Custom date growth windows on get_growth

A custom date growth call on Trends MCP passes {name, recent, baseline} objects inside percent_growth and returns the snapped dates the series actually holds. On the August 24, 2026 pull, bitcoin on google search printed index 33.0 on 2026-08-22 against 59.0 on 2025-08-23, a named YoY of -44.07% that matched the 12M preset because those two dates are the same pair. A Since June 1 window requested baseline 2026-06-01 and snapped to 2026-05-30, where bitcoin sat flat at 33.0. steam Elden Ring requested recent 2026-08-01 and snapped to 2026-07-01 (index 1.9, peak players 46,694). The free plan includes 100 requests per month.

What custom date growth data shows right now

From the Trends MCP pull of August 24, 2026, named custom windows sat beside preset 30D, YTD, and 12M rows in the same get_growth response, and the print that matters is where recent_requested diverges from recent_date. google search weekly points used 2026-08-22 when that day was requested. steam monthly points would not.

Live Google Trends board from the same session, type Google Trends, as_of 2026-08-24T20:01:36 UTC, count 25, used as the keyword pool for several custom windows:

Ranks 16 through 25 were xrp price, fortnite admin panel codes, supreme, insidious out of the further, lunar eclipse august 2026, boy left on mount fuji, prince harry, aircraft pilot, state farm dividend, and publix blueberry recall.

Custom versus preset prints, google search unless noted, recent_date 2026-08-22 and 261 weekly points except where a source calendar is named:

youtube bitcoin used the same 2026-08-22 recent_date: 30D +32.14% (37.0 from 28.0), YoY -26.0% (37.0 from 50.0 on 2025-08-23), August open +37.04% (37.0 from 27.0 on 2026-08-01). For search-only presets without custom objects, see the Google Trends API.

Exact custom date get_growth calls

One get_growth body can mix preset strings and {name, recent, baseline} objects, and the response keeps recent_requested / baseline_requested on custom rows so a snap to 2026-05-30 is visible next to the June 1 that was asked for in the same JSON payload. Extra windows still ride on one source-plus-keyword unit.

{
  "mode": "get_growth",
  "source": "google search",
  "keyword": "bitcoin",
  "percent_growth": [
    "30D",
    "3M",
    "YTD",
    "12M",
    {"name": "YoY", "recent": "2026-08-22", "baseline": "2025-08-23"},
    {"name": "August open", "recent": "2026-08-22", "baseline": "2026-08-01"},
    {"name": "Since June 1", "recent": "2026-08-22", "baseline": "2026-06-01"}
  ]
}

{
  "mode": "get_growth",
  "source": "steam",
  "keyword": "Elden Ring",
  "percent_growth": [
    "30D",
    "12M",
    {"name": "YoY monthly", "recent": "2026-08-01", "baseline": "2025-08-01"},
    {"name": "Since June 1", "recent": "2026-08-01", "baseline": "2026-06-01"}
  ]
}

{
  "mode": "get_growth",
  "source": "wikipedia",
  "keyword": "ChatGPT",
  "percent_growth": [
    "30D",
    "3M",
    {"name": "July 1", "recent": "2026-08-01", "baseline": "2026-07-01"},
    {"name": "June to July", "recent": "2026-07-01", "baseline": "2026-06-01"}
  ]
}

Dates are YYYY-MM-DD. The name field is optional and echoes back on the result row. Extra windows still ride on one source-plus-keyword unit. For comma-separated sources instead of custom dates, see the multi-source growth API.

Why a named YoY can match 12M or miss it

A named YoY window is only a 12M duplicate when the requested dates already sit on the preset pair, which is what happened for bitcoin, fortnite, chatgpt, and GTA 6 on google search in this pull: recent 2026-08-22 versus baseline 2025-08-23, the same two dates 12M selected on its own.

bitcoin makes the trap obvious. 12M and named YoY both printed -44.07% (33.0 from 59.0). August open printed +13.79% (33.0 from 29.0 on 2026-08-01). 30D printed +6.45% (33.0 from 31.0 on 2026-07-25). Since June 1 printed 0.0% after snapping baseline to 2026-05-30. Four true answers, four windows, one index of 33.0. fortnite 12M / YoY was +75.0% (35.0 from 20.0) while August open was +34.62% (35.0 from 26.0) and Since June 1 was +118.75% (35.0 from 16.0). chatgpt 12M / YoY was -10.23% (79.0 from 88.0) while YTD was +27.42% (79.0 from 62.0 on 2026-01-03) and 30D was flat at 79.0 versus 79.0.

steam does not keep that coincidence. Elden Ring 12M used recent 2026-07-01 / 46,694 peak players versus 2025-07-01 / 61,614, index 1.9 from 3.6, -47.22% index and -24.22% volume. The named YoY monthly window requested 2026-08-01 versus 2025-08-01. recent snapped to 2026-07-01. baseline stayed 2025-08-01 (index 2.1, volume 50,646). The print was -9.52% index and -7.8% volume. That is July 2026 versus August 2025, not August versus August. Quote both requested and used dates. For live Steam ranks rather than monthly growth, see the Steam Most Played API.

Board spikes that presets flatten

Preset 30D can hide an intra-month move that a custom August open window catches, which is why flock cameras on google search rose 112.77% over 30D (100.0 from 47.0 on 2026-07-25) and only 69.49% from August 1 (100.0 from 59.0), two different stories about when the spike actually started.

insidious sat at Google Trends rank 5 with google search index 61.0 on 2026-08-22. 30D was +916.67% from 6.0 on 2026-07-25. August open was +662.5% from 8.0 on 2026-08-01. YTD was +1933.33% from 3.0 on 2026-01-03. YoY was +1120.0% from 5.0 on 2025-08-23. The title is a release spike, so the custom August open window is the one that matches "since the month began" rather than "versus last July." moderna (board rank 15 as moderna stock) rose 280.0% over 30D (38.0 from 10.0) and 245.45% from August 1 (38.0 from 11.0). Estimated volumes 489,000 versus 129,000 are derived from the index; skip volume_growth. supreme (board rank 18) fell 12.07% over 30D (51.0 from 58.0) and 10.53% from August 1 (51.0 from 57.0), with YTD -47.96% from 98.0 on 2026-01-03.

lunar eclipse (board rank 20) printed index 8.0 with 30D, 3M, and YTD all at +700.0% from baseline 1.0. YoY was +300.0% from 2.0. Small baselines inflate percents. GTA 6 YTD and August open both printed +209.09% because YTD baseline 2026-01-03 and August 1 baseline both sat at 11.0 against recent 34.0. 12M / YoY was +580.0% from 5.0. For the live Google Trends board without custom windows, see the Google Trends Trending Now API.

Limits of custom dates on short calendars

Custom windows fail or go lopsided when the series is monthly and the requested recent date sits past the last point, which is why wikipedia ChatGPT with recent 2026-08-01 and baseline 2026-07-01 returned collapsed_range on a 44-point series whose latest row is 2026-07-01.

The wikipedia 30D preset on the same call used 2026-07-01 versus 2026-06-01, index 47.5 / 6,574 page views versus 33.5 / 4,632, +41.79% index and +41.93% volume. A named June to July window that requested those two dates succeeded and matched 30D exactly. 3M used 2026-04-01 at 34.4 / 4,760 (+38.08%). Asking for August versus July on that series is the error. steam Elden Ring Since June 1 requested recent 2026-08-01, snapped recent to 2026-07-01, kept baseline 2026-06-01, and matched the 30D preset: index 1.9 / 46,694 versus 2.4 / 49,524, -20.83% index and -5.71% volume, 53 monthly points.

Weekly google search and youtube series in this pull had 261 points and accepted 2026-08-22 as recent. They still snapped June 1 to May 30. Always publish recent_date and baseline_date, not only the name. Zero-baseline and low-quality warnings still apply on custom rows the same way they apply on presets. For Wikipedia page-view growth without custom objects, see the Wikipedia pageviews growth API.

Pricing for custom date growth

Free covers 100 requests per month across every source as of August 2026, and extra named windows inside one get_growth call do not add units, so a bitcoin mix of 30D, YTD, 12M, YoY, and August open is still one request. Starter is $19 for 1,000 requests, Pro $49 for 5,000, Business $199 for 25,000. Annual billing saves 20%. Check current pricing before locking a budget. Confirm the billed count on the account page after a mixed preset-plus-custom test rather than guessing from calculations_completed.