How to Get Google Keyword Ideas Using Google Autocomplete SERP API

SERP API
Table of Contents

Can You Use a SERP API for Google Autocomplete Data?

Yes. DataForSEO’s SERP API includes a Google Autocomplete API that returns the suggestions Google shows in its search box for a seed keyword, with metadata attached to every suggestion. You send a keyword, a location, and a language, and get back structured JSON with the suggestions, usually up to 15 per request.

Results depend on the location and language you set. Our system ignores search history and other personalized factors, and every result includes a check_url you can open in an incognito window to compare the data with what Google shows.

What Google Autocomplete Data the SERP API Returns

Each response holds one result object for your seed keyword. It carries the request context (keyword, location_code, language_code, check_url, datetime) and an items array with the suggestions. Each item can include the following fields:

  • suggestion: the autocomplete phrase, for example “coffee grinder for espresso”.
  • rank_group and rank_absolute: the suggestion’s position in Google’s dropdown.
  • relevance: Google’s relevance score for the suggestion, from 500 to 2000, where a higher value means a more relevant suggestion. Returned for the chrome and chrome-omni clients.
  • suggestion_type: the suggestion type, such as QUERY. Also returned for chrome and chrome-omni.
  • search_query_url: a Google search URL for the suggestion, localized to your location and language.
  • thumbnail_url: the image Google displays next to some suggestions. Returned for gws-wiz and gws-wiz-serp client types.
  • highlighted: an array of the words Google highlights within the suggestion. Returned for gws-wiz, psy-ab, and gws-wiz-local clients.

There’s one thing you won’t find in the response: search volume. Autocomplete shows what people type, not how often they type it. We’ll add volume and difficulty to the suggestions further down.

How to Get Google Autocomplete Suggestions via API

To keep things straightforward, start with the Live method. You send one POST request and receive the suggestions in the same response, with a turnaround of up to 6 seconds on average.

The request takes these parameters:

  • keyword (required): your seed keyword, up to 700 characters. Our Help Center explains the rules and limitations of the keyword field in detail.
  • location_code or location_name (required): the search location, for example 2840 for the United States.
  • language_code or language_name (required): the search language, for example en.
  • cursor_pointer (optional): the cursor’s position within the keyword. It defaults to the end of the phrase, and changing it is the easiest way to get more ideas out of one seed.
  • client (optional): the Google search client to emulate. Both the suggestions and the returned fields can differ by client.
  • tag (optional): your own label for matching tasks with their results.

The client parameter accepts the following values:

  • chrome: Google Search opened in Chrome. Returns relevance and suggestion_type.
  • chrome-omni: the Chrome address bar. Also returns relevance and suggestion_type.
  • gws-wiz: the Google Search home page. Returns thumbnail_url and highlighted.
  • gws-wiz-serp: the search box on a Google results page. Returns thumbnail_url.
  • gws-wiz-local: Google Local. Returns highlighted.
  • psy-ab: may be used when Google Search is opened in Chrome. Returns highlighted.
  • safari and firefox: Google Search opened in those browsers.
  • img and products-cc: Google Images and Google Shopping search.
  • toolbar and youtube: clients that return XML and JSONP, respectively.

Here’s a request for the seed “coffee grinder” in the United States. It uses the chrome client, so the response includes relevance scores.

Request example:

[
  {
    "keyword": "coffee grinder",
    "location_code": 2840,
    "language_code": "en",
    "client": "chrome"
  }
]

Endpoint used: https://api.dataforseo.com/v3/serp/google/autocomplete/live/advanced (see the Live Google Autocomplete Advanced documentation)

The response below came from a live request. We collapsed 12 of the 15 suggestions to keep it readable.

Response example:

{
  "version": "0.1.20261103",
  "status_code": 20000,
  "status_message": "Ok.",
  "time": "1.2548 sec.",
  "cost": 0.002,
  "tasks_count": 1,
  "tasks_error": 0,
  "tasks": [
    {
      "id": "11111123-7555-0139-0000-c18e5e4390e0",
      "status_code": 20000,
      "status_message": "Ok.",
      "time": "1.1382 sec.",
      "cost": 0.002,
      "result_count": 1,
      "path": ["v3", "serp", "google", "autocomplete", "live", "advanced"],
      "data": {
        "se_type": "autocomplete",
        "se": "google",
        "api": "serp",
        "function": "live",
        "client": "chrome",
        "keyword": "coffee grinder",
        "language_code": "en",
        "location_code": 2840,
        "device": "desktop",
        "os": "windows"
      },
      "result": [
        {
          "keyword": "coffee grinder",
          "type": "autocomplete",
          "se_domain": "google.com",
          "location_code": 2840,
          "language_code": "en",
          "check_url": "https://www.google.com/search?q=coffee+grinder&hl=en&gl=US&uule=w+CAIQIFISCQs2MuSEtepUEUK33kOSuTsc",
          "datetime": "2026-11-11 11:23:17 +00:00",
          "spell": null,
          "refinement_chips": null,
          "item_types": ["autocomplete"],
          "se_results_count": 0,
          "items_count": 15,
          "items": [
            {
              "type": "autocomplete",
              "rank_group": 1,
              "rank_absolute": 1,
              "relevance": 1250,
              "suggestion": "coffee grinder jenks",
              "suggestion_type": "QUERY",
              "search_query_url": "https://www.google.com/search?q=coffee+grinder+jenks&hl=en&gl=US&uule=w+CAIQIFISCQs2MuSEtepUEUK33kOSuTsc",
              "thumbnail_url": null,
              "highlighted": null
            },
            {
              "type": "autocomplete",
              "rank_group": 2,
              "rank_absolute": 2,
              "relevance": 601,
              "suggestion": "coffee grinder for espresso",
              "suggestion_type": "QUERY",
              "search_query_url": "https://www.google.com/search?q=coffee+grinder+for+espresso&hl=en&gl=US&uule=w+CAIQIFISCQs2MuSEtepUEUK33kOSuTsc",
              "thumbnail_url": null,
              "highlighted": null
            },
            {
              "type": "autocomplete",
              "rank_group": 3,
              "rank_absolute": 3,
              "relevance": 600,
              "suggestion": "coffee grinder near me",
              "suggestion_type": "QUERY",
              "search_query_url": "https://www.google.com/search?q=coffee+grinder+near+me&hl=en&gl=US&uule=w+CAIQIFISCQs2MuSEtepUEUK33kOSuTsc",
              "thumbnail_url": null,
              "highlighted": null
            },
            ... 12 more items with the same structure
          ]
        }
      ]
    }
  ]
}

The request took 1.25 seconds and cost $0.002. Take a closer look at the relevance values. The top-scored suggestion, “coffee grinder jenks” (1250), is a narrow query, while “coffee grinder for espresso” scored 601. Relevance helps you sort suggestions inside one response, but it doesn’t measure demand.

If you’d like to check the response structure before spending anything, our Sandbox lets you test the endpoint for free.

Get More Keyword Ideas with the cursor_pointer parameter

Here’s where it gets interesting. Google predicts text at the cursor position, not only at the end of a query. Move the cursor, and Google completes a different part of the phrase.

The cursor_pointer value is the number of characters before the cursor, counting from 0:

|coffee grinder    "cursor_pointer": 0
coffee| grinder    "cursor_pointer": 6
coffee grinder|    "cursor_pointer": 14 (default)

To place the cursor at the start of the phrase, add cursor_pointer to the same request:

Request example:

[
  {
    "keyword": "coffee grinder",
    "location_code": 2840,
    "language_code": "en",
    "client": "chrome",
    "cursor_pointer": 0
  }
]

We ran “coffee grinder” with three cursor positions, and each one surfaced a different kind of modifier:

  • End of the phrase (default): words added after the seed, such as “coffee grinder for espresso”, “coffee grinder near me”, and “coffee grinder walmart”.
  • Position 0: words added before the seed, mostly product types and brands, such as “best coffee grinder”, “manual coffee grinder”, “burr coffee grinder”, and “breville coffee grinder”.
  • Position 6: words inserted between “coffee” and “grinder”, such as “coffee bean grinder” and “coffee maker with grinder”.

The three requests returned 45 suggestions, of which 34 were unique. In our run, each extra position added less than the one before it: position 0 brought 14 new ideas, while position 6 brought 5, since most of its suggestions repeated the default list. In practice, word boundaries are the best place to start: position 0, each space between words, and the end of the phrase.

How to Turn Autocomplete Suggestions into Keyword Ideas

Raw suggestions need some work before they become a keyword plan. Using the Google Autocomplete API for keyword research comes down to four steps.

1. Collect suggestions from several cursor positions. The Python script below sends one Live request per cursor position, merges the results, and drops duplicates. It also records which position surfaced each idea first, so you can see what every extra request added.

import requests

# Use your API credentials from https://app.dataforseo.com/api-access
AUTH = ("login", "password")
URL = "https://api.dataforseo.com/v3/serp/google/autocomplete/live/advanced"

def get_suggestions(keyword, cursor_pointer=None):
    task = {"keyword": keyword, "location_code": 2840,
            "language_code": "en", "client": "chrome"}
    if cursor_pointer is not None:  # leave it out to keep the cursor at the end
        task["cursor_pointer"] = cursor_pointer
    response = requests.post(URL, auth=AUTH, json=[task]).json()
    # a failed task returns "result": null, so fall back to an empty list
    for result in response["tasks"][0]["result"] or []:
        for item in result["items"] or []:
            yield item["suggestion"]

seed = "coffee grinder"
# end of the phrase, start of the phrase, and every space between words
positions = [None, 0] + [i for i, char in enumerate(seed) if char == " "]

ideas = {}
for position in positions:
    for suggestion in get_suggestions(seed, position):
        ideas.setdefault(suggestion, position)  # keep the position that found it first

print(f"{len(ideas)} unique ideas from {len(positions)} requests")
for suggestion, position in ideas.items():
    print(suggestion, "| cursor_pointer:", "end" if position is None else position)

On our live data from November 2026, the three requests produced 34 unique ideas at a total cost of $0.006.

2. Remove suggestions that don’t match your topic. Some suggestions share your words but not your intent. “Coffee grinder dance move” refers to a breakdancing move. Retailer modifiers (“walmart”, “amazon”, “target”) and local ones (“near me”, “nearby”) call for judgment: they matter if you sell grinders and add little to a review site.

3. Group the rest by modifier. Once the list is clean, patterns show up quickly. The coffee grinder run splits into four groups:

  • Product type: burr, manual, hand, electric.
  • Brand: Breville, Baratza, Cuisinart, KitchenAid, OXO, Fellow, and others.
  • Use case: “coffee grinder for espresso”, “coffee maker with grinder”.
  • Evaluation: “best coffee grinder”, “coffee grinder reviews”.

Each group points to a content type. Product types fit category pages or explainers, brands fit comparison pages, use cases fit dedicated guides, and evaluation queries fit buying guides.

4. Add search volume and difficulty. Autocomplete tells you how people phrase a search. To decide which ideas deserve a page, you also need demand and competition. Send your shortlist to the Keyword Overview endpoint of our DataForSEO Labs API, which returns search volume, keyword difficulty, and search intent for every keyword in one request.

Endpoint used: https://api.dataforseo.com/v3/dataforseo_labs/google/keyword_overview/live

Request example:

[
  {
    "keywords": [
      "coffee maker with grinder",
      "coffee bean grinder",
      "burr coffee grinder",
      "best coffee grinder",
      "manual coffee grinder",
      "electric coffee grinder",
      "coffee grinder for espresso",
      "breville coffee grinder",
      "baratza coffee grinder",
      "coffee grinder reviews"
    ],
    "location_code": 2840,
    "language_code": "en"
  }
]

Here’s what came back for the top of the shortlist, with monthly search volume in the United States and keyword difficulty on a scale from 0 to 100:

  • “coffee maker with grinder”: 18,100 searches, difficulty 1
  • “coffee bean grinder”: 18,100 searches, difficulty 28
  • “burr coffee grinder”: 18,100 searches, difficulty 24
  • “best coffee grinder”: 14,800 searches, difficulty 24
  • “manual coffee grinder”: 12,100 searches, difficulty 4
  • “electric coffee grinder”: 9,900 searches, difficulty 11
  • “coffee grinder for espresso”: 8,100 searches, difficulty 20

Notice where the strongest ideas came from. “Coffee maker with grinder” pairs 18,100 monthly searches with a difficulty of 1, and it appeared only at cursor position 6. Position 0 returned “manual coffee grinder” (difficulty 4) in the word order people search with, while the default list only had “coffee grinder manual”, which could equally mean an instruction booklet. A single default request would have missed both.

Volume also shifts with the season, so look at the monthly breakdown before you commit to a topic. The same response includes monthly_searches: for “burr coffee grinder”, it ranged from 49,500 in November 2025 to 8,100 in September 2026.

If you need a keyword suggestions API that returns metrics out of the box, the Keyword Suggestions and Keyword Ideas endpoints in DataForSEO Labs return keywords with volume and difficulty already attached. Autocomplete complements them with current phrasing taken straight from Google’s search box.

To run this workflow on your own seed keywords, try for free and test every request from this article in the API playground first. If a client value or cursor position behaves in an unexpected way, our support team will do their best to help.

Was this article helpful?