Google Trends Scraper: Compare & Bulk

Get Google Trends interest over time for one keyword or thousands. Run keywords on their own scale, compare up to 5 on a shared scale, or put any number on one shared scale in bulk mode. Add interest by region and related queries if you need them.

You pay per keyword that returns data, and nothing for empty, failed or suspect results. Every run checks that Google served real data before it bills you.

Why this one

• One flat price per keyword: no start fee, no per-GB fee and no per-data-point fee.
• You are never billed for empty, failed or fake data, and every run checks for it.
• Compare up to 5 keywords, or bulk hundreds, on one shared scale.
• A clear status per keyword, so nothing fails silently.
• Works with AI agents and integrations: summary rows come first in the dataset.

Quick start

1. Type your keywords in Search terms, or paste a link in Import terms from a link.
2. Pick a Country and a Time range. Optionally turn on Include interest by region or Include related queries.
3. Click Start. When the run ends, open the Summary view, or export the dataset as CSV.

What you get

One summary row per keyword. These are real rows from a cloud run (US, past 12 months):

keyword	average	peak date	latest	change %	trend
iphone	60.54	2026-01-25	51	-17.58	down
taylor swift	24.83	2026-06-28	15	+73.61	up
sourdough	69.58	2026-01-18	61	-0.83	flat
matcha	75.96	2026-04-12	61	-11.85	down

With regions on, "air fryer" in the US over the past 12 months gave Wyoming 100, Kansas 36 and Nebraska 35 (real run).

With related queries on, the top related queries for "air fryer" include "air fryer chicken" and "ninja air fryer" (real run).

Values are Google's relative interest (0 to 100), not search volume.

Pricing

Pay per event. You pay for results, not for time.

Event	Price	Charged when
keyword	$0.002	Interest over time came back for one keyword, place and time range (success or low_precision).
region	$0.001	Region rows came back for that keyword.
related-queries	$0.001	At least one top or rising query came back for that keyword.

Prices shown are the current list prices; Apify shows the live price on the Actor page.

What 1 keyword actually costs

One keyword costs $0.002. Each add-on you switch on adds $0.001 for that keyword, and only when that add-on returned data. So a keyword with regions and related queries costs $0.004 at most.

The plain facts:

• No start fee.
• No per-GB fee.
• No per-data-point billing: a keyword with 53 weekly points and one with 2,000 hourly points cost the same.
• You are never charged for empty results (no_data), failed keywords or suspect data (suspect_decoy). Related queries marked partial (rising list held back) are not charged either; the keyword and region events still are.
• If Google blocks a keyword's region or related-queries data after retries, the keyword's timeline is still returned and charged, and the missing add-on is marked failed (see region_failure_reason and related_failure_reason on the summary row) and not charged.
• The bulk anchor keyword, its pre-pass and the data check at the start of each run are free.
• Duplicate keywords with the same place, time range, category and search type are fetched once and billed once.
• The Actor respects your max total charge. It works out how many keywords the budget covers before it starts, and lists the rest as skipped in RUN_SUMMARY.

What it does

• Interest over time for any list of keywords, in a country, a subregion or worldwide, over any time range from the past hour to 2004 to present, or a custom date range.
• Three modes. single gives each keyword its own 0 to 100 scale. compare puts up to 5 keywords on one shared scale. bulk puts any number of keywords on one shared scale by comparing them through an anchor keyword.
• Optional add-ons: interest by region (countries, subregions or cities) and top and rising related queries.
• A summary row per keyword: average, peak, latest value, change over the last periods, trend, and a breakout flag.
• Honest statuses. Every keyword ends as success, no_data, low_precision, failed or suspect_decoy, so you can see what happened to each one. Related queries can also end as partial.
• Inputs you can paste: keywords, Google Trends explore links, or a CSV / Google Sheets link with one keyword per row.

Google Trends reports relative interest on a 0 to 100 scale, not search volume. See the FAQ.

Who it is for

• SEO teams checking seasonality, rising queries and which topic is bigger.
• E-commerce sellers comparing product names by country and finding the regions that search for them most.
• Analysts pulling long keyword lists into a spreadsheet or warehouse on a schedule.
• AI agents that need a clear input, a clear status per keyword and a bill that matches what came back.

Example inputs

Field keys below are the real input keys. Everything except the search terms is optional.

SEO: seasonality and rising queries

{
  "searchTerms": ["air fryer"],
  "mode": "single",
  "geo": "US",
  "timeRange": "past_5_years",
  "includeRelatedQueries": true
}

E-commerce: which product name is bigger, and where

{
  "searchTerms": ["air fryer", "slow cooker", "rice cooker"],
  "mode": "compare",
  "geo": "GB",
  "timeRange": "past_12_months",
  "includeRegions": true,
  "regionLevel": "auto"
}

Analysts: a long list on one scale

{
  "importUrl": "https://docs.google.com/spreadsheets/d/YOUR_SHEET_ID/edit",
  "mode": "bulk",
  "timeRange": "past_90_days",
  "timezone": "America/New_York",
  "dateFormat": "plain",
  "maxKeywords": 500
}

AI agents: one keyword, one call

{
  "searchTerms": ["iphone 17"],
  "mode": "single",
  "geo": "US",
  "timeRange": "past_30_days",
  "includeSummary": true
}

This returns the full series (one timeseries row per date) plus one summary row. For a compact answer, read only the row where type is summary: it has the average, peak, latest value, change and trend.

FAQ

Can I look up a keyword that contains a comma? The data is fetched correctly, but Google Trends links treat a comma as a separator between keywords. So the source_url for such a keyword opens a search split into several keywords, not the one you sent.

Is this search volume? No. Google Trends gives relative interest, from 0 to 100, not search volume. 100 is the peak for the keyword (or for the busiest keyword in a compare or bulk set) in the place and time range you picked. It does not tell you how many searches that is.

Why do I get no_data for a keyword? Google shows no data for keywords that are too rare in the place and time range you chose. Try a wider range, a larger place (worldwide instead of a small subregion), or a more common spelling. You are not charged for it.

Why is city data missing for some keywords? City data is best for popular keywords. Small keywords often return few cities or none. Use regionLevel: "auto" for countries, subregions and metro areas, which have more data.

How do I handle very small keywords? Very small keywords are best run in their own bulk job. On a shared scale with a large keyword, a small one rounds down to 0 or 1 everywhere. Run it with other small keywords and the scale fits them.

What does suspect_decoy mean? We have observed Google serving altered data to some traffic. The Actor checks every run and every row for signs of it, and does not bill rows flagged as suspect. If a keyword comes back as suspect_decoy, you are not charged for it. If the check at the start of a run fails, the run stops and charges nothing.

What is the difference between the modes? In single mode every keyword has its own 0 to 100 scale, so a tiny keyword and a huge one both peak at 100. In compare and bulk mode they share one scale, so you can see which is bigger. Compare takes up to 5 keywords. Bulk takes any number by comparing them through an anchor keyword.

Can I get trending now, daily trends or related topics? No. This Actor returns interest over time, interest by region and related queries.

Advanced

Sample output

The dataset has one row type per type field. In the dataset, summary rows come first, then region, related query and timeseries rows. A view (for example "Summary") picks the columns but still lists every row, so in a view's CSV keep the rows whose columns are filled, or call the API with view=summary and a limit equal to your keyword count to get only the summary rows. These rows were produced by the Actor's own row builders from recorded test data for the keyword "air fryer" (US, past 12 months, single mode). The region rows come from a real run for "iphone" with the same settings.

timeseries (one row per date):

{
  "type": "timeseries",
  "keyword": "air fryer",
  "status": "success",
  "geo": "US",
  "time_range": "today 12-m",
  "category": 0,
  "search_type": "web",
  "mode": "single",
  "unit_id": "2f0714b2927b",
  "source_url": "https://trends.google.com/trends/explore?q=air%20fryer&geo=US&date=today%2012-m",
  "date": "2025-10-12T00:00:00+00:00",
  "value": 53,
  "is_partial": false,
  "scale_note": "",
  "is_anchor": false
}

The last point of a range is often still being collected, so it comes back with "is_partial": true. It is left out of the summary.

summary (one row per keyword):

{
  "type": "summary",
  "keyword": "air fryer",
  "status": "success",
  "geo": "US",
  "time_range": "today 12-m",
  "category": 0,
  "search_type": "web",
  "mode": "single",
  "unit_id": "2f0714b2927b",
  "source_url": "https://trends.google.com/trends/explore?q=air%20fryer&geo=US&date=today%2012-m",
  "average": 66.58,
  "peak_value": 100,
  "peak_date": "2026-04-12T00:00:00+00:00",
  "latest_value": 57,
  "change_pct": -14.45,
  "trend": "down",
  "breakout": false,
  "data_status": "success",
  "failure_reason": null,
  "related_status": "success",
  "region_status": "success",
  "region_failure_reason": null,
  "related_failure_reason": null,
  "is_anchor": false
}

change_pct compares the last 4 complete periods with the 4 before them. trend is up, down or flat (flat is within 10% either way). breakout is true when the latest complete period is the peak and at least twice the average.

region (one row per region, highest first):

{
  "type": "region",
  "keyword": "iphone",
  "status": "success",
  "geo": "US",
  "time_range": "today 12-m",
  "category": 0,
  "search_type": "web",
  "mode": "single",
  "unit_id": "f81e55cdde10",
  "source_url": "https://trends.google.com/trends/explore?q=iphone&geo=US&date=today%2012-m",
  "level": "subregion",
  "region_name": "Wyoming",
  "region_code": "US-WY",
  "value": 100
}

A second row from the same run: "region_name": "Louisiana", "region_code": "US-LA", "value": 63.

related_query (list is top or rising; a top row has "list": "top", "rank": 1, "query": "air fryer chicken", "value": 100, "formatted_value": "100"):

{
  "type": "related_query",
  "keyword": "air fryer",
  "status": "success",
  "geo": "US",
  "time_range": "today 12-m",
  "category": 0,
  "search_type": "web",
  "mode": "single",
  "unit_id": "2f0714b2927b",
  "source_url": "https://trends.google.com/trends/explore?q=air%20fryer&geo=US&date=today%2012-m",
  "list": "rising",
  "rank": 18,
  "query": "ninja glass air fryer",
  "value": 650,
  "formatted_value": "+650%"
}

Getting one row type only

The dataset has four views in the Apify Console: Interest over time, Summary, Interest by region and Related queries. Each view shows the columns for one row type. Apify dataset views cannot filter rows by value, so a view lists every row and leaves the columns that do not apply empty.

To get one type only, filter on the type field after you export. In Google Sheets, add a filter on the type column. With the Apify API or a client, request the dataset items and keep the rows where type equals summary (or timeseries, region, related_query). The fields parameter limits which columns come back.

Statuses

Each row carries a status. The summary row also repeats the keyword's status as data_status, and adds related_status and region_status for the add-ons, plus related_failure_reason and region_failure_reason when an add-on failed or, for partial, was held back.

Status	Meaning	Billed?
success	Interest over time came back and passed the checks.	Yes
low_precision	Data came back, but the keyword peaked under 5 on its scale (in bulk, also when the anchor was weak in its batch), so its values are coarse.	Yes
no_data	Google has no data for this keyword, place and time range.	No
failed	The keyword could not be fetched. The summary row has a failure_reason.	No
suspect_decoy	The data looked altered, so the Actor will not vouch for it.	No
partial	Only for related_status: top related queries were returned, rising ones were held back as suspect (their rows are suspect_decoy), and related queries were not charged. related_failure_reason says why.	No (related queries only)

Keywords with no_data, failed or suspect_decoy always get a summary row, even if you switch the summary off, so every keyword you sent is visible in the dataset.

low_precision is billed because data was returned. It means the keyword peaked under 5 on its scale, so its values are coarse (a single step from 1 to 2 looks like a doubling). In compare mode that is the shared scale. In bulk mode it is also set when the anchor was weak (peaked under 5) in the keyword's batch, so the shared scale is unreliable, and scale_note says which case applies. single mode never sets it. If you have very small keywords, run them in their own bulk job so they share a scale with each other and not with a giant.

Run summary and status message

While a run is going, the status message shows progress. When it ends, the message is a plain one-line recap, for example:

Done: 12 success, 1 no_data. Charged: 12 keyword.

It lists how many keywords ended in each status, what was charged, and why any keywords were skipped (your max cost, a time limit, or your maxKeywords limit).

A fuller record is saved in the run's key-value store under RUN_SUMMARY. It holds the count per status, the events charged per type, the estimated run cost, the canary result, the anchor used in bulk mode, skipped keywords and any warnings.

unit_id and source_url

Every row also carries unit_id and source_url. unit_id is a 12-character id built from the keyword, place, time range, category and search type, so the same keyword with the same settings always gets the same id. source_url is the Google Trends link for that keyword and those settings. In compare and bulk mode the link lists the keywords that were fetched together, so it opens the same comparison. unit_id stays the same from run to run. In the rare case that the platform restarts a run and the same rows are pushed twice, dedupe on unit_id (together with type and date, or region_name, or list and rank). source_url opens the same keyword, place, time range, category and search type on Google Trends, so you can check any number by eye.

All inputs

All inputs and what they do:

Field (key)	What it does
Search terms (searchTerms)	Keywords, or pasted Google Trends explore links. A link's own country, time range, category and search type apply to its terms.
Import terms from a link (importUrl)	A CSV link or a Google Sheets link shared as "anyone with the link" (a "Publish to web" link also works). One keyword per row in the first column.
Mode (mode)	single, compare (up to 5 keywords) or bulk (any number). Default single.
Anchor term (anchorTerm)	Bulk only. Leave empty to let the Actor pick a steady one.
Country or region (geo)	US, GB, or a subregion such as US-CA. Empty means worldwide.
Time range (timeRange)	From past_hour to all_time, or custom with startDate and endDate. Default past_12_months.
Category (category)	One top-level Google Trends category. Default all categories.
Search type (searchType)	web, youtube, news, images or shopping.
Include interest by region (includeRegions)	Adds region rows. Billed as an extra event.
Region detail (regionLevel)	auto or city. Auto follows Google's default level for the area: countries for worldwide, subregions (states) for a country, metro areas for a US state.
Include related queries (includeRelatedQueries)	Adds top and rising related queries. Billed as an extra event.
Include summary row (includeSummary)	One summary row per keyword. On by default.
Timezone (timezone)	IANA name such as Asia/Manila. Default UTC.
Date format (dateFormat)	iso or plain.
Maximum keywords (maxKeywords)	Safety limit for scheduled runs. Extra keywords are skipped and listed in the run summary.

Integrations

You can run the Actor from other tools with the same input. Each tool starts the Actor and reads its dataset. The setup screens differ by tool and change over time, so use each tool's Apify integration and paste the input below.

Zapier

Use the Apify integration to run the Actor, then read the dataset items it produced.

{
  "searchTerms": ["coffee", "tea"],
  "mode": "compare",
  "geo": "US",
  "timeRange": "past_12_months"
}

Back come the dataset rows: one summary row per keyword plus the timeseries rows. Filter on type in a later step if you only want the summaries.

Make

Use the Apify module to run the Actor and then get its dataset items.

{
  "importUrl": "https://docs.google.com/spreadsheets/d/YOUR_SHEET_ID/edit",
  "mode": "bulk",
  "timeRange": "past_90_days",
  "maxKeywords": 200
}

Back come the same rows as in the sample output, one item per row, which you can map to your next module.

Google Sheets

To send keywords in, use importUrl with a sheet shared as "anyone with the link", or a sheet published to the web (File, Share, Publish to web). To get results out, open the run's dataset in the Apify Console, choose Export, and pick CSV, then open the file in Google Sheets (File, Import). The dataset also has a CSV download link you can use the same way. The "All fields" CSV export lists columns alphabetically; export a view (Summary, Interest over time, Interest by region, Related queries) for tidy columns.

{
  "importUrl": "https://docs.google.com/spreadsheets/d/YOUR_SHEET_ID/edit",
  "mode": "single",
  "geo": "US"
}

Back comes a dataset you can export as CSV and open in a sheet. Filter the type column to see one row type at a time.

Apify MCP from Claude

Connect Claude to the Apify MCP server, then ask Claude to run this Actor. The input is the same JSON.

{
  "searchTerms": ["iphone 17", "pixel 10"],
  "mode": "compare",
  "geo": "US",
  "timeRange": "past_30_days"
}

Claude gets the dataset rows back, including the status for each keyword and the summary row, so it can tell a keyword with no data from a failed one.

Issues

Report problems in the Issues tab; the developer answers within 2-3 days. Include the run link and the keyword that went wrong if you can.