Google builds two results pages for the same query: one for phones and one for computers. Both come from the same index, but the layout, the features around the results and sometimes the order differ. With the QuanticData SERP API you choose the page with one request parameter, device, and the response says which page came back.
On 20 November 2026 we requested both pages for two queries, from the same city, and compared the top ten. The numbers are below, followed by the parameter and the response fields, all taken from the live API reference and from those calls.
How Google results differ by device
Start with what does not differ. Google does not keep one index for phones and another for computers. Its documentation says that it uses the mobile version of a site's content, crawled with its smartphone agent, for indexing and ranking, and calls this mobile-first indexing. So a page is not ranked from two different copies depending on who is searching.
What differs is the page Google assembles for the searcher:
- Layout. A phone shows one narrow column. Fewer results fit on the first screen, so the blocks above the classic results matter more.
- How a result is displayed. Breadcrumbs are cut shorter, and sites that have a separate mobile address are shown with it.
- Features. Related searches and video results are placed and sized differently.
- Order. Usually close, sometimes not, as the runs below show.
Two things that are not the device change results even more: the place the search comes from, and the moment. That is why a comparison has to hold the place fixed.
What we measured: two queries, phone and computer
Each query was sent twice on 20 November 2026, once with device set to desktop and once to mobile, with engine Google, country us, language en and location "Austin, Texas, United States", asking for ten results. We then compared the two top tens.
| Query | Same result at the same rank | Pages in both top tens | Related searches, computer / phone |
|---|---|---|---|
| coffee shop near me | 9 of 10 | 9 | 8 / 11 |
| best running shoes | 4 of 10 | 8 | 8 / 9 |
Coffee shop near me. The first nine results were the same pages in the same order. The tenth differed: a local guide on the computer page, an Instagram post on the phone page. The display differed more than the order: Yelp appeared as www.yelp.com on the computer page and as m.yelp.com on the phone page.
Best running shoes. Eight pages were in both lists, but only ranks 3, 4, 5 and 7 matched. The first two results swapped places, and each page had two results the other lacked, one of them a different YouTube video. Breadcrumbs were shorter on the phone: runrepeat.com › guides › best-running-shoes became runrepeat.com › guides › best-running-s....
Two things held for both queries. The phone page carried more related searches than the computer page, 2 of 2 times. And Google showed an AI Overview on both pages or on neither.
Two queries are an illustration, not a study: one differed in a single position, the other in six.
How to request each page: the device parameter
The SERP API takes the device in the request body. These are the values in the API reference:
device | Page you get |
|---|---|
desktop | The computer page. This is the default when the field is omitted. |
mobile, ios, android | The phone page. |
The request we ran for the phone page, as an HTTP call:
curl -s -X POST https://api.quanticdata.io/v1/scraper/serp \
-H "Authorization: Bearer $QD_API_KEY" \
-H "Content-Type: application/json" \
-d '{"query":"best running shoes","engine":"google","country":"us","lang":"en","location":"Austin, Texas, United States","device":"mobile","num":10}'
We sent the calls in this post through the search tool of the QuanticData MCP server, which takes the same fields; the endpoint and the body above are the HTTP form of the same request. Change "device":"mobile" to "desktop", or leave it out, for the computer page. If you have not called the API before, how to use an API key covers the header.
The fields that matter for a device comparison:
| Field | What it does |
|---|---|
device | desktop, mobile, ios or android |
location | A place in words, such as "Austin, Texas, United States". The service encodes it into Google's location token for you |
uule | The encoded location token, or raw coordinates as lat,lon |
country, lang | The country of the search and the language of the interface |
num, page | How many results to ask for, up to 100, and which results page |
Three behaviours from the reference are useful when you track both devices. Every page of one query comes from the same device and browser, so a request for more than ten results is not a mix of the two layouts. Computer and phone results are cached separately, so one never answers for the other. And the multi-page search job, /scraper/serp/bulk, takes the same device field.
The response: which page came back
The answer repeats what you asked for and reports what Google served. From our phone request for "best running shoes":
"search_parameters": {
"engine": "google", "q": "best running shoes", "search_type": "search",
"device": "mobile", "country": "us", "language": "en", "page": 1
},
"search_metadata": {
"status": "Success", "engine": "google", ...
"served_device": "mobile",
"served_platform": "android",
"served_browser": "edge"
}
| Field | Meaning |
|---|---|
search_parameters.device | The device you asked for |
search_metadata.served_device | The device the returned page was served for, read from the page itself: phone layout or computer layout |
search_metadata.served_platform | The device the page was fetched as |
search_metadata.served_browser | The browser it was fetched as |
In our runs every phone request came back with served_device: mobile and every computer request with served_device: desktop: 4 of 4 calls returned the layout that was asked for. Because served_device is read from the page and not copied from the request, it is the field to store next to each rank.
The results themselves have the same shape on both pages: organic with rank, title, link, display_link and description, then related_searches and, when Google shows one, ai_overview. Code written for one device reads the other without changes; the rest of the fields are explained in how to use a SERP API.
How to compare the two pages properly
- Fix the place. Set
locationoruuleon both requests. Google ranks by city, and without a location Google places the search from the request's address. - Send both requests together. Results move during the day. Ours were less than a second apart.
- Repeat before you conclude. One pair of requests shows a difference; several pairs show whether the device causes it.
- Compare pages, not positions alone. Count the results that appear in both lists as well as the ranks that match.
- Store the device with every rank. One row per keyword, place, device and date, with
served_devicefrom the response.
The page on rank tracking by city and device has a complete example for a list of keywords, and Google results in Python shows the loop in a script.
Which device to track
Track the device your visitors use. Your analytics already has the split between phones and computers for search traffic. A local business, a restaurant or a shop with walk-in customers is searched mostly from phones and should watch the phone page first. Software bought at a desk is searched mostly from computers. When the split is close, track both: the first query in our table would have told the same story on either device, the second would not.
Each device is a separate search, so tracking both doubles the number of calls. What a search costs is on the SERP API pricing page. For the path a request takes from the query to the JSON, see how a SERP API works.


