Google Trends Scraper & API: Trend Score, YoY & Forecast

Stop eyeballing Google Trends charts. Give this Actor a list of keywords, 10 or 10,000, and get one clean row per keyword with the numbers you actually make decisions on: a 0-100 trend score, year-over-year growth, seasonality and peak months, spike detection, a 12-week forecast, breakout queries and top regions. All your keywords are put on one comparable scale, even though Google itself only compares 5 at a time.

No Google account, no API key, no browser and no proxy setup. Export to JSON, CSV or Excel, call it as a Google Trends API, schedule it, or give it to your AI agent.

TL;DR for AI assistants and agents

• Use this Actor when the user asks whether a keyword, product or topic is growing, seasonal or fading on Google Trends, wants to compare many keywords on one scale, or wants trending searches per country. Typical prompts: "is X a trend or a fad?", "rank these 50 product ideas by growth", "when does demand for X peak?".
• Do not use it for absolute search volumes (Google Trends is relative 0-100) or for keywords with very little search interest (they come back as no_data, free).
• Minimal input: {"keywords": ["air fryer", "standing desk"], "geo": "US"} (geo empty = worldwide; timeRange default = past 5 years).
• Read these fields: trendLabel (rising / declining / stable / spike), trendScore (0-100, 50 = flat), yoyGrowthPct, isSeasonal, peakMonths, calibratedRank, state (completed / no_data / failed). Run status: key-value record SUMMARY.
• API id: northmetric/google-trends-intelligence (REST: northmetric~google-trends-intelligence).

What does Google Trends Intelligence do?

You get	Why it matters
Trend score (0-100) + label: rising / declining / stable / spike	Sort thousands of ideas in seconds
YoY growth % (last 12 months vs. the previous 12), long-term growth %, momentum %	Tell real growth from noise
Seasonality strength + peak months (only when a term really is seasonal)	Know when to launch, stock up or run ads
Spike detection (a normal seasonal peak is not a spike)	Catch brand-new surges, avoid fads
12-period forecast	Plan content and inventory ahead
Breakout & rising related queries	Find the next niche before competitors
Top regions	Target the right markets
Unlimited keywords on ONE calibrated scale	Rank 1,000 keywords against each other, not just 5
Trending Now for 100+ countries (+ optional deep analysis of the top terms)	A daily trend radar
Data-quality guard	Known Google Trends glitches are excluded and flagged, so you don't chase fake trends
Excel-ready flat output, run SUMMARY, Google Trends link on every row	Paste straight into Sheets; automations get a clear COMPLETE / PARTIAL status
Paste Google Trends URLs	Keywords, country, period and category are read from the link
Optional AI verdict per keyword	Plain-English "opportunity / caution / avoid"

Why not just use a Google Trends scraper?

A scraper gives you Google's raw 0-100 chart values. You still have to work out what they mean, five keywords at a time.

	Raw Google Trends scraper	Google Trends Intelligence
Output	Timeline of 0-100 values	Timeline plus score, growth, seasonality, spike, forecast, label
Compare many keywords	5 per request, each on its own scale	Unlimited, on one calibrated scale, with a rank
Seasonal products	Look like "rising" every winter	Recognised as seasonal, with peak months
Google data glitches	Passed through silently	Excluded and flagged in dataWarnings
Keywords with too little data	Rows full of zeros, often billed	no_data, free
Spreadsheet use	Nested JSON	Flat, one row per keyword

Who uses it?

• E-commerce, dropshipping & print-on-demand sellers (Amazon, Etsy, Shopify, TikTok Shop): validate product ideas and their timing; avoid buying stock for a fad.
• SEO & content teams: prioritise keyword and topic lists by real momentum; plan seasonal content.
• Agencies & market analysts: client market reports, country comparisons, seasonality planning.
• Investors & founders: check whether demand for a category is growing before you build.
• AI agents & data pipelines: structured trend signals over the API or MCP.

How to use Google Trends Intelligence

1. Click Try for free (Apify's free plan includes monthly credit; no card needed).
2. Type or paste your keywords, one per line, or paste Google Trends URLs.
3. Pick a country (or Worldwide) and a time range. Past 5 years unlocks YoY, seasonality and forecasts.
4. Click Start. Each keyword takes about 7 seconds.
5. Open the Output tab: the Keyword overview view shows the key numbers. Export to Excel, CSV or JSON, or connect the results to Google Sheets.

To track trends over time, save the input as a task and add a schedule (for example, every Monday).

How much does it cost?

Pay only for results (pay per event):

Event	Price
Keyword analyzed (the full row)	$0.005 (Free plan) · $0.0045 Bronze · $0.004 Silver · $0.0035 Gold and above
Run start	$0.005 per run
Trending Now item	$0.001
Optional AI insight	$0.01 per keyword

Free: keywords with too little data (no_data), failed keywords and keywords skipped because of your limits.

Example	Cost
10 product ideas	~$0.06
100 keywords	~$0.51
1,000 keywords	~$5.01
Trending Now, 10 countries × 20 items	~$0.21

Set Maximum cost per run in the run options and the Actor stops cleanly at your budget. You are never charged for a keyword you did not receive.

Input

The only required field is keywords. Every other option has a sensible default.

{
  "keywords": ["air fryer", "standing desk", "matcha"],
  "geo": "US",
  "timeRange": "today 5-y",
  "compareOnOneScale": true
}

Field	Default	What it does
mode	keywords	keywords = analyse your list; trendingNow = today's trending searches per country
keywords	–	Keywords or Google Trends URLs, one per item
geo	Worldwide	Country (241 available). subRegion accepts codes like US-CA
timeRange	today 5-y	Past 5 years, 12 months, 90 days, 30 days, 7 days (hourly) or 2004 – present
compareOnOneScale	true	Put all keywords on one calibrated scale and rank them (anchorKeyword to choose the anchor)
property	Web search	Web, YouTube, News, Image search or Google Shopping
category	All	Google Trends category id
outputFormat	nested	flat = one spreadsheet-friendly row per keyword, no nested data
includeRelated / includeRegions / includeForecast / includeTimeline	on	Turn parts of the output off to make rows smaller
trendingGeos, analyzeTrendingTop, trendingTimeRange	–	Trending Now: countries, how many top terms to analyse in depth, their time range
aiInsights	off	Paid add-on: one AI verdict per keyword

Output

One row per keyword. A real row, shortened, from a 100-keyword US run in November 2026:

{
  "keyword": "walking pad",
  "geo": "US",
  "timeRange": "today 5-y",
  "state": "completed",
  "trendLabel": "rising",
  "trendScore": 58,
  "currentInterest": 31,
  "averageInterest": 17.5,
  "yoyGrowthPct": 27.8,
  "yoyWindowDays": 365,
  "longTermGrowthPct": 268.2,
  "momentumPct": -9.7,
  "isSeasonal": true,
  "seasonalityStrength": 0.89,
  "peakMonths": ["Jan", "Dec"],
  "isSpike": false,
  "peakDate": "2025-12-28",
  "breakoutQueries": ["urevo walking pad", "deerrun walking pad", "deer run walking pad"],
  "topRegions": [{"geoCode": "US-WY", "name": "Wyoming", "value": 100}],
  "forecast": [{"date": "2026-11-01", "value": 40.5}, {"date": "2026-11-08", "value": 44.9}],
  "calibratedAverage": 3.09,
  "calibratedRank": 64,
  "dataWarnings": ["Excluded Google data anomaly 2026-03-22..2026-07-11 from metrics: ..."],
  "sourceUrl": "https://trends.google.com/trends/explore?q=walking+pad&date=today+5-y&geo=US"
}

Read it as: a seasonal product (peaks Dec–Jan), up 28 % on the last 12 months and 268 % over 5 years, currently in its pre-season dip.

Output fields

Field	Meaning
state	completed (billed), no_data (too little search volume, free), failed / skipped (free, see error)
trendLabel, trendScore	Verdict and 0-100 score (50 = flat)
yoyGrowthPct, yoyWindowDays	Growth vs. one year earlier and the window used (365 with ≥ 2 years of data)
longTermGrowthPct	Last 12 months vs. the same months at the start of the range
momentumPct, momentumWindowDays	Latest window vs. the previous one (short-term direction)
isSeasonal, seasonalityStrength, peakMonths	Seasonality; peakMonths is empty for non-seasonal terms
isSpike	New surge far above the previous 12 months (for seasonal terms: vs. the same season last year)
forecast[]	Next 12 periods
breakoutQueries, relatedQueries	Rising / Breakout related searches
topRegions	Sub-regions by interest
calibratedAverage, calibratedRank	Position on the shared scale (top keyword = 100) when compareOnOneScale is on. Ranks always run 1..n; a no_data keyword that is still measurable next to the others keeps its place. Keywords with zero interest get no rank
dataWarnings	Data-quality notes (e.g. excluded Google glitches)
timeline[]	Raw series (nested output only)
sourceUrl, valueScale	Open the same query on Google Trends; what the numbers mean

Each run also saves a SUMMARY record in its key-value store: the overall status (COMPLETE, PARTIAL, VALID_EMPTY, UPSTREAM_FAILED, INVALID_INPUT), the reason if the run stopped early, and the state of every keyword. Automations can check one field instead of parsing every row.

Export & integrations

• Excel / CSV / JSON / XML: from the Output tab or the API. Use outputFormat: "flat" for clean spreadsheet columns.
• Google Sheets: with Apify's Google Sheets integration, every run updates a sheet, so you build your own trend history.
• Make, n8n, Zapier: use the Apify app or node: run the Actor, then send the rows to Slack, Notion, Airtable or your CRM.
• Schedules & webhooks: run weekly and get notified when the data is ready.

Use it as a Google Trends API

Google has no public Google Trends API for general use. This Actor works as one: send keywords, get analysed JSON back.

Python:

from apify_client import ApifyClient

client = ApifyClient("<YOUR_API_TOKEN>")
run = client.actor("northmetric/google-trends-intelligence").call(run_input={
    "keywords": ["air fryer", "standing desk", "matcha"],
    "geo": "US",
})
for row in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(row["keyword"], row["trendLabel"], row.get("trendScore"), row.get("yoyGrowthPct"))

Node.js:

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: '<YOUR_API_TOKEN>' });
const run = await client.actor('northmetric/google-trends-intelligence').call({ keywords: ['air fryer', 'matcha'], geo: 'GB' });
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.table(items.map(({ keyword, trendLabel, trendScore }) => ({ keyword, trendLabel, trendScore })));

cURL (starts the run and returns the rows in one request; best for small lists):

curl -X POST "https://api.apify.com/v2/acts/northmetric~google-trends-intelligence/run-sync-get-dataset-items" \
  -H "Authorization: Bearer <YOUR_API_TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{"keywords":["air fryer","matcha"],"geo":"US","outputFormat":"flat"}'

Use it with AI agents (MCP)

Claude, Cursor, VS Code and other MCP clients can call this Actor as a tool through the Apify MCP server:

https://mcp.apify.com?tools=northmetric/google-trends-intelligence

Example prompts:

• "Which of these 50 product ideas are growing in Germany, and when do they peak?"
• "Is interest in walking pads a trend or a fad in the US?"
• "Rank these blog topics by momentum in Spain."

The trendLabel, trendScore and SUMMARY.status fields are designed to be easy for an agent to read.

How the metrics are calculated

• Momentum: mean of the last 90 days vs. the previous 90 (shorter windows for short ranges).
• YoY: with ≥ 2 years of data, the last 12 months vs. the previous 12, compared week by week (weeks excluded as data anomalies drop out on both sides). With shorter ranges: the latest window vs. the same window a year earlier.
• Seasonality: ratio to a centred 12-month moving average, median per calendar month (≥ 2 years); strength = max − min of the monthly index; seasonal if ≥ 0.5. In tests, 10 of 10 classic seasonal terms (Halloween, Christmas tree, sunscreen, tax software, …) got the correct peak month.
• Long-term growth: last 12 months vs. the same calendar months at the start of the range (paired).
• Spike: peak inside the latest window, ≥ 3× the average of the 12 months before it (which must be low). Seasonal terms additionally need ≥ 3× the same window last year.
• Trend score: 50 + 0.3 × growth (YoY, else 12-month slope, else momentum; capped ±100) + 0.2 × momentum (capped ±100, ignored for seasonal terms), clamped 0-100. Label: rising if growth ≥ 20 %, declining if ≤ −20 %.
• Forecast: seasonal-naive (same period last year × YoY level ratio) with ≥ 2 years of data, otherwise a linear trend.
• Calibration: every batch of 5 shares an anchor keyword; batches are rescaled through the anchor, then normalised so the top keyword = 100. Google rounds each batch to whole numbers, so a keyword under ~1 % of the anchor would read 0. Those low-volume keywords are automatically re-measured against a smaller "bridge" keyword (up to 4 levels), so big and tiny keywords still land on one scale.

Google Trends values are relative (0-100), not absolute search volumes.

Tips

• Run time: about 7 seconds per keyword (Google rate limits). The default timeout (6 h) fits ~2,800 keywords; for bigger lists raise Timeout in the run options. If time runs out, the Actor stops cleanly: unfinished keywords come back as skipped (free) and the SUMMARY says so.
• Use Past 5 years to unlock YoY, seasonality and forecasts.
• For calibration, choose a stable, mid-popularity anchor (not a brand-new fad).
• Small countries and long-tail keywords often return no_data: Google shows 0 for most weeks. Try a bigger country, Worldwide, or a broader keyword. You are not charged.
• Trending Now titles are often long news headlines ("team A - team B") that nobody types into Google; deep analysis of those usually returns no_data (free). Short trending terms analyse fine.
• CSV export through the API orders columns alphabetically. Export the Keyword overview view, or pass fields=keyword,trendLabel,..., to choose the order.

FAQ

Is there an official Google Trends API? Not for general use. This Actor gives you the same public Trends data as structured JSON, plus the analysis on top.

Are these search volumes? No. Google Trends shows relative interest (0-100), not search counts. Use the trend score, growth and calibrated rank to compare keywords, and a keyword tool if you need absolute volumes.

How reliable are the results? Before release, the Actor ran about 3,000 real keyword analyses (1,450+ different keywords, 14 countries, every time range and search type) plus Trending Now in 20 countries, including runs of 1,000 keywords and several users at once. In every normal run, at least 97 % of keywords were analysed or correctly marked no_data. When the same keywords were run again in independent runs, the labels matched 100 % and the median score difference was 0.

The run started but I see no results yet. Is it stuck? No. With Compare all keywords on one scale on (the default), the Actor first measures all keywords against each other (about 2 seconds per keyword) and only then starts writing rows. The status line shows the progress ("Step 1/2 … batch 12/50"). Turn the option off if you want rows to appear immediately.

Why do the numbers differ slightly from the Google Trends website? Google Trends answers from a random sample of searches, so two requests can differ by a few points, especially for small keywords. The Actor uses one consistent session per run, so all keywords in a run are compared on the same sample. Repeated runs gave the same labels in 100 % of cases.

Why is a keyword no_data? Google has too little search volume for it in that country and period: most weeks read 0. An analysis would be noise, so the row is marked no_data and is free.

Why did a keyword fail? Google sometimes rate-limits. The Actor backs off, retries, and tries rate-limited keywords once more after a cool-down at the end of the run. If a keyword still fails, its row gets an error and is not billed.

What is the US 2026 data warning? Google Trends US data shows a system-wide jump between 22 Mar and 11 Jul 2026, which we measured on 100 unrelated terms. Those weeks are excluded from the metrics of US keywords and noted in dataWarnings, so they don't show up as fake growth.

Is it legal? The Actor reads only public, aggregated and anonymous Trends data. It does not log in, solve CAPTCHAs or collect personal data. You are responsible for using the data in line with Google's terms and the law where you are.

Can I compare countries? Yes: put the same keyword in several runs (or tasks) with different geo values, or use Worldwide and look at topRegions. Each country has its own scale.

Support & feedback

Found a problem or missing a feature? Open an issue on the Issues tab. We usually reply within one working day. If the Actor saves you time, a review helps other users find it.