Google AI Overviews by country: length, sources and local citations compared

If an AI Overview answers a search, who sees which answer depends on where they are. For a brand that sells in more than one market, that raises a practical question: is the overview in the UK or India the same one you see in the US, or a different one, citing different sites?

We ran 10 searches in five countries and compared the overviews on length, sources and local citations. This guide has the numbers, a function to profile your own searches, and what it means for tracking. We used the Google Search API in October 2026. Overviews change from run to run, so yours will differ.

How we ran it

We used 10 searches: what is inflation, best vpn, how to lower blood pressure, how to start a podcast, project management software, best credit card, how does a vpn work, how to change a tire, noise cancelling headphones and what causes headaches. Each ran in us, gb, au, ca and in, in English, on desktop. That is 50 requests, and all 50 completed.

It is a small set we picked, checked once. The figures describe it and nothing wider.

Overviews show almost everywhere

48 of the 50 searches had an AI Overview. The two without were noise cancelling headphones in the US and in Australia. All 10 searches had one in the UK, Canada and India.

Length and sources

The UK overviews were the longest and cited the most, and the Indian ones were the shortest, at about two thirds of the US length, with the fewest sources. One Canadian overview was only 26 words, and one Indian one ran to 484.

Take these with the usual caution about instability. When we repeated what is inflation in the US while testing the code, it came back at 171 words and 10 sources, against 103 words and 4 sources in the study. A single check is a sample. We found the same movement between runs in which sites AI Overviews cite.

Nothing matches the US text

We compared each country’s overview with the US one for the searches that had both:

None of the 36 overviews was word for word the same as the US one. And about a quarter of the cited domains in each country, between 24% and 28%, were also cited in the US one. That is similar to what we saw in Google results by country, where the sources the overviews cited overlapped more than the organic results did.

For what is inflation, the first sources in each country show why:

us  federalreserve.gov, en.wikipedia.org, imf.org
gb  bankofengland.co.uk, ons.gov.uk, halifax.co.uk
au  en.wikipedia.org, federalreserve.gov, commbank.com.au
ca  en.wikipedia.org, bankofcanada.ca, facebook.com
in  en.wikipedia.org, britannica.com, imf.org

The UK overview led with the Bank of England and the Office for National Statistics, Canada’s included the Bank of Canada, and Australia’s cited the Commonwealth Bank. Wikipedia appeared in four of the five.

Local citations

We counted citations to sites on the country’s own domain endings: .co.uk and .uk for the UK, .com.au and .au for Australia, .ca for Canada and .in for India:

The local sites included bankofengland.co.uk, nhs.uk, ons.gov.uk and money.co.uk in the UK, finder.com.au, commbank.com.au and healthdirect.gov.au in Australia, bankofcanada.ca, heartandstroke.ca and ratehub.ca in Canada, and airtel.in and reliancedigital.in in India.

This is a lower bound. A local publisher on a .com domain is not counted, and we did not measure the US. Even so, the order is informative: Australian and UK overviews leaned on local sites, while Canadian and Indian ones mostly cited international ones. An international brand with no local domain is not shut out, but local sites make up a larger share of the citations in some markets than in others.

Do the overviews name the country?

Some do. We looked for the country’s name or currency in the text:

For best credit card, the opening words were:

gb  The best credit cards in the UK depend on what you need—whether that is fee-free spending abroad, cashback, ...
au  The best credit card in Australia depends on your main financial goal, with top options split across ...
ca  The top credit cards in Canada vary depending on whether you want flexible travel points or cash back: ...
in  The best credit cards in India depend on your primary spending categories, such as online shopping, ...
us  The top credit cards span several key categories based on rewards and fees: Travel & Lounge Cards ...

All four non-US overviews named their country in the first sentence. The US one did not. The Indian overview also cited paisabazaar.com as its first three sources, three separate links to one site, so the number of sources is not always the number of different sites.

The UK headphones overview used the pound sign. We did not look for the US, because American matches in the text were brand names such as American Express.

The code

Put your key in an environment variable:

export CRAWLERDATA_API_KEY="YOUR_API_KEY"

import os
import time
from concurrent.futures import ThreadPoolExecutor
from urllib.parse import urlparse

import requests

API_URL = "https://api.crawlerdata.com/v1/jobs"
HEADERS = {"Authorization": f"Bearer {os.environ['CRAWLERDATA_API_KEY']}"}

LOCAL_SUFFIXES = {
    "gb": (".co.uk", ".uk"),
    "au": (".com.au", ".au"),
    "ca": (".ca",),
    "in": (".in",),
}


def google_search(query, country="us", retries=4):
    body = {"endpoint": "google-search", "params": {"query": query, "country": country}, "queue": "live"}
    for attempt in range(retries):
        resp = requests.post(API_URL, json=body, headers=HEADERS, timeout=250)
        if resp.status_code == 429:
            time.sleep(int(resp.headers.get("retry-after", 10)))
            continue
        resp.raise_for_status()
        job = resp.json()["job"]
        if job["status"] == "completed":
            return resp.json()["result"]
        if "retry" in (job["error"] or "").lower():
            time.sleep(5 * (attempt + 1))
            continue
        raise RuntimeError(job["error"])
    raise RuntimeError("Search failed after retries")


def domain(link):
    return (urlparse(link or "").hostname or "").removeprefix("www.")


def aio_profile(query, country):
    """Length, sources and local-site citations of the AI Overview for one search in one country."""
    aio = google_search(query, country)["ai_overview"]
    if not aio:
        return {"query": query, "country": country, "overview": False}
    domains = [domain(s["link"]) for s in aio["sources"]]
    suffixes = LOCAL_SUFFIXES.get(country, ())
    local = [d for d in domains if d.endswith(suffixes)] if suffixes else []
    return {
        "query": query,
        "country": country,
        "overview": True,
        "words": len(aio["text"].split()),
        "sources": len(domains),
        "local_sources": len(local),
        "domains": sorted(set(domains))[:5],
    }


def compare_overviews(query, countries=("us", "gb", "au", "ca", "in"), workers=3):
    with ThreadPoolExecutor(workers) as pool:
        return list(pool.map(lambda c: aio_profile(query, c), countries))

For what is inflation, a repeat of the search gave:

us  171 words, 10 sources, 0 local   blackrock.com, britannica.com, clevelandfed.org, en.wikipedia.org, federalreserve.gov
gb  199 words,  4 sources, 2 local   bankofengland.co.uk, en.wikipedia.org, imf.org, ons.gov.uk
au  118 words,  4 sources, 2 local   commbank.com.au, en.wikipedia.org, federalreserve.gov, rba.gov.au
ca  228 words,  4 sources, 1 local   bankofcanada.ca, en.wikipedia.org, facebook.com, investopedia.com
in  199 words,  4 sources, 0 local   britannica.com, en.wikipedia.org, imf.org, instagram.com

The local-site count for the US is always 0, because the function has no country-code domain to count there.

What to do with it

• Profile each market separately. An overview in one country is not a proxy for another.
• Check whether you are cited per country. The AI Overview tracker takes a country argument.
• Local domains matter in some markets. If you sell in Australia or the UK, the overview leaned on local sites.
• Repeat important checks. Length and sources moved a lot between runs.
• Read the first sentence. It often says which market the answer is for.

What it costs

Every search is 1 credit, in every country, on every queue. Starter is $1.00 per 1,000 requests and the largest pack is $0.30 per 1,000, see pricing.

Failed requests are not charged.

Next steps

You now have numbers on how AI Overviews differ between five markets, a function that profiles any search, and a reason to track each country on its own. The Google Search API page lists every parameter with a full example response.