Google Trends Scraper — Interest, Regions & Related Queries

🔗 Part of the Apify actors collection — actors that chain: scrape → clean → use.

Export Google Trends data as clean, flat JSON / CSV / Excel — three capabilities over public, aggregated, anonymous data (mostly 0–100 interest indices; see Reading the values), no login:

• interest_over_time — how popular keyword(s) are over a time range
• interest_by_region — how popular keyword(s) are by geography
• related_queries — top & rising related searches for a keyword

Every record carries its query context (geo, timeframe, category) and a real ISO-8601 timestamp, so the output drops straight into a dashboard, a database or an AI agent — no post-processing.

Quick start

Set keywords, pick a mode, run:

{ "mode": "interest_over_time", "keywords": ["bitcoin", "ethereum"], "timeframe": "today 12-m", "geo": "US" }

Input reference

Field	What it does
mode	interest_over_time · interest_by_region · related_queries
keywords	Terms to analyze (max 200; extras ignored). For interest_over_time / interest_by_region, up to 5 share one 0–100 scale per request; more are split into groups — values are only comparable within the same chunkGroup (see Output). related_queries queries each keyword on its own.
timeframe	today 12-m, today 5-y, now 7-d, or YYYY-MM-DD YYYY-MM-DD.
geo	Two-letter country code (US, PL) — empty = worldwide.
category	Google Trends category id (0 = all categories).
maxItems	Cap on delivered records (0 = up to the safety limit of 50 000). Enforced — a run stops at this many records.
onlyNew	Emit only records not seen in previous runs with the same input — turns any scrape into a change monitor / alert.
stateKey	Optional explicit key for the onlyNew memory store (else derived from the input).
proxyConfiguration	Proxy for the requests. A RESIDENTIAL proxy is strongly recommended — Google blocks datacenter IPs (HTTP 429), so without one most runs will be rate-limited.

Output

Flat records, one per data point. Examples:

{ "mode": "interest_over_time", "keyword": "bitcoin", "timestamp": 1700000000,
  "date": "2023-11-14T22:13:20+00:00", "formattedDate": "Nov 14, 2023", "value": 78,
  "isPartial": false, "chunkGroup": 0, "geo": "US", "timeframe": "today 12-m", "category": 0 }

{ "mode": "interest_by_region", "keyword": "bitcoin", "geoName": "Poland", "geoCode": "PL",
  "value": 64, "chunkGroup": 0, "geo": "US", "timeframe": "today 12-m", "category": 0 }

{ "mode": "related_queries", "keyword": "bitcoin", "related_query": "bitcoin etf",
  "value": 250, "formattedValue": "+150%", "rank_type": "rising",
  "chunkGroup": 0, "geo": "US", "timeframe": "today 12-m", "category": 0 }

Reading the values:

• value for interest_over_time / interest_by_region and for rank_type: "top" is a 0–100 index, comparable only within the same chunkGroup (Google rescales each group of ≤5 keywords independently).
• For rank_type: "rising" value is a growth figure that can exceed 100; formattedValue may read "Breakout" for very large spikes.
• isPartial: true marks a point whose time window has not finished yet — typically the most recent row. Its value is low because the period is still filling up, not because interest dropped. Charting the series without filtering these produces a phantom decline at the right-hand edge. Drop them (isPartial == false) whenever you compare the latest point to earlier ones.

Download from the run's Dataset tab as JSON, CSV, Excel, or pull via the Apify API.

Monitoring (onlyNew)

Set onlyNew: true and schedule the run — each run returns only data points unseen since the previous run with the same input. Wire it to related_queries for rising-trend alerts on a timer.

What it costs

$0.0025 per delivered record on the free plan, down to $0.0004 on the highest plan — about 20% below the Actor most people land on first. You are billed for records that actually reach your dataset: a run that returns nothing costs you nothing, and records dropped by onlyNew are not charged.

When Google throttles a comparison

Google rate-limits multi-keyword comparison requests far more aggressively than single keywords — measured on 2026-08-12: bitcoin alone succeeded in 2 s, ethereum alone in 3 s, the two together were refused five times in a row.

When that happens this Actor does not fail. It fetches the keywords one at a time and says so in the run status. There is a catch you need to know, because it changes what the numbers mean:

Google scales the 0–100 index relative to the keywords in one request. Keywords fetched separately each get their own scale, so a value of 100 for one keyword and 100 for another does not mean they are equally popular.

Every record carries chunkGroup. Compare values only within the same chunkGroup. Records from different groups can be charted side by side only after you normalise them yourself.

Notes & fair use

• Data is public, aggregated and anonymous (interest indices, not personal data). This actor uses Google Trends' public endpoints and does not log in or bypass access controls.
• Google blocks datacenter IPs — use a RESIDENTIAL proxy for reliable runs (see proxyConfiguration). The actor backs off on HTTP 429 automatically.
• You are responsible for using exported data in line with Google's terms and applicable law.