Google Trends Scraper: Queries, Regions & Trending Now

Extract Google Trends data at scale: search interest over time, interest by country and region, related queries (top + rising), and real-time Trending Now searches, up to 100 keywords per run, any country, any timeframe.

If it returns nothing, you pay nothing

The usual way a Google Trends scraper wastes your money is that it runs for twenty minutes, hits Google's rate limits, and hands you an empty dataset, after billing you the whole time.

This one is built so that can't happen:

• You are charged per row of real data. Billing on dataset writes is switched off entirely, so a failed keyword produces an error row that costs $0.00. No data, no charge.
• Memory is capped at 512 MB (most Actors default to 4 GB). Compute is billed as memory × time, so a slow run here costs a fraction of what the same run costs elsewhere.
• No headless browser. Plain HTTP against Google's own endpoints, runs finish in about a minute, not twenty.
• One bad keyword never kills the run. Each keyword is isolated; the other 99 still return.
• Your run charge limit is honored. Set a maximum charge and the Actor stops cleanly at it instead of burning compute for revenue it can't collect.

Rotating proxies and staged backoff handle Google's rate limiting, which is what causes most empty results in the first place.

When to run it (measured, 2026-07-29)

Google gates trends.google.com/trends/api/* behind a session cookie, and it gates it harder during US and European working hours. This is not a claim about this Actor, it is what the endpoint does to everyone, which is why pytrends 429, pytrends 404 and gtrendsR: status code was not 200 are perennial Stack Overflow questions.

What was measured, on the same keyword, the same day:

03:17 UTC	full results in 41 seconds
12:22-13:37 UTC	429 on every attempt
Waiting 30s, 60s, 120s between tries	still 429
Datacenter proxy, residential proxy, and no proxy at all	identical: all blocked
The same request from an ordinary home connection, same minute	200

Two things follow, and they are worth knowing before you build on any Google Trends tool:

• Backoff does not buy you through it. Two full minutes of waiting changed nothing. An Actor that "retries harder" is just billing you for the wait.
• Switching proxy tier does not help. Datacenter, residential and Apify's own egress all fail identically, while an ordinary home connection succeeds in the same minute. Google is judging the address, and every address a cloud platform can offer for this endpoint is already spent - the residential pool included, because it is shared by thousands of scrapers. Paying for residential bandwidth here buys nothing.

If you need it during the day: bring your own proxy

The Proxy input accepts Custom proxy URLs. That is the one setting here that changes the outcome, because the block is on the address and every address a cloud platform can offer is already spent. A proxy on IPs that are not shared with thousands of other scrapers gets the same 200 an ordinary home connection gets.

{
  "keywords": ["matcha"],
  "proxyConfiguration": { "useApifyProxy": false, "proxyUrls": ["http://user:pass@your-proxy:8000"] }
}

Nothing else in the input moves the needle. Turning Apify Proxy on or off, or switching it between datacenter and residential, was measured to make no difference at all.

The practical answer is to schedule it outside those hours. Apify Schedules can run this nightly; roughly 00:00-06:00 UTC has been reliable. For ad-hoc research during the day, expect error rows, which, as below, cost you nothing.

What you get

Data type	What it returns
Keyword comparison	Up to 5 keywords on one shared 0-100 scale, with each one's average, peak, and rank
Interest over time	0-100 search-interest score per week/day/hour for your timeframe
Interest by region	Score per country, state, or US metro area
Related queries	Top related searches + rising searches (with growth %, including "Breakout")
Trending Now	Real-time trending searches for any country, with approximate traffic and linked news

What this Actor deliberately does not return

Two Google Trends data types are widely advertised by scrapers and do not actually work, because Google itself stopped returning them:

Not included	Why
Related topics	Google's endpoint returns an empty ranked list on both of its backends. Every result would be blank.
City-level interest	Google returns 200 cities with hasData: false on every one. Region (51) and US metro (210) do return real data, and both are supported.

Shipping an option that silently returns nothing is the fastest way to waste your money, so these were removed rather than left in the input schema to look impressive. Everything the schema offers returns data.

Use cases

• SEO & content research: find rising queries before they peak; catch "Breakout" terms early
• E-commerce & dropshipping: validate product demand by country before you spend on ads
• Market research: compare seasonality and regional demand across markets
• Trading & finance signals: track search interest for tickers, brands, and events
• AI agents & LLM pipelines: clean JSON via API, ready for automated workflows
• News & social monitoring: Trending Now with attached news stories, refreshed in real time

Compare keywords head-to-head

Set compareKeywords: true and every keyword is scored against the same ceiling, so the numbers are directly comparable, the question Google Trends is most often asked:

{ "keywords": ["nike", "adidas", "new balance", "hoka"], "compareKeywords": true, "geo": "US" }

#1  nike           avg 53.6   peak 100
#2  adidas         avg 23.0   peak  52
#3  new balance    avg 16.3   peak  34
#4  hoka           avg  6.6   peak  14

Each row carries the full timeline plus averageInterest, peakInterest, and rank, so you can chart it or sort it without post-processing.

Input example

{
  "keywords": ["matcha", "protein powder", "creatine gummies"],
  "dataTypes": ["interestOverTime", "relatedQueries", "interestByRegion"],
  "geo": "US",
  "timeframe": "today 12-m",
  "trendingNowGeo": "US"
}

Output example (one row per keyword × data type)

{
  "keyword": "matcha",
  "dataType": "relatedQueries",
  "geo": "US",
  "timeframe": "today 12-m",
  "top": [{ "query": "matcha latte", "value": 100 }],
  "rising": [{ "query": "matcha shortage", "formattedValue": "Breakout" }]
}

Export as JSON, CSV, Excel, or feed it straight into Google Sheets, Make, Zapier, LangChain, or your own code via the Apify API.

Pricing

Pay per result: you are charged a small fee per successfully returned data row (keyword × data type). Failed lookups cost nothing. A run with 10 keywords × 2 data types = 20 results.

There is no official Google Trends API

Google has never shipped one. There is no key to request, no quota page, and no endpoint in the Cloud console. What exists is the set of internal endpoints the trends.google.com page calls to render itself, and every tool in this space including this one is a typed wrapper over those.

That is worth knowing before you pick anything here, because it sets what is actually true of all of them:

• No API key exists to obtain. Any product offering you a "Google Trends API key" is issuing its own key for its own proxy.
• Values are relative, 0 to 100, never absolute search volume. Google does not publish volume through this surface at all.
• The endpoints rate-limit hard by IP, which is the single thing that decides whether a run works. See the proxy note below; it is not optional for keyword analysis.

Timeframes: now 1-H, now 4-H, now 1-d, now 7-d, today 1-m, today 3-m, today 12-m, today 5-y, and all for 2004 to present.

Data types: interest over time, interest by region (country, region or DMA resolution), related queries, rising queries, and Trending Now, which is the real-time surface and is the one path that needs no proxy of your own.

Any country, via geo, plus category and property filters (web, images, news, YouTube, Shopping).

Trending Now across markets

Pass several countries and each one runs in the same job:

{ "trendingNowGeo": "US, GB, DE, IL" }

Every row carries its own geo and its rank within that country, so a row ranked 1 in Israel and a row ranked 1 in Germany are both firsts rather than being renumbered into one list. That is the comparison people actually want, and it used to take one run per market.

approxTrafficMin is the numeric form of Google's approxTraffic, which it publishes as 500+, 20K+ or 2M+. Those sort alphabetically and cannot be filtered, so the number is there for "anything above 50,000".

Each trend also carries the news stories Google attached to it, with title, source, URL, snippet and image, which is usually enough to see why something is trending without following the link.

This path needs no proxy of your own. It reads Google's trending RSS feed, which is not rate-limited on Apify, unlike the keyword endpoints below.

Notes & fair use

• Google Trends values are relative (0-100 within your query), not absolute search volumes.
• By default each keyword is normalized independently. Set compareKeywords: true to put up to 5 on one shared 0-100 scale (see above).
• Regions with no search data are omitted rather than returned as zeros, so every row you get is a real signal.
• This Actor collects publicly available, aggregated, anonymous data only, no personal data is touched.

FAQ

Why do I sometimes see a keyword with an error field? Google aggressively rate-limits. The Actor retries with rotating proxies and exponential backoff; in the rare case a keyword still fails, it's reported transparently and not charged.

Does it support any country? Yes. Worldwide by default, or any ISO country/region code (US, GB, DE, IL, US-CA).

Can I schedule it? Yes. Use Apify Schedules to run it hourly or daily and deliver fresh trend data to your stack automatically.