Track Whether Your Brand Is Cited in ChatGPT, Gemini, Google AI Mode and AI Overviews
Four actors, one job: ask the questions your buyers ask ChatGPT, Gemini, Google AI Mode and Google AI Overviews, and record whether your brand is named and your site is cited.
The actor referenced in this article. Pay only for results delivered.
Your buyers no longer stop at ten blue links. They ask ChatGPT which CRM to pick, read the AI Overview above Google’s results, or open Google’s AI Mode tab and follow the conversation. Each of those answers names a few brands and cites a few pages. If your brand is not among them, you lose the click before the buyer ever sees your site, and you may never find out. Checking by hand does not scale: the answers change from one ask to the next, differ by country, and there are four places to look.
Try it live: Google AI Overviews Scraper. Pay per result delivered.
This guide covers four actors that ask those questions for you and return one row per answer, with the full text, the cited sources and a check of your brands and domains against it. Run them on a schedule and you get a record of when you are named, when you are cited, and which sites the AI trusts instead of you.
Which actor reads which answer
| Actor | What it reads | Best for | Gold price per 1,000 rows |
|---|---|---|---|
| Google AI Overviews Scraper | The AI Overview block on the normal Google results page | Whether Google shows an overview for your keywords, who it cites, and your organic rank on the same page | $8.00 |
| Google AI Mode Scraper | Google’s separate AI Mode tab | Longer conversational answers, product cards and the follow up questions Google asks back | $8.70 |
| ChatGPT Search Scraper | chatgpt.com in a fresh logged out chat, Web search on | What ChatGPT recommends and links for buying questions | $9.00 (US), $75.00 (other countries) |
| Gemini Search Scraper | gemini.google.com as a logged out visitor | Share of voice across repeated asks, cited versus searched sources | $2.70 |
A short way to choose:
- You get most of your traffic from Google search. Start with AI Overviews. It runs on your existing keyword list, and the same row holds the organic results, so “cited but not ranking” and “ranking but not cited” show up without a second tool.
- You sell products people compare. Add AI Mode. Its product cards come back with name, price, merchant, rating and review count, in Google’s order.
- Your buyers research in ChatGPT. Use the ChatGPT Search Scraper with your brand and competitors in
brandsToTrack. Each row says whether each brand was mentioned, linked or cited. - You want a steady number for a report. Gemini is the cheapest per answer and can ask each prompt up to 10 times, which smooths out the answer to answer swings.
What one row holds
All four return one row per question and drop free status rows (summary, info, error) into the same dataset with a _type field, so filter on _type being absent.
- Google AI Overviews:
ai_overview_shown,ai_overview_text(the whole answer, including the part behind Show more),sourceswithposition,url,domainandcited_in_answer,cited_domains,tracked_domainswithcited,source_positionandorganic_position,tracked_brandswithnamed_in_answerandmentions,citation_state, andorganic_results,people_also_askandrelated_searches. - Google AI Mode:
answer_text,answer_markdown,answer_blocks,follow_up_questions,sources,products(price, merchant, rating, reviews),tracked_domains,tracked_brandsandcitation_state. - ChatGPT:
answer_text,answer_markdown,web_search_used,sourceswithposition,domainandsection,cited_domains,entities(brands and products ChatGPT named), andtracked_brandswithmentioned,mention_count,first_position,rank_among_tracked,citedandlinked. - Gemini:
answer_text,answer_plain_text,sourcessplit intosection: "cited"(withcited_text, the sentence it supports) andsection: "searched",cited_domains, andbrand_mentionswithmentioned,mention_count,rank,first_sentenceandcited. ASHARE_OF_VOICErecord sums up the whole run.
The citation field is the one to chart. citation_state on the Google actors reads cited, others_cited, overview_no_sources or no_overview (AI Mode uses cited, others_cited or answer_no_sources). The day a query flips from no_overview to others_cited is the day it starts costing you clicks.
The inputs
Google AI Overviews, the main actor of this guide:
{
"queries": ["best crm for small business", "hubspot vs zoho", "what is a sales funnel"],
"countryCode": "us",
"languageCode": "en",
"trackDomains": ["yourbrand.com", "competitor.com"],
"trackBrands": ["Your Brand", "Competitor"],
"includeOrganic": true,
"resolveLinks": true
}
Google AI Mode takes the same shape. To check several markets in one run, use searches:
{
"searches": [
{ "query": "best crm for startups", "countryCode": "gb", "languageCode": "en" },
{ "query": "bestes crm für kleine unternehmen", "countryCode": "de", "languageCode": "de" }
],
"trackDomains": ["hubspot.com"],
"trackBrands": ["HubSpot"]
}
ChatGPT joins a brand’s spellings and its domain with a bar:
{
"prompts": ["best crm software for small business", "hubspot alternatives"],
"brandsToTrack": ["HubSpot", "Zoho|Zoho CRM|zoho.com", "Salesforce"],
"country": "US",
"webSearch": true,
"maxConcurrency": 2
}
Gemini lists each brand as name, other names and website, separated by commas:
{
"prompts": ["best crm for a small business", "is hubspot worth the price"],
"brands": ["HubSpot, hubspot.com", "Zoho CRM, Zoho, zoho.com"],
"country": "US",
"language": "en",
"repeatEachPrompt": 3
}
One script for all four
This asks the same questions on all four surfaces and prints, for each brand, how often it was named and how often its site was cited.
from collections import defaultdict
from apify_client import ApifyClient
client = ApifyClient("YOUR_APIFY_TOKEN")
QUESTIONS = ["best crm for small business", "hubspot alternatives", "which crm is easiest to set up"]
BRANDS = {"HubSpot": "hubspot.com", "Zoho": "zoho.com", "Pipedrive": "pipedrive.com"}
def answers(slug, run_input):
run = client.actor(f"themineworks/{slug}").call(run_input=run_input)
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
if "_type" not in item: # skip the free summary, info and error rows
yield item
results = [] # (engine, brand, named, cited)
for engine, slug in [("AI Overviews", "google-ai-overviews-scraper"), ("AI Mode", "google-ai-mode-scraper")]:
for item in answers(slug, {
"queries": QUESTIONS,
"countryCode": "us",
"languageCode": "en",
"trackDomains": list(BRANDS.values()),
"trackBrands": list(BRANDS.keys()),
}):
if item.get("check_status") != "ok":
continue # deferred or partial: unknown, so not counted as a loss
cited = {d["domain"]: d.get("cited", False) for d in item.get("tracked_domains", [])}
named = {b["brand"]: b.get("named_in_answer", False) for b in item.get("tracked_brands", [])}
for name, domain in BRANDS.items():
results.append((engine, name, named.get(name, False), cited.get(domain, False)))
for item in answers("chatgpt-search-scraper", {
"prompts": QUESTIONS,
"brandsToTrack": [f"{name}|{domain}" for name, domain in BRANDS.items()],
"country": "US",
"webSearch": True,
}):
for b in item.get("tracked_brands", []):
results.append(("ChatGPT", b["brand"], b.get("mentioned", False), b.get("cited", False) or b.get("linked", False)))
for item in answers("gemini-search-scraper", {
"prompts": QUESTIONS,
"brands": [f"{name}, {domain}" for name, domain in BRANDS.items()],
"country": "US",
"language": "en",
}):
for b in item.get("brand_mentions", []):
results.append(("Gemini", b["brand"], b.get("mentioned", False), b.get("cited", False)))
totals = defaultdict(lambda: [0, 0, 0]) # answers, named, cited
for engine, brand, named, cited in results:
t = totals[(engine, brand)]
t[0] += 1
t[1] += int(named)
t[2] += int(cited)
for (engine, brand), (n, named, cited) in sorted(totals.items()):
print(f"{engine:13} {brand:10} named in {named}/{n} cited in {cited}/{n}")
Save the results list with a date each week and you have a trend. For the pages to win a citation from, count domain across sources where the actor marks a source as cited: those are the review sites and lists the AI leans on for your category.
What it costs
These four cost more per row than most actors on this site, and it is worth being plain about that. A row here is a whole AI answer that has to be written before it can be read, not a line in a listing. Prices per 1,000 rows:
| Actor | Event | Free plan | Gold and above | Start fee per run |
|---|---|---|---|---|
| Google AI Overviews | ai-overview or query-no-overview | $13.00 | $8.00 | $0.0085 |
| Google AI Mode | ai-mode-answer | $12.00 | $8.70 | $0.003 |
| ChatGPT, asked from the US | chatgpt-answer | $15.00 | $9.00 | $0.003 |
| ChatGPT, any other country | local-country-answer | $95.00 | $75.00 | $0.003 |
| Gemini | answer-scraped | $4.50 | $2.70 | $0.003 |
In every case the start fee is charged only after the first row reaches your dataset, so a run that gets no answers costs nothing. Questions that got no answer, failed tries, the summary, info and error rows, and runs past your max cost are not charged. On AI Overviews a deferred row is free, but a query Google answered with no overview is charged like one with an overview, because “no overview” is a real result worth tracking.
A worked example: 50 questions, once a week, on all four surfaces from the US. On the Free plan that is $0.6585 for AI Overviews, $0.603 for AI Mode, $0.753 for ChatGPT and $0.228 for Gemini, about $2.24 a week. On Gold and above the same run is about $1.44. Setting Gemini’s repeatEachPrompt to 3 triples its line. Asking ChatGPT from the UK instead of the US moves its line from $0.75 to $4.75 on the Free plan, so keep non US prompts to the questions that really differ by market.
Limits worth knowing
- Answers change between asks. One reading is a sample. Track
cited_domains,citation_stateand brand rank over several runs. - Signed out views only. ChatGPT returns the guest experience, and Gemini answered with
3.5 Flash-Litein every test. Paying users of either can see different answers. - Not every query gets an answer. On AI Overviews 12% of US queries in our tests were deferred. On AI Mode 32 of 38 queries got a whole answer, and the rest were mostly long finance and comparison answers Google did not finish. Both are free and worth rerunning later.
- Brand checks are text matches. They look for the names and domains you list, so give distinctive spellings and the domain, not a common word.
- AI Overviews reads the desktop page. There are no mobile results.
- Sources are not always there. ChatGPT and Gemini decide per question whether to search the web. Plain knowledge questions often come back with no sources at all.
Related
- Google AI Overviews Scraper, Google AI Mode Scraper, ChatGPT Search Scraper and Gemini Search Scraper
- Using Google Trends to Find Untapped SEO Opportunities
- How to Build a Competitor Intelligence System Using Web Scrapers
- Track Which Sites Rank in Google Images for Your Keywords
- Audit Core Web Vitals for Hundreds of URLs at Once
Explore the scraper referenced in this article: inputs, outputs, and pricing, then run it on Apify.
Frequently asked questions
Which of the four should I run first? +
Google AI Overviews, if your traffic comes from Google search. It showed on 81% of the US queries in our tests, it returns the organic results of the same page in the same row, and it reports deferred pages as unknown instead of counting them as a lost overview.
Why are these actors more expensive per row than a normal scraper? +
Each row is a whole AI answer that has to be written before it can be read. Google AI Mode takes 15 to 70 seconds per answer, and ChatGPT needs a real browser and, outside the US, a residential connection in the country you ask from. On Gold and above a row costs between $2.70 and $75 per 1,000.
Do I need accounts or API keys for ChatGPT, Gemini or Google? +
No. All four read what a signed out visitor sees. That also means the answers come from the models those services give logged out visitors, which can differ from what a paying subscriber gets.
Why does the same question give a different answer next week? +
All four services write a new answer each time. Track the pattern over several runs: which domains are cited, whether your brand is named, and its rank, rather than the exact text. Gemini can also ask one prompt up to 10 times in a run.
Audit Core Web Vitals for Hundreds of URLs at Once
Run Lighthouse on a whole list of pages, mobile and desktop, and get scores, Core Web Vitals and the top fixes as one spreadsheet row per page, with or without your own Google key.
Build a Labeled Image Dataset From Google Images Search Terms
Turn one search term per class into a labeled image manifest with Google Images Scraper: full size URLs, width, height and the source page to check.
Track Which Sites Rank in Google Images for Your Keywords
Use Google Images Scraper to see which domains hold the top image positions for your keywords in each country, then rerun weekly to catch movement.
Autotrader UK and Canada Data: Used Car Prices, Mileage and Dealer Contacts
Pull Autotrader UK and AutoTrader.ca listings into one sheet with price, mileage, specs and dealer phone, for pricing work, stock checks and dealer lead lists.