Do Google AI Mode answers change by location and language?
Summarize with ChatGPTClaudePerplexity
TL;DR
- Google AI Mode answers and citations can change with country, city, language-related context, account history and other Search signals.[9][33]
- Google publishes supported countries and languages separately, so a country-language inventory is more accurate than assuming one market represents an entire language.[25]
- SE Ranking found 19% URL overlap for repeated general local queries within the same city versus 2% across cities in its specific five-city test; this is not a universal variance rate.[22]
- Logged-out or personalization-off testing does not eliminate location effects, because Google documents localized results and device, IP and activity signals separately.[2]
- A defensible tracker records market, city, prompt language, account state, collection method, conversation state and repeated evidence for every AI Mode result.[31][2]
Yes. Google AI Mode can vary by country, city, language-related context, account history, device location and other Search signals. Published tests have found different cited URLs and local recommendations across locations, but the available evidence does not support one universal “location variance” percentage.[1][2]
For practitioners, the right approach is to track defined country–city–language cells, document account and collection state, repeat each test location, and compare answer text and citations—not only visibility scores.
Google announced AI Mode availability in more than 200 countries and territories in October 2025, but availability, languages and optional features do not expand identically everywhere.[3][4] Google’s current Help page lists countries and languages separately, so a supported language should not be treated as proof that AI Mode works in every country where that language is spoken.[5][4]
What is Google AI Mode by country, city and language?
Google AI Mode is a Search experience that generates answers and links using multiple related searches, or query fan-out, across subtopics and data sources.[6] Because Google can use location, language settings, previous activity and other contextual signals, two users asking similar questions may receive different answers or sources.[1][2]
Location is not limited to a manually selected country. Google documents device location, saved home and work addresses, previous activity and IP address as possible location signals.[7] Without precise device-location permission, Google can still estimate a general area, and a browser location cookie may retain the last device location for up to six hours.[7]
Language is a separate tracking dimension. Google’s AI Mode Help page publishes a language list independently from its country and territory list.[2] A German-language prompt in Germany, an English-language prompt in Germany and an English-language prompt in the United States should therefore be treated as different test conditions.
Account state adds another layer. Google says AI Mode personalization can reference previous searches and activity saved in Search Services History, while Personalized Recommendations can affect Search experiences.[8][9] Turning personalization off does not remove all localized or contextual differences.[10]
Where is Google AI Mode available?
Google’s current Help page is the best source for checking individual countries and territories.[7][9] Its published list includes the United States, Canada, Mexico, Brazil, Argentina, Chile, Colombia, Peru, the United Kingdom, France, Germany, Italy, Spain, the Netherlands, Poland, Switzerland, India, Japan, South Korea, Indonesia, Singapore, Australia, New Zealand, the United Arab Emirates, Saudi Arabia, Israel, South Africa and other markets.[9][11]
The list is not equivalent to a statement that every feature works identically in every market. Google separately announced additional languages, countries, Search Labs access and optional capabilities over time.[9]
| Date | Google-announced change | Scope |
|---|---|---|
| March 5, 2025 | AI Mode introduced as an opt-in Labs experiment | Initial access for invited Google One AI Premium subscribers; the cited launch did not provide a complete country-language list.[12][2] |
| May 1, 2025 | AI Mode Labs access opened more broadly | United States access, not the later general rollout.[1][2] |
| May 20, 2025 | General AI Mode rollout began | United States, without Labs signup.[9][1] |
| June–July 2025 | India rollout | English-language Labs access began in June, followed by broader rollout in July. |
| July 28, 2025 | UK rollout began | Google Search and the Google app.[13] |
| August 21, 2025 | Expansion to more than 180 new countries and territories | English-language expansion.[2] |
| September 8, 2025 | Five additional languages launched | Hindi, Indonesian, Japanese, Korean and Brazilian Portuguese.[9] |
| October 7, 2025 | More than 40 new areas and 35-plus new languages announced | Total availability reached more than 200 countries and territories.[14][15] |
| October 8, 2025 | Seven additional Indian languages announced | Bengali, Kannada, Malayalam, Marathi, Tamil, Telugu and Urdu.[16][9] |
Google’s Search Live and Personal Intelligence announcements should not be treated as new base AI Mode launches. Search Live expanded an interaction feature to locations and languages where AI Mode was already available, while Personal Intelligence had its own geographic and language scope.[2][17]
How does Google use location in AI Mode?
Google documents several location signals for Search, including IP address, device location, saved addresses and activity associated with previous searches.[7][18] Precise device location can support nearby results, while a general location estimate may still be available when precise permission is disabled.[7]
Google also says Search results can differ because of language settings and localized results even when Personalized Recommendations are disabled.[10] This means “personalization off” is not the same as “location off.”
For local queries, the distinction is especially important. A generic query such as “best plumbers” may depend heavily on inferred location, while “best plumbers in Manchester” provides an explicit geographic instruction. The explicit city reduces one ambiguity but does not guarantee identical answers across runs.
Google’s local guidance identifies relevance, distance and prominence as major local-ranking factors and recommends accurate Business Profile information, current hours, review responses and useful media.[19] Google separately recommends keeping Business Profile and Merchant Center information up to date for AI features.[20][21]
These recommendations do not constitute a published AI Mode ranking formula. They do show why local tracking should inspect business details and recommendations alongside cited website URLs.
What have independent AI Mode tests observed?
SE Ranking’s five-city AI Mode study is the clearest direct evidence in the report. It tested 5,000 queries across New York, Los Angeles, Washington, Houston and Denver, with three runs per city.[22]
For general local queries, the study reported 19% URL overlap and 35% domain overlap within the same location, compared with 2% URL overlap and 24% domain overlap across locations.[22] For queries that explicitly named a location, overlap was considerably higher: nearly 48% of URLs and 55% of domains within a location, compared with 43% of URLs and 49% of domains across locations.[22]
Those figures describe that study’s queries, locations and collection method. They are not a universal AI Mode variance rate.[23]
The study also found imperfect overlap within the same city, which means a difference between two cities cannot automatically be assigned entirely to geography.[22] A tracking program needs repeated observations in each location to establish the normal same-location baseline.
Cloro’s state-level study collected 279 successful responses from eight local-intent prompts across six US states and used UULE for AI Mode location settings.[24] It documented differences in local recommendations but used single calls for each prompt, engine and state cell, so it does not establish a stable state-level variance estimate.[24]
A Profound analysis of 3.25 billion citations found citation-pattern differences across native-language country cohorts, including AI Mode.[23] However, its country groups also differed in prompt language and prompt mix, so it is evidence of country-language cohort differences rather than a controlled language-only experiment.[2]
An ACL Anthology study found AI Overviews for 81% of US queries compared with 65% of German queries in its test environment.[2] The study used logged-out, clean environments and country-specific virtual-machine IPs, but it did not directly test AI Mode.[2]
What should an AI visibility tracker record for each Google AI Mode result?
The following table separates Google’s documented mechanisms from what published tests and vendor methods have actually established.
| Factor | What Google documents | What tests observed | What to set in a tracker |
|---|---|---|---|
| Country | Search can use IP address and other location signals; AI Mode publishes a supported-country list.[2] | Country cohorts can show different citation patterns, but country and language may be confounded.[23] | Country or territory, eligibility date and collection route. |
| City | Google documents general location use and local Search signals.[22][7] | SE Ranking found different URL overlap across cities; same-city runs were not perfectly identical.[22] | City for location-sensitive queries, plus repeated runs. |
| Region or state | Google’s general location documentation supports regional context.[5] | Cloro observed different state-level local recommendations using location-targeted collection.[24] | State, region or province where local decisions differ. |
| Prompt language | Google lists supported languages separately from countries.[25] | Native-language country cohorts showed different citation patterns, but the design did not isolate language alone.[23] | Original prompt language, translation relationship and interface language separately. |
| Device location | Google can use precise device location when permission is enabled and estimate a general area otherwise.[7] | Published vendor studies do not consistently disclose device-permission state.[5] | Permission state, if known; otherwise “Not published.” |
| IP location | Google identifies internet address as a possible Search location signal.[7] | Cloro used UULE for AI Mode; other vendors publish different routing methods.[24] | Requested location, actual routing method and IP/proxy state. |
| Account history | AI Mode personalization can use previous searches and Search Services History.[9][8] | Many vendor methods do not publish login or history state.[26][5] | Signed-in/out state, history settings and connected apps. |
| Personalization | Personalized Recommendations can affect results; disabling them does not remove all localization.[8][10] | Ahrefs documents collection without stored user context or personalization.[26] | Personalization setting and whether the vendor explicitly disables context. |
| Conversation state | AI Mode can use earlier conversations and preferences in supported personalization experiences.[9] | Published location tests do not consistently disclose whether conversations were fresh.[1] | Fresh conversation, follow-up conversation or unknown. |
| Surface | AI Mode and AI Overviews can use different models and techniques.[6] | Academic tests of adjacent Search surfaces cannot be treated as direct AI Mode tests.[27] | Store AI Mode, AI Overviews and Gemini as separate datasets. |
Which AI visibility tracking tools offer country, city and language controls for Google AI Mode?
Vendor controls vary from country selection to city targeting, and their methods are not interchangeable.
SE Visible documents country, region and city targeting, with 42 countries listed in its documentation.[25] Promptwatch documents monitor-country selection and publishes a broader method using geolocated residential proxies, matching browser locale and timezone; its separate AI Mode implementation details are not fully published.[5]
Peec documents per-prompt country locations, ISO country codes and an IP-address run-location choice.[4] Its prompt-discovery documentation also includes location and language fields, but that does not establish a city-level collection control for every existing prompt.[8][5]
Semrush documents 220-plus countries and territories for current Prompt Tracking and separately advertises location-language options in Brand Performance.[11] Its historical AI Mode Position Tracking announcement described US desktop tracking down to city, state or ZIP code, but that announcement concerns a particular product and period.[28]
Ahrefs Brand Radar documents corpus locale behavior and says it does not use stored user data, prior context, personalization or pre-prompting.[26] Its documentation does not establish a custom AI Mode city-control field in the reviewed material.
Finseo tracks Google AI Mode as one of its core engines, with its own tracker page, and tracks prompts in 40+ countries and their languages, each with its own competitors and sources.[22][5] Finseo is hosted in Frankfurt, Germany with EU data residency, is GDPR-compliant with a data processing agreement (DPA) and is ISO 27001 certified. For enterprise teams it offers SSO via SAML, roles per project, an audit trail and unlimited seats, according to its enterprise page.
A tracker specification should therefore record the vendor product, geography granularity, prompt language, IP or proxy method and account-state disclosure. “Country tracking” and “city tracking” are not equivalent datasets.
How should practitioners design a location plan?
A defensible plan should treat geography, language, personalization and volatility as separate variables.
- Define the market cell: record country, region or city, prompt language and intended audience.
- Choose the exact surface: keep AI Mode separate from AI Overviews and Gemini.
- Separate implicit and explicit local intent: test generic queries such as “best accountants” separately from queries naming a city.
- Set the location method: document browser location, IP routing, UULE, vendor location field or another mechanism.
- Record account state: note signed-in or signed-out status, Search Services History, Personalized Recommendations, location permissions and connected apps.
- Start fresh conversations: use new conversations for baseline prompts and store follow-up conversations separately.
- Separate country and language tests: hold country constant when comparing translated prompts; hold language constant when comparing locations.
- Repeat every cell: compare city-to-city changes with repeated runs in the same city. The available evidence does not establish one universally sufficient repeat count.[2][22]
- Store answer evidence: retain answer text, cited URLs, cited domains, brand mentions, local recommendations, timestamps and requested location.
- Annotate method changes: record changes in Google eligibility, vendor product, routing, account handling and prompt translations before interpreting trend lines.
This design prevents a common analytical error: attributing every answer difference to location. Google documents localization and personalization separately, while published studies also show run-to-run instability.[15] In Finseo’s own German-language (DACH) citation-stability study, only 42.7% of Google AI Mode cited URLs were still cited the next day for the same prompt.
What does this mean for local SEO and AI visibility?
For local businesses, tracking should measure more than whether a domain appears as a citation. Record whether AI Mode recommends the business, displays correct hours or prices, shows ratings, offers directions or presents other local actions.
Google says AI Mode can research local costs, appointment availability, local stock, pricing and reservations, subject to feature-specific eligibility.[29] Google’s earlier AI Mode announcement also described visual place cards with ratings, reviews and opening hours for local businesses.[30]
Website citations, Business Profile information and local recommendations are related but distinct outputs. Accurate Business Profile and Merchant Center data remain practical foundations, while Google does not promise inclusion from any single optimization action.
Frequently asked questions
Does Google AI Mode give different answers in different countries?
Yes. Google documents location and language-related Search differences, and published country-cohort research has found different citation patterns.[23] Country comparisons should still document prompt language and account state because those variables may change at the same time.
Does Google AI Mode give different answers in different cities?
It can. SE Ranking’s direct AI Mode city test found different URL and domain overlap across five US cities, while also finding imperfect repeatability within individual cities.[22] City-level tracking should therefore include repeated runs rather than treating one response as definitive.
Does changing the prompt language change AI Mode citations?
It can, but the reviewed evidence does not isolate a universal language-only effect. Profound found differences across native-language country cohorts, but country, language and prompt mix were not independently controlled.[23] For a cleaner test, hold the country and account state constant while comparing equivalent prompts in different languages.
Are AI Mode results different when personalization is turned off?
They may still differ because Google says localized results and language settings can affect Search independently of Personalized Recommendations.[10] Turning personalization off removes one class of account-based context, not every location or language signal.
Does logged-out Google AI Mode remove personalization?
Not necessarily every contextual signal. Google documents location estimates, previous device activity and browser location information that can affect Search even without a signed-in account.[7] A tracker should record logged-out status separately from location and history controls.
How should a company track Google AI Mode by city?
Create separate country-city-language cells for locations where recommendations, customers or regulations differ. Use the same prompt set, document the routing and account state, repeat each cell, and save answer text and cited URLs alongside visibility metrics.[31]
Which AI visibility tools track Google AI Mode by country or city?
SE Visible documents country, region and city targeting, with 42 countries listed in its documentation.[25] Peec documents per-prompt country locations, ISO country codes and an IP-address run-location choice.[4] Semrush documents 220-plus countries and territories for current Prompt Tracking.[12] Country tracking and city tracking should not be treated as equivalent datasets.
Does Ahrefs Brand Radar use personalization when collecting AI answers?
No. Ahrefs Brand Radar documents corpus locale behavior and says it does not use stored user data, prior context, personalization or pre-prompting.[26] Its documentation does not establish a custom AI Mode city-control field in the reviewed material.
Does Semrush track Google AI Mode by location and language?
Semrush documents 220-plus countries and territories for current Prompt Tracking and separately advertises location-language options in Brand Performance.[11] Its historical AI Mode Position Tracking announcement described US desktop tracking down to city, state or ZIP code. That announcement concerns a particular product and period.[32]
Sources
- 1AI Mode Research: Sources, Volatility, & Differences between AIO and Organic Search — seranking.com
- 2Characterizing Web Search in the Age of Generative AI — aclanthology.org
- 3Where Search Labs & experiments are available - Google Search Help — support.google.com
- 4Setting up your prompts — docs.peec.ai
- 5Which AI Visibility Tools Track Brand Mentions Across Multiple Countries and Languages? | Promptwatch — promptwatch.com
- 6AI Features and Your Website | Google Search Central | Documentation | Google for Developers — developers.google.com
- 7Understand & manage your location when you search on Google - Android phone & tablet - Google Search Help — support.google.com
- 8Get started with Search Services History & Personalized Recommendations - Google Search Help — support.google.com
- 9Get AI-powered responses with AI Mode in Google Search - Computer - Google Search Help — support.google.com
- 10Manage Personalized Recommendations in Search services - Google Search Help — support.google.com
- 11AI Visibility Toolkit: Boost Brand Visibility in AI Search — semrush.com
- 12How to Track Your Google AI Mode Visibility with Semrush — semrush.com
- 13AI Mode now available on Google Search in the UK — blog.google
- 14AI Mode in Google Search expands to more than 40 new areas — blog.google
- 15Does Location Matter with AI Search? What 31,200 AI Answers Show | A/GEO Research — ageoresearch.com
- 16Supercharging Search for India: New Languages in AI Mode & Search Live debuts — blog.google
- 17Personal Intelligence: How AI Mode in Search personalizes responses for you - Google Search Help — support.google.com
- 18How Google uses location information – Privacy & Terms – Google — policies.google.com
- 19Tips to improve your local ranking on Google - Google Business Profile Help — support.google.com
- 20Google's Guide to Optimizing for Generative AI Features on Google Search | Google Search Central | Documentation | Google for Developers — developers.google.com
- 21AI Mode in Google Search: Updates from Google I/O 2025 — blog.google
- 22How Volatile Are AI Mode Results? 2025 Local Search Test — seranking.com
- 23How query language reshapes AI citations | Profound — tryprofound.com
- 24How AI Answers Change by State: A Data Study | cloro — cloro.dev
- 25SE Visible FAQ – Knowledge Base — help.seranking.com
- 26What is Brand Radar, and how to use it? | Help Center - Ahrefs — help.ahrefs.com
- 27Google AI Mode visibility tracker - Peec AI — peec.ai
- 28Track AI Mode in Position Tracking — semrush.com
- 29Use AI Mode to check local availability and pricing - Computer - Google Search Help — support.google.com
- 30Try Google Search’s AI Mode in Labs — blog.google
- 31Prompt tracking methodology to measure AI visibility — semrush.com
- 32Expanding AI Overviews and introducing AI Mode — blog.google
- 33How Promptwatch collects data — promptwatch.com