Google Trends Scraper & API - Trending Now & Trending Searches
Pricing
from $0.30 / 1,000 results
Google Trends Scraper & API - Trending Now & Trending Searches
Daily and real time trending searches from Google Trends Trending Now, with search volume (100+ countries): active vs ended trends, trend duration. Keywords mode: interest over time, by region, related queries (rising, breakouts). Google Trends API and pytrends alternative. $0.01/run + $0.0005/item.
Pricing
from $0.30 / 1,000 results
Rating
0.0
(0)
Developer
Orbots
Maintained by CommunityActor stats
0
Bookmarked
21
Total users
14
Monthly active users
10 days ago
Last modified
Categories
Features
Share
Unofficial tool. Orbots is not affiliated with, sponsored by, or endorsed by Google. "Google" and "Google Trends" are trademarks of Google LLC, used here only to describe the public data this Actor collects.
Get what the world is searching right now: daily and real time trending searches from Google Trends Trending Now, with estimated search volume, for 100+ countries, plus interest over time, interest by region and related queries for any keyword. An unofficial Google Trends API alternative and a maintained pytrends alternative: no browser, no login, no item cap. You only pay for results that are actually delivered.
Why this Actor
Google retired the old dailytrends and realtimetrends endpoints. They now return 404, and tools still built on them silently return nothing for trending data. This Actor reads the Trending Now list from the Trending Now page itself, and keyword data from the classic Explore API of Google Trends (see Keywords mode for its limits).
- High volume per run. Large markets return the most. In one run on 2026-11-05 (about 20:20 UTC), the 24-hour lists held 480 trending terms for the US, 453 for India, 440 for Germany, 397 for Japan, 315 for Brazil and 309 for the UK. With California and Paris added, that run returned 2,672 rows in 137 seconds. The US list moves a lot from day to day: 341 to 898 terms in our 31 daily runs from 2026-10-06 to 2026-11-05 (median 518). Smaller markets return far fewer: across 125 countries, the median 7-day list held 289 terms on 2026-11-05. Yield depends on how much is actually trending in that market, and every row is billed (see Pricing), so we publish the measured numbers rather than a flattering range.
- Search volume included, on every term. Volume was present on all 2,672 rows of the run above, and on all 64,759 terms of the 7-day lists of 125 countries on 2026-11-05 (for example
100000,500000). Related searches come along when Google provides them: on 24% of the rows of that run (US 32%, Germany 18%), and on 29% of the terms in those 7-day lists (31% in the 10 largest markets, 23% in the 25 smallest). Most trending scrapers return only the term. - Category, rank and explore link on every term. Each row is classified into Google's own topic categories (Sports, Entertainment, Business and Finance, and more), carries its position in the feed and a direct link to the term's Google Trends page. Filter to one vertical with the
categoriesinput, no post-processing needed. - Active trends vs ended trends, with trend duration. Every term says whether it is still trending right now (
is_active) and, once it has ended, exactly how long it lasted (duration_minutes, plusstarted_atandended_at). A ranked list alone cannot tell a two-hour spike from a two-day story. Measured on the US 24-hour list of the run above: of 480 terms, 96 were still active and 384 had ended, median life 2 hours 30 minutes. In a US 48-hour list pulled the same evening, 1,200 of 1,347 terms had already ended. Field coverage is complete on ended terms. - Built for schedules. Turn on
onlyNewand each run returns only trends you have not received before, per country and window. When runs overlap (a 7-day window pulled daily, or a 24-hour window pulled every few hours), you pay for each trend once instead of once per run. - No item cap. Some competitors cap at 1,000 items per run. This one does not.
- Honest billing.
maxItemsis respected exactly, and nothing is charged for a result that was not returned.
Reliability, measured
- 293 of 301 third-party runs succeeded, lifetime (as of 2026-11-06). The public 30-day figure on this page read 148 of 153 that day; the two failures of 5 November 2026 fell in the hours between Google's change to the Trending Now page and release 0.1.23, which fixed it the same day. The denominator is still modest and we publish it anyway. The logs of a failed third-party run are not visible to us (Apify shows a user's runs to the developer only with the user's consent), so release 0.1.21 closed the failure paths that were ours to close; the changelog lists them.
- The design behind that number: when Google throttles part of a request, the run delivers the countries that answered and names the ones that did not, instead of failing everything; a country that times out, drops the connection or answers with something unexpected is skipped and named the same way; invalid input (a country code that is not an ISO code, keywords mode with no terms) ends as an explained empty run that costs nothing, not as a failure; a quiet day with
onlyNewfinishes as a success with a message, not a fake failure; the session and the IP are rotated together whenever Google refuses one. - For scale comparison, the same public stat on
memo23/google-trends-scraperreads 90.2% (147 of 163 runs, last 30 days, as of 2026-11-06). Every Actor's success rate is public on its page. Check it before you put anything on a schedule.
What you get
Trending Now mode (default)
{"type": "trending_now","term": "japan earthquake","rank": 3,"geo": "US","started_at": 1785224400,"ended_at": 1785238800,"is_active": false,"duration_minutes": 240,"search_volume": 200000,"formatted_volume": "200K+","growth": 1000,"category": "Law and Government","categories": ["Law and Government"],"related_terms": ["japan earthquake today", "earthquake japan", "kumamoto"],"trends_url": "https://trends.google.com/trends/explore?q=japan+earthquake&date=now+1-d&geo=US&hl=en-US","window_hours": 24,"scraped_at": "2026-07-29T12:53:25Z"}
Every term arrives ranked (1 = hottest), classified into Google's own topic categories (Sports, Entertainment, Business and Finance, and more), with the volume both as a number you can sort by and as the label the interface shows (200K+), plus a ready-made link to the term's explore page. Windows: past 4 hours, 24 hours, 48 hours or 7 days. Pass several countries in one run.
is_active is true while the term is still trending and ended_at is null; once Google closes the trend, ended_at is filled and duration_minutes tells you how long it held. Sort by duration_minutes to separate a passing spike from a story with legs, or keep is_active == true to act only on what is live at this minute.
Want just one vertical? Set categories (e.g. ["Sports"]) and only trends in those categories are delivered: the same 19 topic categories Google itself assigns, matched with no post-processing on your side. A term Google did not categorise is excluded while the filter is on, and rows skipped by the filter are never charged. If the filter eats every row, the run still succeeds, with a message saying exactly that.
Only what is new: onlyNew
Set onlyNew: true and the Actor remembers, per country and per window, which trends it has already delivered to you, and returns only the ones it has not. A schedule then behaves the way you would want it to: the first run gives you the whole list, and every run after that gives you only the additions. How many additions you get depends on how much your runs overlap. The 24-hour window holds the trends that started in the last 24 hours, so a run one day later shares almost none of them: on a daily 24-hour schedule the additions are about a full list each day (see Pricing).
A term that trends again later counts as new, because history is keyed by term and start time, so a recurring topic is not swallowed. History is kept for 7 days in your own account, one entry per country and window. If a run finds nothing new, it finishes successfully with a message saying so, and charges you nothing: no rows, no start fee. Leave onlyNew off to always get the full list.
Keywords mode
{ "type": "interest_over_time", "keyword": "chatgpt", "date": "Jul 27 - Aug 2, 2025", "timestamp": 1753574400, "value": 91, "is_partial": false, "property": "web" }{ "type": "interest_by_region", "keyword": "chatgpt", "region": "California", "region_code": "US-CA", "value": 5, "property": "web" }{ "type": "related_queries", "keyword": "mortgage rates", "ranking": "top", "rank": 1, "query": "mortgage rates today", "value": 100, "formatted_value": "100", "is_breakout": false, "link": "https://trends.google.com/trends/explore?q=mortgage+rates+today&date=today+12-m&geo=US", "property": "web" }{ "type": "related_queries", "keyword": "mortgage rates", "ranking": "rising", "rank": 1, "query": "50 year mortgage rates", "value": 550, "formatted_value": "+550%", "is_breakout": false, "link": "https://trends.google.com/trends/explore?q=50+year+mortgage+rates&date=today+12-m&geo=US", "property": "web" }
date is Google's own label for the period and can contain non-ASCII dashes and spaces; parse dates from timestamp. Compares up to 5 terms at a time; longer keyword lists are split into groups of 5 automatically. Time ranges from the past hour to 2004-present. Interest can be measured on five Google surfaces (regular Web Search, Image Search, News, YouTube or Shopping, via the property input), which most tools in this category do not expose.
Related queries are an extra block: add related_queries to dataTypes (it is off by default, so existing schedules keep their cost). For each term you get Google's top list (the queries searched most often together with it, scored 0-100) and its rising list (the queries whose searches grew the most, with the growth in percent), up to 25 rows each, in Google's own order (rank 1 = first). formatted_value is the label Google shows, in the language you pick. is_breakout is true when Google shows Breakout instead of a percentage, which it does for growth above 5,000%; a translated label counts too. link opens the related query on Google Trends. When Google returns an empty related-queries list for a term (typical of a term with little search volume in that country and time range), the term simply gets no related rows, and the status message names it. Related topics are not offered: Google answers that list empty to automated sessions.
Read this before choosing Keywords mode. Keyword queries go through the classic Explore service of Google Trends, which limits every IP address. Measured in November 2026 on residential addresses: in our rate-limit test 166 of 166 requests were answered once the session cookie was in place, and in 15 of our own runs on the Apify platform the same day (10,177 rows) Google refused 2 of 187 requests and one connection dropped; each time the Actor moved to a fresh address and fetched only the missing part, so no row was lost. The busiest address served 18 requests without a refusal. The Actor warms up one session per address and lets it serve every block of a query and the queries after it; it moves to a fresh address after 12 data calls, or the moment Google refuses one, and then fetches only what is still missing, so rows already collected are kept. If Google refuses two groups of terms in a row on every address, the run stops asking for what keeps being refused (the remaining groups, or just their related queries) instead of spending more addresses on it, and the status message names what was left out. By default keywords mode runs on residential proxy from the country in geo, because in our tests Google refused datacenter addresses on keyword queries almost at once; any other proxy setting you choose is respected. Long lists take time: 20 terms took about 2 minutes with the two default data types and 5 to 7 minutes with all three (one country, past 12 months; the slower run had four refusals to recover from), so by our estimate the default 1-hour run timeout fits about 150 terms with all three, fewer worldwide or over 5 years. Split longer lists across runs or raise the timeout. Google has announced that the classic Explore is "expiring soon", without a date. If it goes away, keywords mode stops working until it is rebuilt on a replacement. When Google answers the classic API with "not found" or redirects it to the new Explore, the run stops once two addresses have shown it, with a message that names the cause, instead of retrying address after address. Trending Now does not use that service, is not affected, and remains the faster mode. A run that collects nothing costs you nothing (see Pricing).
Sub-regions, not just countries
geo and trendingGeos also accept sub-region codes, which most tools in this category do not expose: US-CA (California), US-NY, FR-75 (Paris), GB-ENG, BR-SP. A sub-region returns a narrower, more local list. In a 24-hour run on 2026-11-05, US-CA returned 228 trending terms and FR-75 returned 50.
Input
| Field | What it does |
|---|---|
mode | trending (default) or keywords |
trendingGeos | Country codes for Trending Now, e.g. ["US","BR","GB"] |
trendingHours | 4, 24, 48 or 168 |
categories | Trending mode: keep only trends in these Google topic categories (19 available, the ones Google itself assigns). Empty = all |
keywords | Terms to analyse in keywords mode |
geo | Country code; empty means worldwide |
timeframe | Time range for keyword analysis |
property | Surface to measure: web (default), images, news, youtube, shopping |
dataTypes | interest_over_time, interest_by_region (the default) and related_queries (top and rising related queries per term, off by default) |
onlyNew | Trending mode: return only trends not delivered in an earlier run. Made for schedules |
maxItems | Hard cap, respected exactly, never charged beyond |
proxyConfiguration | Keywords mode replaces the default setting with residential proxy from the geo country; any other setting here (a group, a country, your own proxies or no proxy) is used as you set it. Trending mode works on the default (see below) |
When a run returns nothing
Zero rows is not always an error, and this Actor does not pretend otherwise:
| Situation | Run status | Charged |
|---|---|---|
| Google throttled every country you asked for | Failed | nothing |
| Some countries throttled, the rest answered | Succeeded: the reachable countries are delivered, the throttled ones are named in the log | only the delivered rows |
| The endpoint answered but returned no data at all | Failed | nothing |
onlyNew is on and nothing is new since your last run | Succeeded, with a message saying so | nothing |
Your run charge limit (maxTotalChargeUsd) is below the start fee plus one row | Succeeded, with a message saying so | nothing |
Keywords mode, only related_queries asked, and Google answers them empty for the terms | Succeeded, with a message saying so | nothing |
| Keywords mode, Google refuses two groups of terms in a row on every address | Failed if nothing was delivered yet; otherwise Succeeded, with the terms left out named in the status message | only rows delivered before it |
| Keywords mode while the classic Explore is gone (Google is retiring it) | Failed once two addresses see the classic API answer "not found" or redirect to the new Explore, with a message saying so | only rows delivered before it |
That distinction matters if you run this on a schedule: a quiet day should not page you, and a real breakage should.
Proxy is required, not optional
Google rate-limits by IP, so by default every run goes through Apify Proxy. Keywords mode sets itself up on the RESIDENTIAL group, from the country in geo (from any country for CN, since Google does not answer from mainland China), unless you change the proxy setting in the input: in our tests Google refused datacenter addresses on keyword queries almost at once. If residential proxy is not available to the run, keywords mode falls back to the proxy set in the input and says so in the log. Trending Now reads one page per country, after a short session warm-up, and works on the default proxy. In both modes the Actor moves to a fresh IP whenever Google refuses one and re-establishes the session from scratch, which is why it keeps returning data where simpler scrapers fail.
Notes and limits
- Related topics are not included. Google answers the related topics list empty to automated sessions (checked again in November 2026), and we would rather ship what works than advertise a field that returns empty. Related queries are included: see Keywords mode.
- Trending Now is far faster than keyword analysis; if you need breadth, start there.
- Data is public, aggregated search interest. No personal data is collected.
Pricing: what a run actually costs
Pay per event: $0.01 per run, charged only when the first row is delivered, plus a per-row fee that scales down with your Apify plan tier (FREE/BRONZE/SILVER $0.0005, GOLD $0.0003, PLATINUM $0.00015, DIAMOND $0.0001). Platform compute is on us: nothing lands on your Apify bill beyond these events. A run that delivers zero rows costs $0.00: no rows, no start fee. Every row type costs the same, related queries included.
What one row is. In Trending Now mode, one row is one trending term. In Keywords mode, one row is one data point: a point of a term's curve (one week, over the past 12 months), a region, or a related query. So in Keywords mode a term costs as many rows as Google returns points for it, and the period you pick sets most of the price.
Trending Now: cost per run
Rows measured in our own pulls. Cost at FREE to SILVER, start fee included.
| What you run | Rows | Cost |
|---|---|---|
| US, 24-hour window (the default), one run | 341 to 898 in 31 daily runs from 2026-10-06 to 2026-11-05, median 518 | $0.18 to $0.46, median $0.27 |
| US, 24-hour window, once a day for 30 days | 16,511 in the 30 days to 2026-11-05 | about $8.56 a month |
| UK, 24-hour window, one run | 287 to 321 in our runs from 2026-10-30 to 2026-11-05 | $0.15 to $0.17 |
| Eight locations (US, India, Germany, Japan, Brazil, UK, California, Paris), 24-hour window, one run | 2,672 (2026-11-05) | $1.35 |
| US, 4-hour window, one run | 29 to 136 (2026-10-30 and 2026-11-02) | $0.02 to $0.08 |
| US, 48-hour window, one run | 1,347 (2026-11-05) | $0.68 |
| 7-day window, one run, eight largest markets | 1,852 (Brazil) to 2,646 (US) on 2026-11-05; the US list covered less than 5 days | $0.94 to $1.33 |
| A run that delivers zero rows | 0 | $0.00 |
At GOLD / PLATINUM / DIAMOND the median US run above costs about $0.17 / $0.09 / $0.06.
onlyNew does not make a day of new trends cheaper. The 24-hour window holds the trends that started in the last 24 hours, so two runs a day apart share almost none of them, and the daily schedule above costs about the same with onlyNew on. It saves money when runs overlap: pull the 7-day window every day, or the 24-hour window every few hours, and onlyNew charges each trend once instead of once per run. The 7-day list holds fewer trends per day than the 24-hour list (in the US on 2026-11-05: 471 trends that had started in the previous 24 hours, against 652 in the 24-hour list pulled the same hour), and in the US it stops at about 2,600 terms, less than 5 days. So switching windows changes what you get, not only what you pay.
Keywords mode: cost per term
Measured in the US on 2026-11-04 and 2026-11-05. Add $0.01 once per run, not per term.
| Period and data | Rows per term | FREE to SILVER | GOLD | PLATINUM | DIAMOND |
|---|---|---|---|---|---|
| Past 12 months (the default), interest over time and by region (the default) | about 104 (53 + 51) | $0.052 | $0.031 | $0.016 | $0.010 |
| Past 12 months, all three blocks | up to about 154 (53 + 51 + up to 50 related) | up to $0.077 | up to $0.046 | up to $0.023 | up to $0.015 |
| Past 5 years, interest over time only | about 262 | $0.131 | $0.079 | $0.039 | $0.026 |
Interest by region is one row per region: 51 in the US; worldwide (an empty geo) it lists every country, 250 in our run of 2026-11-02, so a worldwide term costs about three times as much (303 rows against 104, past 12 months, curve and regions). Other periods bill the same way, one row per point Google returns, and Google sets the spacing of the points: weekly over the past 12 months and 5 years, daily over the past 30 or 90 days, hourly over the past 7 days, and in steps of minutes over the past day or hours. So a shorter period does not mean fewer rows. Long lists add up too: 100 terms over 5 years, curve only, are 26,200 rows, $13.11 at FREE to SILVER. If that is your job, an Actor that bills one result per term costs less (see below).
Compared honestly (public prices, as of 2026-11-06)
The three Actors bill different units, so this table compares the same jobs, not the price of one result.
| This Actor | apify/google-trends-scraper | memo23/google-trends-scraper | |
|---|---|---|---|
| One billed result | one row: a trending term, or one data point of a keyword | one search term, with its curve, regions and related queries inside | your choice: one term with every chart inside (perTerm), one chart (nested), or one data point (flat, used when an API call leaves the field out) |
| Price per result | FREE to SILVER $0.0005, GOLD $0.0003, PLATINUM $0.00015, DIAMOND $0.0001 | FREE $0.003, BRONZE $0.001, SILVER $0.0005, GOLD $0.0003, PLATINUM $0.00015, DIAMOND $0.0001 | $0.00075 at every tier |
| Start fee | $0.01, only if rows are delivered | none | $0.001 per GB of run memory from 2026-11-16 ($0.00005 before), also on runs that deliver nothing |
| Platform compute | included | billed to your account on top | included |
| Trending Now, 518 rows (our median US 24-hour list), FREE to SILVER | $0.27 | no Trending Now output | about $0.39 |
| The same 518 rows at GOLD / DIAMOND | $0.17 / $0.06 | no Trending Now output | about $0.39 |
| Keywords, per term (start fees apart), past 12 months, curve and regions, FREE | $0.052 | $0.003 plus compute | $0.00075 (perTerm), about $0.078 (flat) |
| Keywords, per term (start fees apart), past 5 years, curve only, FREE | $0.131 | $0.003 plus compute | $0.00075 (perTerm), about $0.20 (flat) |
| The same two keyword jobs at SILVER | $0.052 and $0.131 | $0.0005 plus compute | $0.00075 (perTerm) |
Start fees come on top of the keyword rows, once per run: $0.01 here, $0.001 per GB of run memory at memo23 from 2026-11-16, none at the official Actor. A 1-term, 12-month run in the US with the default data types therefore costs $0.062 here at FREE to SILVER.
Read that table the honest way:
- Trending Now. The official Actor does not list Trending Now among its outputs; its README points to a separate Actor for trending searches. memo23 offers it at $0.00075 per trending term against our $0.0005. With its start fee from 2026-11-16 ($0.001 against our $0.01), it costs less on runs under 36 rows (under 20 on GOLD, fewer on PLATINUM and DIAMOND), and we cost less above that.
- Keywords. memo23's
perTermoutput costs less per term than we do, at every tier and for every period: $0.00075 a term, all charts in one row. So does the official Actor's price per term before its compute: $0.003 on FREE, down to $0.0001 on DIAMOND. We did not measure that compute. We bill every data point as its own row, so a 12-month term costs $0.052 here at FREE to SILVER and a 5-year curve $0.131. For any keyword list, a per-term Actor is the cheaper choice; the gap grows with the number of points per term. Our row costs less than memo23'sflatrow ($0.0005 against $0.00075), which matters only if you want one row per data point. - A blanket "cheapest" claim would be false, so we are not making one.
Changelog
- 0.1.24 (2026-11-06): billing guard. A run whose charge limit (
maxTotalChargeUsd) cannot cover the start fee plus one row now ends as an explained empty run with nothing charged, before any proxy or fetch, instead of delivering rows without the start fee or charging the start fee without delivering a row. A row the limit did not allow is no longer counted as delivered, so the status message and theonlyNewhistory are exact when a run stops at its limit, and the status message now says when that happened. README: the pricing section now gives the cost per run in Trending Now and per term in Keywords mode, from volumes measured in November 2026, and compares the three Actors on the same jobs; the volume figures elsewhere were re-measured too. The input descriptions oftimeframeandtrendingHoursnow say how many rows to expect, and the one ofonlyNewsays when it saves money. Nothing else changed in how runs work or what they cost. - 0.1.23 (2026-11-05): Trending Now works again. On 5 November 2026 Google moved the Trending Now list into the Trending Now page itself, and the feed this Actor used to call started answering with an error, so trending runs found no terms and failed. The Actor now reads the list from that page, with the same rows and fields as before; if the page ever comes without the list, it falls back to the old feed, and if both come back empty it switches address instead of reporting a country with nothing trending. Keywords mode, inputs and prices are unchanged.
- 0.1.22 (2026-11-02): keywords mode rebuilt on new measurements, plus maintenance. Keywords: one warmed-up session per address now serves every block of a query and the queries after it; the Actor moves to a fresh address after 12 data calls or the moment Google refuses one, and fetches only what is missing, so rows already collected are kept. A redirect to Google's verification page, or a page where data should be, now switches address instead of dropping the group. Keywords mode runs on residential proxy from the
geocountry by default (any other proxy setting in the input is respected; trending mode is unchanged). Newrelated_queriesdata type: top and rising related queries per term, off by default, charged like any other row. If Google refuses two groups of terms in a row on every address, the run stops asking for what keeps being refused, and the status message names what was left out; a group lost entirely is now named there too. If Google retires the classic Explore by answering "not found" or redirecting to the new Explore, keywords runs stop once two addresses show it, with a message saying so. When every target is refused, the error now names the proxy the run used and what to change. Maintenance: a trending answer cut short in transit now switches address instead of dropping the country. The Output tab of runs started on this version or later loads the results in the Apify Console: the result links point to the dataset (all results) and to its views (trending terms, trend lifecycle, keyword interest) instead of carrying query parameters the Console viewer rejects. ThetrendingTermsCsvoutput link is gone, because the Console cannot display a CSV link: add format=csv to the trending link for the same rows as CSV, or export from the Storage tab. The Actor SDK, its API client and crawlee are pinned to the versions this build was validated with. Prices are unchanged, and so are trending rows and inputs; the reliability and price comparison sections were re-measured. - 0.1.21 (2026-09-07): failure isolation. A country that times out, drops the connection or answers with something unexpected no longer ends the run: it is skipped and named, like a throttled one (before, only throttling and URL errors were isolated). Keyword groups get the same treatment, and a transport error inside a keyword cycle now rotates the IP instead of aborting. Invalid input (keywords mode without terms, a country code that is not an ISO code, an unknown category) ends as an explained empty run with nothing charged instead of a failed one. Reliability section re-measured.
- 0.1.20 (2026-08):
categoriesinput (keep only trends in the topic categories you ask for, straight from the feed's own classification; runs where the filter matches nothing succeed with a message and cost nothing); pricing table with dated competitor comparison; measured reliability section; real example input for API and MCP callers. - 0.1.19 (2026-08-13): resilience. A partially throttled request no longer fails the run (the countries that answered are delivered, the refused ones are named in the log); IP rotation now also covers session warm-up and the trending feed; README numbers re-measured on live pulls.
- 0.1.18 (2026-07-30):
onlyNewcross-run memory, per country and window, kept 7 days; trend lifecycle fieldsstarted_at,ended_at,is_active,duration_minutes; zero-row semantics split: real breakage fails the run, "nothing new" succeeds with a message. - 2026-07-29: first public release. Trending Now with estimated volume for 100+ geos and sub-regions, keyword interest over time and by region.