Google Trends Scraper: Interest Over Time, Regions & Related

Google Trends data for any keyword without a browser: interest over time, interest by region, related queries and related topics, plus the country's "Trending now" feed. The Actor calls the same JSON endpoints the Trends website calls (explore → widget tokens → widget data), so a keyword's full series arrives in about 2 seconds and a 70-keyword run finishes in under 90 seconds on a 256 MB datacenter run. No login, no API key, no Puppeteer. Pay per result.

Built for people who were burned by browser-based Trends scrapers that time out or fail half their runs: the Actor paces itself around Google's rate limit (a 429 is paused and retried, never a failure), rotates sessions, dedupes, and writes a SUMMARY record with every counter you need to trust the data.

What it does

For every keyword you pass it returns one record per selected dataset:

Dataset	One record per keyword with…
interestOverTime	the complete 0–100 series (points[] with ISO dates and isPartial), plus average, peak, latest, pointCount, resolution (MINUTE/HOUR/DAY/WEEK/MONTH — chosen by Google from the time range)
interestByRegion	every country/region/city/DMA with data, sorted by value, plus topRegion and regionCount
relatedQueries	the Top list (25 queries with index 0–100) and the Rising list (25 queries with +250%-style growth or Breakout), each with a Trends link
relatedTopics	the same shape for topics (title, type, Google mid) — Google currently returns this list empty for many terms, in which case no record is charged
trendingNow	the country's "Trending now" feed (no keywords needed): ~10–20 trending searches with approximate traffic, start time, picture source and the headline news items behind them

Values are Google's relative index (100 = the keyword's own peak in the chosen range and location), exactly as the Trends site shows them. Each keyword is queried on its own, so series are not rescaled against each other.

Input

Field	Type	Default	Notes
keywords	string[]	["chatgpt"]	Search terms or Google topic ids (/m/0k8z). Up to a few hundred per run; ~1 request per keyword plus one per dataset.
geo	string	"" (worldwide)	ISO country (US, GB, DE, IN), region (US-CA, GB-ENG) or US DMA (US-CA-807). Also picks the trendingNow country.
timeRange	string	today 12-m	now 1-H, now 4-H, now 1-d, now 7-d, today 1-m, today 3-m, today 12-m, today 5-y, all, or a custom YYYY-MM-DD YYYY-MM-DD. Aliases: 7d, 30d, 12m, 5y.
datasets	string[]	interest over time, by region, related queries	Any of interestOverTime, interestByRegion, relatedQueries, relatedTopics, trendingNow.
category	integer	0	Trends category id (0 = all; 5 Computers & Electronics, 71 Food & Drink, 12 Business, 45 Health, 20 Sports — the cat= number in the Trends URL).
maxItems	integer	100	Total records cap (1–10 000).
proxyConfiguration	object	Apify datacenter proxy	Datacenter proxies pass; Google only rate-limits, which the Actor paces around.

Agent-friendly: a single keyword with the default datasets is one record per dataset in ~2–3 seconds, for about $0.011.

Output

interestOverTime (abridged):

{
  "kind": "interestOverTime",
  "id": "interestOverTime:3f1c9e0b2a77",
  "url": "https://trends.google.com/trends/explore?q=ozempic&date=today+5-y",
  "keyword": "ozempic", "geo": "", "timeRange": "today 5-y", "category": 0, "property": "",
  "resolution": "WEEK", "hasData": true, "pointCount": 262,
  "average": 57.1,
  "peak":   { "date": "2024-05-26T00:00:00.000Z", "value": 100 },
  "latest": { "date": "2026-11-01T00:00:00.000Z", "value": 44, "isPartial": true },
  "firstDate": "2021-10-31T00:00:00.000Z", "lastDate": "2026-11-01T00:00:00.000Z",
  "points": [ { "date": "2021-10-31T00:00:00.000Z", "value": 3, "isPartial": false }, "…" ],
  "source": "trends.google.com/trends/api/widgetdata/multiline",
  "scrapedAt": "2026-11-06T17:48:40.000Z"
}

relatedQueries carries top[] and rising[] ({ query, value, formattedValue, isBreakout, url }), interestByRegion carries regions[] ({ geoCode, name, value }) and topRegion, trendingNow carries title, rank, approxTraffic ("20K+" → 20000), approxTrafficText, publishedAt, picture, pictureSource and news[] ({ title, url, source }, headline snippets only).

The key-value store's SUMMARY record lists every counter: keywords explored, records per dataset, emptyResults (datasets Google returned empty for — not charged), noDataKeywords, rateLimited, blocked, requestsFailed, time to first record, p95 latency, peak concurrency and compute seconds. Dataset views: Overview, Interest over time, Trending now.

Pricing

Pay per event: $0.005 per run + $0.002 per record (Store discounts: −10 % Starter, −20 % Scale, −30 % Business). A record is a whole series or table, not a data point. Examples: one keyword with the three default datasets ≈ $0.011; 100 keywords × 3 datasets = $0.605; the trending feed for one country ≈ $0.03. A 256 MB run of 70 keywords took 85 s of compute (≈ $0.007 at the Starter tier). Compare: the official browser-based Actor charges $0.006 per record and fails about half of its public runs; the current PPE leader charges $0.025 per start + $0.002 per record.

Monitoring (scheduled runs)

Schedule the Actor with your keyword list and timeRange: "now 7-d" (hourly points) or today 3-m (daily points) and feed the dataset into mdcurt/dataset-diff-monitor to get alerted when latest.value jumps, a new query appears in rising[] with isBreakout: true, or a term enters the trendingNow feed. Record ids are stable per keyword × geo × range × category, so diffs line up run to run.

Speed (live tests, 2026-11-06, datacenter egress)

Run	Result
1 keyword, interestOverTime only	1 record, first record 2.2 s, 2 requests
3 keywords worldwide, 5 years, all four keyword datasets	9 records (related topics empty → not charged), first record 2.6 s, 15 requests
trendingNow for GB	10 records in 0.5 s, 1 request
2 keywords, US-CA, custom range 2025-01-01 2025-06-30	4 records (daily resolution), first record 2.5 s
70 keywords × 3 datasets	210 records, 280/280 requests OK, 0 rate limits, 0 blocks, first record 2.2 s, 85 s compute (148 records/min)
200 keywords × 3 datasets (sustained)	see the Limitations: Google throttles a single IP after roughly 150–300 calls; the Actor slows to ~1 request/s and finishes with zero failed requests

Tutorials

• How to track competitors on Google Trends automatically (no browser, $0.035 a run) — also in examples/competitor-search-interest-monitor/ in the Actor source: a real 15-record sample dataset (5 running-shoe brands, 2026-11-08), how to read it, and the dataset-diff-monitor recipe that turns it into a weekly Slack digest. The ready-made input is the example task Monitor competitor brands' Google search interest by US state on this Actor's Example tasks tab.

For AI agents

Pass one thing: {"keywords": ["<term>"]} (optionally "geo": "US"). Back come three records in about 2–3 seconds — interestOverTime (average, peak.date, latest.value, the full points[]), interestByRegion (topRegion, regions[]) and relatedQueries (top[], rising[] with isBreakout). One lookup costs about $0.011 ($0.005 start + 3 × $0.002). The Actor runs with limited permissions and no login, so it is usable through Apify's MCP server (https://mcp.apify.com/?tools=mdcurt/google-trends-scraper) and through agentic payments without an Apify account. Pass maxItems: 1 with datasets: ["interestOverTime"] for the cheapest single-number answer.

Limitations

• Google Trends is sampled, relative data; Google may answer "not enough data" (an all-zero series, hasData: false) for rare terms.
• relatedTopics comes back empty for many terms since Google changed the Trends backend; empty datasets are not charged.
• Keywords are explored one at a time (no 5-term comparison rescaling yet) and only web search (property youtube/news/images/shopping is on the roadmap).
• One IP sustains roughly 1–3 requests per second before Google answers HTTP 429. The Actor pauses, lowers concurrency, retries on another session and never counts a 429 as a failure, but very large runs (500+ keywords) take several minutes. Residential proxies are not needed.
• The Actor never requests the robots-disallowed /trends/explore HTML page; it reads the JSON API behind it. Trends data is Google's; use it within Google's terms.

Changelog

• 1.0.0 (2026-11-06) — first release: interest over time, interest by region, related queries/topics, trending-now feed; cookie bootstrap per session, 429 pacing, six-tier Store discounts.