization weight redistributed onto content and behavior. Google's own system now relies more on visible content language, structured data, local backlinks, and user behavior patterns. Localization got harder in 2025, not easier. Domain shortcuts carry less weight (https://www.namesilo.com/blog/en/seo/do-country-code-domains-still-drive-local-seo-success).

User-facing personalization grew in parallel. Google Maps, Local Pack, and country-restricted results now depend on the requester's geolocation more than ever. The same query can produce different SERPs for two users in the same city, per Search Everywhere's 2026 personalization analysis (https://www.searcheseverywhere.com/blog/google-ai-overviews-in-2026-search-data).

Quarterly geo audit. Open a private browser and query your top 30 commercial terms from three locations: your home IP, a VPN endpoint in your primary market, and a VPN endpoint in your secondary market. Record which version of your site ranks in each.

Cross-reference against hreflang return chains. Open each regional URL, view source, confirm the hreflang tag points to the correct sibling and that the sibling returns a reciprocal tag. One missing return tag breaks the cluster. Fix before publishing new variants.

Compare organic performance across locales in GSC. Filter by country. Compare impressions, CTR, and average position per locale. A locale with stable impressions but low CTR has a title or schema problem. A locale with dropping impressions has a ranking problem.

Note the gap. This post synthesizes 2025-2026 data from four sources: Google Search Central, Heroes of Digital, NameSilo, and Search Everywhere. Google's exact requester-locale ranking algorithm remains undisclosed. Replication required on your own site before you ship.