Google Trends API: Interest, Regions, Related & Trending avatar

Google Trends API: Interest, Regions, Related & Trending

Pricing

from $3.50 / 1,000 query results

Go to Apify Store
Google Trends API: Interest, Regions, Related & Trending

Google Trends API: Interest, Regions, Related & Trending

Reliable Google Trends API for AI agents and developers: interest over time, interest by region, related queries and trending searches. Compact JSON with a plain-English summary per keyword. pytrends alternative, MCP-ready.

Pricing

from $3.50 / 1,000 query results

Rating

0.0

(0)

Developer

Datafaro

Datafaro

Maintained by Community

Actor stats

0

Bookmarked

2

Total users

1

Monthly active users

an hour ago

Last modified

Categories

Features

Limited permissions

Share

A reliable Google Trends API for AI agents and developers. Ask for interest over time, interest by country or region, related queries, or what's trending right now. You get compact JSON with a plain-English summary for every keyword, so an agent can answer straight from the summary without reading the data points.

  • Built for AI agents. Six input fields, all with defaults, and errors that say how to fix the input. Works as an MCP tool for Claude, Cursor and any MCP client.
  • Small output. One item per keyword, not one per data point, plus a single OUTPUT JSON for fast run-sync calls.
  • Reliable. Residential proxy with session rotation, exponential backoff with jitter, and per-keyword failure isolation. A broken keyword never sinks the run, and you don't pay for it.
  • pytrends alternative. pytrends was archived in April 2025. This Actor talks to Google Trends directly and needs no Python setup on your side.

What it does

modeWhat you getKeywords
compare (default)Interest over time, all keywords on one shared 0–100 scale, with average, peak, trend and % change1–5
by_regionInterest by region within geo (e.g. US states), or by country when geo is "", each keyword on its own 0–100 scale1–5
relatedRelated search queries, both top and risingexactly 1
trending_nowCurrent trending searches in a country (Google has no worldwide list; empty geo shows the US), with approximate traffic and news linksignored

Input

FieldTypeDefaultDescription
modeenumcomparecompare, by_region, related, trending_now
keywordsstring[]["chatgpt", "claude"]1–5 search terms (exactly 1 for related, ignored by trending_now)
geostring"" (worldwide)Worldwide: "", "WORLD" or "worldwide". A country: ISO code like US. A region: US-CA. trending_now needs a country; with an empty or worldwide geo it shows the US and adds a note.
timeframeenumpast_12_monthspast_day, past_7_days, past_30_days, past_90_days, past_12_months, past_5_years
categoryinteger0Google Trends category ID (0 = all)
propertyenumwebweb, youtube, news, images, shopping

An empty input {} runs the defaults: chatgpt vs claude worldwide over the past 12 months. Example:

{ "mode": "compare", "keywords": ["claude", "chatgpt"], "geo": "ES", "timeframe": "past_12_months" }

Invalid input fails the run immediately with a message you can act on, e.g. Invalid input. mode "related" takes exactly 1 keyword. Nothing is charged.

Output

One dataset item per keyword. For trending_now it's one item for the country. Every item has the same envelope:

FieldMeaning
keyword, mode, geo, timeframe, property, categoryWhat was asked
statusok, no_data (Google has too little search volume) or error (this item failed)
errorWhy it failed, or null
summaryOne sentence answering the question, "No data", or "Failed: …"

compare example:

{
"keyword": "claude", "mode": "compare", "geo": "ES", "timeframe": "past_12_months",
"property": "web", "category": 0, "status": "ok", "error": null,
"summary": "Search interest in \"claude\" in ES over the past 12 months is rising, up 64% from the first to the last quarter; it peaked at 100 on 2026-03-08 and averaged 37 on a 0-100 relative scale, ranking 2 of 2 compared keywords by average.",
"points": [{ "date": "2025-11-16", "value": 3 }, "…", { "date": "2026-11-15", "value": 32, "partial": true }],
"average": 37,
"peak": { "date": "2026-03-08", "value": 100 },
"trend": "rising",
"change_pct_last_vs_first_quarter": 64
}

trending_now example ({"mode": "trending_now", "geo": "US"}), shortened to one trend:

{
"keyword": null, "mode": "trending_now", "geo": "US", "timeframe": "past_12_months",
"property": "web", "category": 0, "status": "ok", "error": null,
"summary": "Top trending searches in US right now: hacker (500+), milo ventimiglia (500+), sonny gray (2000+), albemarle county public schools (500+), anheuser busch (200+).",
"trends": [
{
"query": "hacker",
"approx_traffic": "500+",
"started_at": "2026-11-18T20:40:00Z",
"top_news_title": "British hacker who stole $11m through text message scheme jailed for six years in US",
"news": [
{ "title": "British hacker who stole $11m through text message scheme jailed for six years in US", "url": "https://www.bbc.com/news/articles/cm87987w1qz5o", "source": "BBC" },
{ "title": "British hackers face prison for GTA 6 data breach and crypto thefts", "url": "https://techcrunch.com/2026/11/18/british-hackers-face-prison-for-gta-6-data-breach-and-crypto-thefts/", "source": "TechCrunch" }
]
},
"…"
]
}

timeframe, property and category don't apply to trending_now. They're echoed only to keep every item the same shape.

Mode-specific fields:

  • by_region → regions: [{ "geo_code": "SV", "name": "El Salvador", "value": 100 }, …], sorted from highest.
  • related → related_queries: { top: [{query, value, formatted}], rising: [...] }. For rising entries, formatted is Google's label, e.g. "+250%" or "Breakout".
  • trending_now → trends: [{ query, approx_traffic, started_at, top_news_title, news: [{title, url, source}] }], up to 3 news links per trend. The dataset's Trending view shows one row per trending search.

The key-value store record OUTPUT holds the whole run as one compact JSON: { request, notes, counts: {ok, no_data, error}, items }. notes lists any adjustment made to your request. For example, trending_now with an empty geo gives ["Trending Now has no worldwide view; showing US."], and the same note starts each item's summary.

About the values

Google Trends values are relative, not absolute search volume. 100 is the peak popularity within the request: the chosen keywords, place and time range. 50 means half as popular as the peak, and 0 means too little data. In compare mode all keywords share one scale, so you can compare them. In by_region mode each keyword is scaled on its own. Dates are UTC. past_day and past_7_days return ISO datetimes, other ranges return dates. The newest point can be incomplete and is marked "partial": true, and it's left out of the trend math.

Pricing

One flat price per query, in any mode:

EventPrice
query-result: one query with 1–5 keywords, or one country's trending searches$0.0035

There's no start fee. You're charged only when the query returns data. If every keyword fails or Google has no data, the run is free. A 5-keyword comparison costs the same $0.0035 as a single keyword.

Use with Claude / MCP

Add this Actor as a tool in any MCP client through the Apify MCP server:

https://mcp.apify.com/?tools=<username>/google-trends-api

In Claude Desktop / Claude Code, add it as a remote MCP server and authenticate with Apify. Then ask things like "Is interest in Claude rising in Spain compared to ChatGPT?" The agent calls the Actor and answers from summary.

From code, with the run-sync endpoint, which returns the dataset items directly:

curl -X POST "https://api.apify.com/v2/acts/<username>~google-trends-api/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
-H 'Content-Type: application/json' \
-d '{"keywords":["claude","chatgpt"],"geo":"ES"}'

FAQ

Are these real search volumes? No. Google publishes only relative interest (0–100), and every Google Trends tool returns the same thing. See About the values.

How many keywords per run? Up to 5. Google compares at most 5 terms on one scale. For more, run the Actor several times. Values are only comparable within one run.

How reliable is it? Each request has a 7-second limit. On a rate limit (HTTP 429), a stalled connection or a network error, the Actor switches at once to a fresh residential proxy session that is already warmed up. Server errors are retried with a short backoff. One keyword's failure doesn't affect the others. If every item fails, the run is marked failed instead of returning empty "success". A scheduled self-test checks every mode daily.

Does it return related topics? Not in v1. Google returns an empty related-topics list for automated requests, even for popular keywords, so v1 returns related queries only (top and rising).

How fast is it? In 10 compare runs on Apify, half finished within about 6.5 seconds end to end (start to results, including Actor start-up), and 95% within 8 seconds. When Google rate-limits a request, the Actor switches to a fresh proxy session immediately instead of waiting. Every query is capped at 16 seconds, so a run returns results or a clear error in about 20 seconds at most, well inside the roughly 30 seconds MCP clients wait.

What are the limits? Not supported yet: related topics, more than 5 keywords per run, CSV upload, charts, absolute search volume.