Gemini Answers Scraper

Returns: one row per prompt with prompt, answerText, citations, sources, citedDomains, usedWebSearch, brandMentions. Input: {"prompts": ["What are the best CRM tools for small businesses?"], "brands": ["HubSpot", "Salesforce"], "maxItems": 50} Best for: AI-visibility and GEO tracking, brand and competitor monitoring in AI answers, finding which sites Gemini cites.

{
  "id": "r_e6bc98e1d3ab2126",
  "prompt": "Write a two-line poem about the ocean.",
  "answerText": "Endless blue waves crash against the shore,\nWhispering secrets of the deep ocean floor.",
  "language": "en",
  "usedWebSearch": false,
  "citationCount": 0,
  "sourceCount": 0
}

What does Gemini Answers Scraper do?

Gemini Answers Scraper asks Google Gemini the questions you give it and saves each answer as one dataset row. It works as a Gemini scraper and a simple Gemini API for people who need the answers as data: the full answer text, the inline citations, the list of web sources Gemini searched, and the domains behind them. No login and no API key are needed. Each prompt is a fresh anonymous conversation, so earlier prompts do not influence later ones.

The main use is AI-visibility tracking, also called generative engine optimisation (GEO). You list the questions your customers ask, add the brand names you care about (yours and your competitors), and the Actor tells you for every answer whether each brand is mentioned, how many times, and whether the brand name appears in a cited domain. Run it on a schedule and you can compare the results over time.

It is also useful for research: collecting Gemini answers in many languages, seeing which publications and review sites Gemini relies on for a topic, and building datasets of question and answer pairs.

Answers are not deterministic. Gemini can word an answer differently, cite different sources or skip web search on the same prompt in two runs. Treat one run as a sample and run important prompts more than once.

What data can you get from Gemini?

Every row has the fields below. The citation and source fields are empty arrays when Gemini answered without searching the web, which happens for creative or purely conversational prompts.

Field	Description	Example
id	Gemini response id (r_...), unique per answer	r_e6b596c8d50be85f
prompt	The question sent to Gemini	What are the best CRM tools?
language	Language code from the input	en
answerText	Full answer as markdown, may contain code blocks	When choosing the best CRM...
citations	Inline citations with url, title, domain, siteName, snippet and citedText	[{"url": "https://flowqi.com/..."}]
citationCount	Number of inline citations	11
sources	The web source list with url, title and domain	[{"domain": "pcmag.com"}]
sourceCount	Number of sources	6
citedDomains	Sorted unique domains across citations and sources	["pcmag.com"]
domainSummary	Per domain citation and source counts plus distinct URLs	[{"domain":"pcmag.com","citationCount":2,"sourceCount":1,"urls":["https://pcmag.com/example"]}]
usedWebSearch	True when the answer has citations or sources	true
brandMentions	Per brand: mentioned, count, cited, mentionRank and firstMentionContext (only when brands are given)	[{"brand": "HubSpot", "mentioned": true}]
conversationId	Gemini conversation id (c_...)	c_81994921342a1ebe
responseId	Gemini response id	r_e6b596c8d50be85f
scrapedAt	ISO 8601 time the answer was received	2026-10-16T00:07:13.487Z
sourceUrl	Gemini endpoint the answer came from	https://gemini.google.com/app

How to use Gemini Answers Scraper

1. Open the Actor on Apify and go to the Input tab.
2. Enter one question per line under Prompts. Write them the way a customer would ask Gemini.
3. Optionally add brand and competitor names under Brands to track, and set the Language code (for example de or ja).
4. Set Maximum results to the number of prompts you want answered, then press Start.
5. When the run ends, open the Output tab and export the dataset as JSON, CSV, Excel or XML, or read it through the API.

Tips for good results:

1. Ask full questions that include the buying context, such as "best CRM tools for small businesses", because they give Gemini useful context. Web search and citations are not guaranteed.
2. Put competitor names in the brands list too. The brandMentions field then shows mention counts and first-appearance order per answer.
3. Keep runs modest. The Actor answers prompts one after another with a short pause, and each answer takes roughly 5 to 20 seconds, so 50 prompts take around 15 minutes.
4. Duplicate prompts in the input are answered once.

Input

Field	Type	Required	Description
prompts	array of strings	Yes	Questions to ask Gemini, one dataset item per prompt. Up to 4000 characters each.
brands	array of strings	No	Brand names to flag in each answer. Adds brandMentions to every item.
language	string	No	Language code for Gemini, for example en, de, ja, pt-BR. Default en; malformed codes fall back to en with a warning.
maxItems	integer	No	Maximum number of prompts answered, 1-10000, default 50. Shared across all prompts.
proxyConfiguration	object	No	Proxy settings. Default: Apify Proxy (datacenter).

{
    "prompts": [
        "What are the best CRM tools for small businesses?",
        "Which project management software do you recommend for remote teams?"
    ],
    "brands": ["HubSpot", "Asana", "Pipedrive"],
    "language": "en",
    "maxItems": 50,
    "proxyConfiguration": { "useApifyProxy": true }
}

How brand matching works: a brand counts as mentioned when its name appears in the answer text, ignoring case. count is the number of occurrences. cited is true when the lowercased brand name appears inside any cited or source domain, for example hubspot inside hubspot.com.

The additive mentionRank field orders only the tracked brands by their first appearance in the answer, starting at one. It does not describe a numbered recommendation, sentiment, quality or market position. Two names first appearing at the same position share a rank. An absent brand has a null rank and null firstMentionContext. The context is a verbatim excerpt with up to 100 characters before and after the matching name, so analysts can check how Gemini used the name without opening the full answer. Existing mentioned, count and cited values keep their original meanings.

domainSummary groups the same citations and sources by domain. Each entry contains domain, citationCount, sourceCount and urls. Repeated inline citations are counted separately; the URLs list is deduplicated across both lists. This makes it possible to distinguish a publisher supporting many answer sentences from a site that appears only in the search source list. Domains remain alphabetically ordered, rather than presented as a recommendation ranking. An answer without web sources has an empty domain summary. These calculations use the returned answer without another request or an AI enrichment call.

An empty prompts list ends the run with a clear input error. A prompt that Gemini does not answer after several retries is logged and skipped, and the rest of the batch continues. If nothing could be answered the run still ends successfully with zero items and a status message explaining why.

Gemini's exact English service-error message, “I encountered an error doing what you asked. Could you try again?”, triggers the existing retry path instead of being saved as an answer. Ordinary answers, short creative text and refusals remain unchanged. Other service-error wording is not automatically classified.

Output

Each prompt becomes one item. Here is a real item for a prompt where Gemini did not search the web:

{
    "id": "r_24cd5d14ca6605ec",
    "prompt": "Write a haiku about autumn leaves",
    "language": "en",
    "answerText": "Golden leaves drift down,\nWhispering of cooler days,\nNature falls asleep.",
    "citations": [],
    "citationCount": 0,
    "sources": [],
    "sourceCount": 0,
    "citedDomains": [],
    "usedWebSearch": false,
    "conversationId": "c_5032e07ec26e5b7a",
    "responseId": "r_24cd5d14ca6605ec",
    "scrapedAt": "2026-10-16T00:07:04.728Z",
    "sourceUrl": "https://gemini.google.com/app",
    "brandMentions": [
        {
            "brand": "HubSpot",
            "mentioned": false,
            "count": 0,
            "cited": false
        }
    ]
}

The following selected fields from a local comparison answer show the added visibility metrics and domain summaries:

{
  "id": "r_12d3deb0c489e1b9",
  "prompt": "Compare Notion and Trello for planning a small team project.",
  "brandMentions": [
    {
      "brand": "Notion",
      "mentioned": true,
      "count": 6,
      "cited": false,
      "mentionRank": 1,
      "firstMentionContext": "When planning a small team project, choosing between **Notion** and **Trello** comes down to a core trade-off: **Simplicity vs. Flexibility**. \n\nBoth tools are p"
    },
    {
      "brand": "Trello",
      "mentioned": true,
      "count": 6,
      "cited": false,
      "mentionRank": 2,
      "firstMentionContext": "When planning a small team project, choosing between **Notion** and **Trello** comes down to a core trade-off: **Simplicity vs. Flexibility**. \n\nBoth tools are popular, but the"
    },
    {
      "brand": "MissingBrandXYZ",
      "mentioned": false,
      "count": 0,
      "cited": false,
      "mentionRank": null,
      "firstMentionContext": null
    }
  ],
  "domainSummary": [
    {
      "domain": "aicomparison.ai",
      "citationCount": 5,
      "sourceCount": 1,
      "urls": [
        "https://aicomparison.ai/notion-vs-trello/"
      ]
    },
    {
      "domain": "fastlancer.org",
      "citationCount": 4,
      "sourceCount": 1,
      "urls": [
        "https://www.fastlancer.org/en/fastlancer-blog/notion-vs-trello/"
      ]
    },
    {
      "domain": "rambox.app",
      "citationCount": 2,
      "sourceCount": 1,
      "urls": [
        "https://rambox.app/blog/notion-vs-trello-comparison/"
      ]
    },
    {
      "domain": "toolfinder.com",
      "citationCount": 4,
      "sourceCount": 1,
      "urls": [
        "https://toolfinder.com/comparisons/trello-vs-notion-for-project-management"
      ]
    }
  ]
}

The answer text is the visible markdown of the answer. When Gemini runs code to answer, for example for arithmetic, the text can consist mostly of a code block. Download the dataset as JSON, CSV, Excel, XML or RSS from the Output tab or the API.

How much does it cost to scrape Gemini?

Gemini Answers Scraper is charged per result, which means per prompt answered. You pay only for the answers returned, and maxItems caps how many that can be. If you set a maximum cost per run, the Actor stops cleanly when the limit is reached and keeps everything saved so far. See the Pricing tab for the current price.

Use with AI agents (MCP)

Connect Gemini Answers Scraper to Claude, Cursor or any MCP client through the Apify MCP server:

{
    "mcpServers": {
        "apify": {
            "url": "https://mcp.apify.com/?tools=axiomworks/gemini-answers-scraper"
        }
    }
}

Example prompts:

1. "Ask Gemini 'best CRM tools for small businesses' and tell me whether HubSpot or Pipedrive is cited."
2. "Run these 20 buying questions through Gemini and list the domains it cites most often."
3. "Check how Gemini answers 'Was sind die besten Online-Banken in Deutschland?' in German and flag N26 and DKB."

Integrations and API

Run Gemini Answers Scraper from your own code with the Apify API. Set APIFY_TOKEN to your Apify API token (send it in the Authorization header, never in the URL).

Python

from apify_client import ApifyClient

client = ApifyClient("<YOUR_API_TOKEN>")
run = client.actor("axiomworks/gemini-answers-scraper").call(run_input={
    "prompts": ["What are the best CRM tools for small businesses?"],
    "brands": ["HubSpot", "Salesforce"],
    "maxItems": 50,
})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["prompt"], item["citedDomains"])

JavaScript

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: '<YOUR_API_TOKEN>' });
const run = await client.actor('axiomworks/gemini-answers-scraper').call({
    prompts: ['What are the best CRM tools for small businesses?'],
    brands: ['HubSpot', 'Salesforce'],
    maxItems: 50,
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();

cURL

curl -X POST "https://api.apify.com/v2/acts/axiomworks~gemini-answers-scraper/run-sync-get-dataset-items" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"prompts": ["What are the best CRM tools for small businesses?"], "brands": ["HubSpot"], "maxItems": 50}'

On the Integrations tab you can connect the Actor to Make, Zapier, n8n, Google Sheets and webhooks, and schedule runs, for example weekly, to track how answers change.

FAQ

Do I need a Google account or API key? No. The Actor talks to the public Gemini web endpoint as an anonymous visitor.

Why does the same prompt give different answers each time? Gemini generates a new answer on every request. The sources it searches and cites also change. Run prompts several times and compare the share of runs in which a brand appears.

Why are citations and sources empty for some prompts? Gemini only searches the web when it decides the question needs fresh information. For creative or general prompts it answers from its own knowledge, so usedWebSearch is false and the arrays are empty.

Can I ask in other languages? Yes. Set language to a code such as de or ja and write the prompts in that language. Brand names can use any script. If the optional language value is malformed, the Actor logs a warning and uses en; the output records this effective language. Missing or blank prompts still fail input validation.

Why did a run return fewer items than prompts? maxItems limits the number of prompts answered. Prompts that Gemini did not answer after retries are skipped and listed in the log.

Does it keep a conversation going across prompts? No. Every prompt starts a new anonymous conversation, so there is no follow-up context between prompts.

What if Gemini starts requiring a login? This Actor does not log in. If Gemini stops answering anonymous requests, runs finish with zero items and a status message. Report it on the Issues tab.

Is it legal to scrape Gemini?

The Actor sends questions to the public Gemini web page without logging in and stores the answers it receives. You are responsible for how you use the output, including any personal data it might contain, and for following the terms of the services and laws that apply to you. This is not legal advice.

Gemini Answers Scraper is an independent tool. It is not affiliated with, endorsed by or sponsored by Google or Gemini.

Feedback

Found a bug or need a field that isn't here? Open an issue on the Issues tab and we'll look at it.