Google Trends Scraper & API: Compare 100+ Keywords, Regions, Rising Queries & Trending Now

Get Google Trends data that actually arrives. This Actor collects interest over time, interest by country, state, metro (DMA) and city, top and rising related queries (with breakouts), and Google Trending Now searches with volume and news articles. It can put any number of keywords on one comparable 0–100 scale: Google allows 5 terms per comparison, and the Actor anchors the rest automatically. It runs without a browser and handles rate limits (HTTP 429) for you. There is no start fee, and failed terms are never charged. The output is flat, documented JSON and CSV that works in spreadsheets, BI tools and AI agents (MCP).

Why this Google Trends scraper

	This Actor	Typical Google Trends scrapers
More than 5 keywords on one scale	✅ Automatic anchoring and bridging. 100+ terms on one scale, with a scale-quality label on every term	❌ 5 per request, or "add your own anchor term"
Rate limits (HTTP 429)	✅ Adaptive pacing, backoff with jitter, fresh sessions and automatic proxy escalation (none → datacenter → residential). Proxy cost is included	Runs fail, hang or return empty data
Related queries	✅ Top and rising, growth %, breakout flag and an on-topic flag (Google mixes generic queries into rising lists)	Often missing
Regions	✅ Several levels in one run (states, metros and cities) and a low-volume toggle	Usually one level
Trending Now	✅ Hundreds of trends per country (4 h / 24 h / 48 h / 7 days), categories, active-only filter, volume, growth, start/end time, related searches and up to 3 news articles for every trend	RSS feed only (about 10 items)
Monitoring	✅ New rising queries, breakouts, direction changes, top-region changes, jumps and drops. "Only new/changed" output mode	Rare
Failed terms	✅ Error item with a reason and a retry hint, never charged	Silently missing, or charged anyway
Price	✅ $1.50 per 1,000 term reports (down to $0.75 on Business plans), no start fee, platform and proxy usage included	$2–$25 per 1,000, often with a start fee or browser usage billed on top

What you can use it for

• SEO and content planning: rank 50–500 keyword ideas on one scale, find breakout queries before they get competitive, and plan around seasonal peaks.
• Market and product research: compare brands, products or categories over 5 years, and see which states, metros or cities search the most.
• E-commerce and ads: time campaigns to demand, and spot rising product queries such as "hoka women's transport 2" (+3,350%).
• Newsrooms and social media: pull a daily Trending Now feed per country, with volume, categories and the news stories behind each trend.
• Monitoring and alerts: schedule a weekly run and get only what changed (new breakouts, direction flips, jumps).
• AI agents and LLM pipelines: compact stats, a plain-English insight and stable status fields on every item.

Quick start

1. Enter search terms. These can be keywords, topic IDs such as /m/0d6lp, or Google Trends explore URLs. You can also point Google Sheet at a shared sheet, and the Actor reads the terms from its first column.
2. Choose locations (US, US-TX, GB…; leave it empty for worldwide), a time range, and optionally a category or search type (web, news, images, YouTube, shopping).
3. Optionally add Trending Now countries.
4. Click Start. Results appear in the Output tab, and CSVs are in the key-value store.

Input examples

{ "searchTerms": ["coffee", "tea", "matcha", "cold brew", "kombucha", "yerba mate", "oat milk"], "geos": ["US"], "timeRange": "today 12-m" }

Seven terms on one scale, with regions and related queries for each.

{ "searchTerms": ["solar panels"], "geos": ["US-TX"], "regionLevels": ["DMA", "CITY"], "maxRegions": 10 }

Metro- and city-level demand inside Texas.

{ "trendingNowGeos": ["US", "GB", "IN"], "trendingNowHours": "24", "trendingNowCategories": ["Sports"], "trendingNowSort": "searchVolume", "trendingNowMaxItems": 15 }

Daily Trending Now report, with news articles for each trend.

{ "searchTerms": ["ai agents", "mcp server"], "geos": ["US"], "outputMode": "newOrChanged" }

Breakout alerts. Schedule it, and each run returns (and charges) only what changed.

{ "searchTerms": ["https://trends.google.com/trends/explore?date=today%2012-m&geo=US&q=aeropress,chemex,moka%20pot"] }

Paste any Trends explore URL. Its location, time and category are used.

{ "spreadsheetId": "https://docs.google.com/spreadsheets/d/<your-sheet-id>/edit", "geos": ["US"] }

Keywords from a Google Sheet shared as "Anyone with the link can view" (first column, header optional).

Output

Term report (default dataset, one item per term × location)

Headline fields: avgInterest, latestInterest, peakInterest + peakDate, lowInterest, changePct (first vs last quarter of the range), recentChangePct, trendDirection (rising/falling/stable), topRegion, topRelatedQuery, topRisingQuery + growth, breakoutQueries, and a one-sentence insight. Detail arrays: timeline[], regions[], relatedQueriesTop[], relatedQueriesRising[]. Diagnostics: scale (quality and anchor chain), sections (status of each part), errors[], change (since the last run) and trendsUrl (the same query on Google Trends).

Real item from a 12-term run on the platform on 2026-10-20 (arrays shortened):

{
  "type": "termReport", "term": "hoka", "geo": "US", "geoName": "United States", "timeRange": "today 12-m",
  "status": "ok",
  "sections": { "interestOverTime": "ok", "interestByRegion": "ok", "relatedQueries": "ok", "relatedTopics": "withheld" },
  "avgInterest": 12.33, "latestInterest": 8.15, "peakInterest": 25.34, "peakDate": "2026-04-12",
  "changePct": -8.6, "recentChangePct": -12, "trendDirection": "stable", "granularity": "week",
  "scale": { "type": "batch", "batchSize": 12, "anchorTerm": "running shoes", "quality": "anchored",
             "chain": ["scaled via anchor \"running shoes\""], "rankInBatch": 6 },
  "topRegion": "Wyoming", "topRelatedQuery": "hoka shoes",
  "topRisingQuery": "hoka women's transport 2", "topRisingQueryGrowth": "+3,350%",
  "insight": "Interest in \"hoka\" (United States, past 12 months) is stable (-9% first vs last quarter); peak 25.34 on 2026-04-12, latest 8.15; #6 of 12 compared terms; most popular in Wyoming; top rising query \"hoka women's transport 2\" (+3,350%).",
  "regions": [{ "level": "REGION", "rank": 1, "geoCode": "US-WY", "geoName": "Wyoming", "value": 100 },
              { "level": "REGION", "rank": 2, "geoCode": "US-HI", "geoName": "Hawaii", "value": 33 }],
  "relatedQueriesTop": [{ "rank": 1, "query": "hoka shoes", "value": 100 }, { "rank": 2, "query": "hoka bondi", "value": 45 }],
  "relatedQueriesRising": [{ "rank": 1, "query": "hoka women's transport 2", "growthPct": 3350, "isBreakout": false, "formattedValue": "+3,350%", "containsTerm": true }],
  "timeline": [{ "date": "2025-10-19", "value": 9.5 }, { "date": "2026-10-18", "value": 8.15, "partial": true }],
  "change": { "status": "new" }
}

Values are on the shared batch scale: the most popular term in the batch ("sneakers") peaks at 100. Small terms keep 3 significant digits (for example 0.0312), so they are never flattened to 0.

Trending Now (dataset trending, one item per trending search)

{
  "type": "trendingSearch", "rank": 1, "googleRank": 1, "term": "is kim jong un still alive", "geo": "US", "hoursWindow": 24,
  "searchVolume": 200000, "searchVolumeLabel": "200K+", "growthPct": 1000,
  "startedAt": "2026-10-19T18:00:00.000Z", "endedAt": null, "isActive": true, "durationHours": 20.4,
  "categories": ["Politics"], "relatedQueries": ["kim jong un dead", "did kim jong un die", "kim jong un is he alive"],
  "newsArticles": [{ "title": "Kim Jong Un: Why do rumours about his death keep coming back?", "url": "https://www.bbc.com/news/articles/cmvg6qv1jdydo",
                     "source": "BBC", "publishedAt": "2026-10-20T11:45:49.000Z", "picture": "https://encrypted-tbn1.gstatic.com/images?q=tbn:…" }],
  "change": { "status": "new" }
}

Comparison (dataset comparison, one item per batch)

Every term in the batch is ranked, with avgInterest, shareOfInterestPct, vsLeaderPct, peak, direction and scale quality. Example insight: "sneakers" leads 12 compared terms in United States (past 12 months); #2 "sandals" has 82% of its interest; last "barefoot shoes" has 3.5%.

CSV files and run summary (key-value store)

• TIMELINE.csv (long format) and TIMELINE_WIDE.csv (one column per term, ready for charts)
• REGIONS.csv, RELATED_QUERIES.csv, TRENDING.csv and COMPARISON.csv
• OUTPUT: status of each term, totals, charged events, and connection stats (requests, retries, rate limits absorbed, proxy tier used)

How the shared scale works

Google compares at most 5 terms per request and scales each request to its own peak of 100. For more than 5 terms, the Actor puts the most popular term (or your anchorTerm) into every group of 4 others. It then rescales each group by the anchor's ratio over the overlapping periods. A term that is too small to measure precisely against the anchor (raw value under 10) is measured again next to a closer-sized term (a "bridge"). Finally, the scale is normalised so the top term peaks at 100. Each term reports scale.quality:

• exact: in the same request as the leader.
• anchored: scaled through the anchor.
• bridged: scaled through a closer-sized term.
• approximate: low precision.
• unscaled: no overlap with the anchor.

To get each term on its own 0–100 scale instead, set compareTerms: false.

Monitoring: what changed since the last run

Every term report and trending search carries change.status (new, changed or unchanged) and change.events. The events are new rising queries, new breakouts, a direction change, a top-region change, and interest jumps or drops beyond changeThreshold points. Runs with the same input share a monitor automatically; to pin one, set monitorKey. With Output set to "New + changed items", unchanged items are neither saved nor charged.

No false alarms from Google's sampling. Google samples its rising-queries list, so two requests a few minutes apart can return different subsets of 25. A naive "new query since last run" alert therefore fires on almost every term (we measured 12 of 12 terms flagged on an unchanged rerun). This Actor counts a rising query as new only if it has never been seen for that term, contains the term, and is a Breakout or ranks in the top 15. It must also still be there when the list is requested twice more in the same run; those extra requests happen only when there is a candidate, and you aren't charged for them. Rising-query alerts start on the third run; the first two runs build the baseline. Direction, region and interest changes are reported from the second run. The trade-off is deliberate: a query that Google's samples disagree on is treated as noise, so a genuinely new query can show up a run later than on the website. In exchange, alerts don't cry wolf.

Pricing

Pay per result. There is no start fee, and platform and proxy usage are included.

Event	Free plan	Starter	Scale	Business and above
Term report (one term × one location × one time range)	$1.50 / 1,000	$1.20 / 1,000	$1.00 / 1,000	$0.75 / 1,000
Trending search	$0.40 / 1,000	$0.30 / 1,000	$0.25 / 1,000	$0.25 / 1,000
Timeline row in the optional long-format timeline dataset (off by default)	$0.02 / 1,000	$0.02 / 1,000	$0.02 / 1,000	$0.02 / 1,000
Comparison items, CSV files (including the full timeline), run summary	free	free	free	free
Terms that fail or are below Google's volume floor	free	free	free	free

Examples on the Free plan: 100 keywords in one country cost $0.15. A weekly monitor of 20 keywords costs about $0.13 a month. A daily Trending Now feed of 25 trends in each of 3 countries costs about $0.90 a month. A term with partial data (for example, regions failed but the timeline and queries arrived) is charged as one report, and its sections and errors show what is missing. To cap spending, set Maximum cost per run in the run options. The Actor stops cleanly at the limit and never saves more than you pay for.

Use with AI agents (MCP)

The Actor runs with limited permissions and pay-per-event pricing without a start fee, so it works with the Apify MCP server and agentic payments. Every field has a title and description in the dataset schema. Each item has a plain-English insight and stable status/sections values, so an agent can tell what is missing and why. A compact agent call is {"searchTerms": ["topic a", "topic b"], "geos": ["US"], "includeTimeline": false}, which returns stats without the timeline arrays.

Limits and honest notes

• Related topics: Google currently withholds related topics from automated sessions. Every client we tested got empty topic lists, including a real browser, datacenter IPs and residential IPs. The Actor returns topics whenever Google serves them; otherwise it sets sections.relatedTopics: "withheld", at no extra cost. Related queries work normally.
• Off-topic rising queries: Google sometimes adds generic queries (such as "home workout routines") to rising lists for automated clients. Every rising row has containsTerm. The headline fields use on-topic rows only, and risingQueriesFilter: "containingTerm" removes the rest.
• Sampling: Google Trends values are sampled and rounded, so they can differ slightly between runs and from the website. Small states such as Wyoming often rank first for common terms. A real browser on the Trends website gets the same ranking (checked against a browser-based scraper on the same day).
• Speed: heavy runs are paced to stay under Google's rate limits, at roughly 2–6 seconds per term in large batches. A 12-term batch with regions and related queries took 23 seconds. A 23-term batch on a day of heavy rate limiting took 2 minutes, with 15 rate-limit responses absorbed.
• Bad days at Google: when Google throttles very hard, the run doesn't grind until the platform timeout. It stops starting new requests at a time budget (about 4 minutes plus 30 seconds per term, at most 30 minutes, always ahead of the run timeout) or once a small residential-proxy traffic budget is used, saves everything it already fetched, and marks the rest status: "error" with code: "time_budget" or "traffic_budget". Those items are not charged. Set Max run time (maxRunSeconds) to change the time budget.
• Trending Now news: Google lists up to 3 articles per trend. In our final test, all 75 trends (US, GB, DE) had 3 each.

FAQ

Why do other Google Trends scrapers fail or return empty results? Google rate-limits (HTTP 429) aggressively and often answers automated traffic with a captcha page. This Actor paces requests, retries with backoff and fresh sessions, and moves to Apify datacenter and then residential proxies when needed. The proxy cost is included in the price. If Google still blocks a term after all retries, that term comes back as an error item with a hint, and you are not charged for it.

Can I compare more than 5 keywords? Yes, any number (up to 2,000 per run). See "How the shared scale works".

Why are related topics empty? Google withholds them from automated sessions (see "Limits"). Related queries are not affected.

Why do my numbers differ slightly from the Google Trends website? Google samples its data, so two requests minutes apart can differ by a point or two. Batches of more than 5 terms are also rescaled to the batch leader. trendsUrl opens the same query on Google Trends.

Is there an official Google Trends API? Google's Trends API is in limited alpha (by application only). This Actor reads the same public endpoints that the Trends website uses.

Do I need a proxy? No. The Actor uses Apify Proxy automatically when Google rate-limits it. You can force a proxy group in the advanced settings.

Which time ranges work? Past hour, 4 hours, day, 7 days, 30 and 90 days, 12 months, 5 years, since 2004, and custom ranges such as 2024-01-01 2024-06-30.

Is scraping Google Trends legal? The Actor collects public, aggregated and anonymous statistics. Check the terms that apply to your use.

Found a bug or need a field? Open an issue on the Issues tab. Issues are answered there.