Pinterest Trends Scraper

Scrape Pinterest Trends keyword tables for supported markets and export ranked search trends to JSON, CSV, or Excel. This Pinterest Trends API alternative collects normalized demand, growth values, seasonality, and the date of the underlying data without requiring a Pinterest account for the current public table resource.

Select one or more markets and trend types, then add optional topic and date filters. Each result is a flat keyword record suitable for a spreadsheet, dashboard, or scheduled report. The source can change its public resource or market availability; the actor reports an error if no usable rows are returned.

What data can you extract from Pinterest Trends?

The actor writes one row for each keyword in a requested market and trend table. Numeric values below are the values Pinterest returns. In particular, searchCount is a score in the table response, not an estimate of the number of searches performed by people.

Field	Description	Example
term	Search phrase in the trend table	3 screen wallpaper
country	Market used for the request; grouped markets use their canonical code	US
trendType	Selected table	growing
rank	Position within that market and table, beginning at one	1
normalizedCount	Normalized search score supplied by Pinterest	55
searchCount	Search count score supplied by Pinterest	55
weeklyChange	Week-over-week change value, when supplied	0.5
monthlyChange	Month-over-month change value, when supplied	100.01
yearlyChange	Year-over-year change value, when supplied	100.01
seasonalityScore	Seasonal relevance score	0.3215189
affinity	Interest affinity value, when supplied	null
lookbackWindow	Calculation window requested in the input	30D
requestedEndDate	Date sent to the table resource	2026-09-26
dataEndDate	Latest data date reported in the table response	2026-09-22
pinterestTrendsUrl	Link to the keyword detail view	https://trends.pinterest.com/detail/?terms=3+screen+wallpaper&country=US
scrapedAt	UTC time when this run collected the row	2026-09-26T21:12:32.145884Z

Missing numeric metrics are null. Every row retains the requested window and both dates, so a delayed data refresh is visible in the export. The keyword detail link is useful for review in a signed-in browser; the detail page currently sends anonymous visitors to a sign-in prompt. The ranked table data itself was accessible without signing in during the verified run.

How to scrape Pinterest Trends

1. Choose a country or region, such as US, and one or more tables. growing is the default table.
2. Set maxResultsPerCountry from 1 to 100. The limit applies separately to each distinct market and trend type. Set maxItems if the complete run needs a smaller total cap.
3. Optionally set a lookback window, end date, keyword filters, interest IDs, or ranking method. Keep filters empty on the first run to see the available terms.
4. Run the actor. It collects the public table response for each selected combination and writes one row for each returned keyword.
5. Open the default dataset and export it as JSON, CSV, or Excel. Use country, trendType, and dataEndDate as grouping columns when comparing runs.

For a first check, request 50 US growing trends. A verified run with that input returned 50 rows. Pinterest reported a data end date four days earlier than the requested end date, which is why both dates appear in the output.

Input

Name	Type	Default	Description
countries	String list	US	Markets to query. Some country choices resolve to a larger regional market.
trendTypes	String list	growing	Any of top_monthly, top_yearly, growing, and seasonal.
maxResultsPerCountry	Integer	50	Maximum rows for each market and type, from 1 to 100.
maxItems	Integer	0	Total output cap across all tables; zero means no total cap.
lookbackWindow	String	30D	Calculation window: 7D, 30D, 90D, 180D, or 365D.
endDate	Date string	Current UTC date	Requested last date, in YYYY-MM-DD format.
keywordsToInclude	String list	Empty	Search terms to include in the table request.
keywordsToExclude	String list	Empty	Search terms to remove from the table request.
interestIds	String list	Empty	Top-level Pinterest interest IDs for category-specific rows.
rankingMethod	String	Unset	Optional ordering: TREND_RANK, TOTAL_SEARCH, WEEKLY_CHANGE, MONTHLY_CHANGE, or YEARLY_CHANGE.
proxyConfiguration	Object	Platform proxy	Optional proxy settings.

Example input for two markets and one topic:

{
  "countries": ["US", "CA"],
  "trendTypes": ["growing", "seasonal"],
  "maxResultsPerCountry": 50,
  "lookbackWindow": "30D",
  "keywordsToInclude": ["nails"],
  "maxItems": 150
}

This example can request four separate tables, up to 50 rows in each. The total cap stops the run at 150 saved rows even if all four tables contain 50. A keyword filter can make any table shorter, so the final count may be lower. The actor does not fill missing rows with unrelated topics.

For a category request, use an interest ID such as 935541271955, the top-level beauty interest in the current site configuration. Interest IDs are source identifiers, not category names. Invalid or outdated IDs can return an empty table. Start with one market and one category, check the terms, then expand the request.

The country selector includes standalone markets and country aliases for grouped regions. For example, AT and CH resolve to DE+AT+CH; SE, DK, FI, and NO resolve to SE+DK+FI+NO; and NL, BE, and LU resolve to NL+BE+LU. If a run requests both an alias and its canonical region with the same trend type, the actor collects that table once. The country field contains the canonical region used in the request.

Output example

This is the first item from the verified 50-row US growing-trends run. The large growth values are source values, not a promise of a particular percentage change or future result.

{
  "term": "3 screen wallpaper",
  "country": "US",
  "trendType": "growing",
  "rank": 1,
  "normalizedCount": 55,
  "searchCount": 55,
  "weeklyChange": 0.5,
  "monthlyChange": 100.01,
  "yearlyChange": 100.01,
  "seasonalityScore": 0.3215189,
  "affinity": null,
  "lookbackWindow": "30D",
  "requestedEndDate": "2026-09-26",
  "dataEndDate": "2026-09-22",
  "pinterestTrendsUrl": "https://trends.pinterest.com/detail/?terms=3+screen+wallpaper&country=US",
  "scrapedAt": "2026-09-26T21:12:32.145884Z"
}

The source supplies the ranked table and scores. The actor adds the market, type, dates, link, and extraction timestamp so the result can stand alone outside the website. rank restarts at one for each table. It follows the returned order; changing the selected trend type, date, filters, or ranking method can change that order. A term can appear in more than one table, which is expected when it qualifies for multiple trend types.

Use cases

• Editorial calendars: A content planner can collect growing and seasonal terms every week, then map terms to a publish date. Save dataEndDate with each snapshot so a delayed refresh is not mistaken for a new trend.
• Product exploration: A merchant can filter trend terms around a product category and compare the current list with earlier exports. The actor provides search interest signals, not product sales or inventory data.
• Regional research: A market analyst can run the same type and window across countries, then compare keyword sets by country. Grouped markets should be treated as one region rather than independent country measurements.
• Creative briefs: A campaign team can find rising visual themes, inspect the detail links in a signed-in browser, and turn selected phrases into a brief. The dataset contains keywords and scores, not pins or creative assets.
• Seasonal monitoring: A researcher can schedule seasonal and top_yearly tables to identify terms that recur around holidays or events. A single score should be checked alongside multiple dates before drawing a conclusion.
• Topic lists: A team can maintain include and exclude terms for a subject area, export clean rows, and use those terms as seeds for further research.

How much does it cost to scrape Pinterest Trends?

The current private actor price is $0.001 per result, or $1.00 per 1,000 saved rows. The charge is based on rows delivered to the dataset. A 50-row run costs $0.05 in result events. Four full 50-row tables cost $0.20. Ten full 100-row tables produce 1,000 rows and cost $1.00. A filter that returns fewer rows also produces fewer result events.

There is no separate start event in the current price configuration. A configured spending limit can stop output before the requested count. Prices shown here reflect the current actor configuration and may change in a later version; inspect the actor's price panel before a large run. The maximum of 100 applies to each market and trend type, so a 1,000-row collection requires at least ten distinct tables with enough available results.

Tips for best results

Start with US, growing, and 50 rows to verify the current source response. Add markets and types after checking the first dataset. The actor queries each distinct market and type once. It does not need to load the detail page for each keyword, which keeps requests and run time proportional to the number of tables rather than the number of rows.

Use endDate when comparing historical snapshots, and always retain dataEndDate. Pinterest can return its latest available data date even when the requested date is later. The data date is therefore a more precise label for the underlying observations than the time at which the run completed.

Use keywordsToInclude for a focused topic and keywordsToExclude to remove phrases that match a broad term but are outside the project. The filters are passed to Pinterest's table resource before rows are saved. A filter that is too narrow can return fewer than the requested 50 or 100 rows. Run the same request without filters to see whether the market and table contain data.

Choose a ranking method when the order matters to downstream work. TREND_RANK represents the trend list's rank, while TOTAL_SEARCH, WEEKLY_CHANGE, MONTHLY_CHANGE, and YEARLY_CHANGE ask the source to rank by those metrics. The numeric fields still describe the individual term. Do not treat a high normalized score as a raw count of searches.

For grouped markets, compare runs using the country value in output. If you submit both AT and CH, both refer to the same Germanics table and duplicate requests are skipped. If your analysis requires a country-only measurement, check that the selected market is standalone before combining it with another dataset.

Limits

Pinterest's current table endpoint accepts at most 100 rows per market and trend type. Requests for 125, 200, and 250 rows returned HTTP 400 during source checks. The input schema enforces the verified limit. You can collect more than 100 total rows by selecting multiple distinct market and type combinations; the actor will not imply that extra rows exist in a single capped table.

The source may provide fewer rows for a narrow keyword filter, category, market, or historical date. Some numeric fields are absent for particular terms, and affinity was empty in the verified general US table. Those values remain null. The actor does not infer them from other metrics. Growth values are retained exactly as the table reports them, including high capped-looking values such as 100.01; the source response does not explain every value's calculation.

The public home page responded without a challenge during source checks. Search and detail pages redirected anonymous visitors to a sign-in prompt, but the table JSON resource responded successfully without an account. A future access change can affect this actor. Temporary server failures are retried briefly. An invalid filter or unavailable market can fail a table; in a multi-table run, the remaining tables continue. If no table yields rows, the run fails with a clear error instead of producing a misleading empty success.

The dataset contains trend keywords only. It does not provide pin records, boards, accounts, audience demographics, raw search volume, product conversions, or forecast certainty. Scores should be interpreted as source-specific signals and checked against the original site when used for high-impact decisions.

Use with AI agents (MCP)

Add this actor to any MCP client (Claude, Cursor, VS Code, n8n...) with one URL:

https://mcp.apify.com?tools=mlg14/pinterest-trends-scraper

The actor returns one flat record per trend keyword with rank, growth and seasonality scores and the data date. Keep maxResultsPerCountry small to bound results and cost.

Prompt 1: "What are the fastest growing Pinterest trends in the US right now?"

{
  "countries": [
    "US"
  ],
  "trendTypes": [
    "growing"
  ],
  "maxResultsPerCountry": 30
}

Prompt 2: "Show top monthly Pinterest trends about home decor in Germany and France."

{
  "countries": [
    "DE",
    "FR"
  ],
  "trendTypes": [
    "top_monthly"
  ],
  "keywordsToInclude": [
    "home decor"
  ],
  "maxResultsPerCountry": 30
}

Prompt 3: "Get seasonal Pinterest trends for the UK and Ireland over the last 90 days."

{
  "countries": [
    "GB+IE"
  ],
  "trendTypes": [
    "seasonal"
  ],
  "lookbackWindow": "90D",
  "maxResultsPerCountry": 30
}

FAQ

Does this require a Pinterest account?

No account was required for the public ranked-table resource during the verified run. Opening a keyword detail link in a browser may require sign-in. The actor does not accept account credentials or collect private account data.

Is it legal to collect these trends?

This actor reads publicly available trend-table data. Check the site's terms and applicable privacy or data-use rules for your use case. Do not use exported data for unlawful profiling or unwanted messages. The output consists of public keywords and market-level scores rather than personal records.

Do I need to configure a proxy?

The default platform proxy is configured for requests. A normal successful table request uses direct HTTP through that network path. If the site begins blocking requests, the fetch layer can rotate its session and escalate to a residential proxy. That escalation may affect platform usage, so use focused limits for routine schedules.

How fast is a run?

Run time depends mainly on the number of selected market and type combinations. A single table returns up to 100 rows in one source request. Network delays or retries can extend a run. The verified golden run returned 50 rows; use a small first run to observe timing in your account.

Can I schedule and monitor it?

Yes. Create a recurring run and inspect the dataset count and dataEndDate after each run. A table returning fewer rows can reflect a filter or source change. A failed run with no rows is an explicit signal to review the input or current access.

Can I export to a spreadsheet?

Yes. Download the default dataset as CSV or Excel, or read the JSON dataset through the platform API. Keep the date and market columns when joining exports from different runs. Numeric score fields can be compared within the same table definition, but should not be relabeled as counts of searches.

Why is a metric empty or the returned data date older?

Pinterest does not supply every metric for every keyword, so missing values are null. Its latest published trend date can also lag behind the requested date. The actor records both requestedEndDate and dataEndDate to show that difference. Try a broader filter if the entire table is empty.

Integrations

The default dataset can be read through the platform API, downloaded to a spreadsheet, or sent to a workflow through a webhook. A scheduled run can update a reporting table with the latest rows; preserve the market, trend type, requested date, and source data date as separate columns. For a topic tracker, append each scheduled export rather than replacing previous snapshots, then compare terms only across equivalent market, type, and window settings.

Support

Open an issue on the Issues tab; we reply within 24h and add fields on request.