Pass source google search for a weekly Google Search interest series, or source youtube for a weekly YouTube search-interest series. Same grain, same close date, different peaks and often opposite 12M signs.
Live data as of 2026-10-08
Source google search on get_growth and get_time_series returns a weekly 0-100 Google Search interest series for one phrase. Source youtube returns a weekly 0-100 YouTube search-interest series for the same string. Both pulls on October 8, 2026 used 261 weekly points from 2021-10-09 through 2026-10-03. chatgpt closed 71.0 on Google Search and 54.0 on YouTube. The 12M window moved in opposite directions: Google Search minus 22.83 percent from 92.0, YouTube plus 157.14 percent from 21.0. Omitting percent_growth defaults to 12M on both. The live boards use type Google Trends and type YouTube Trending, not those source strings.
The October 8, 2026 pull is the split that still matters next month: the two sources share weekly grain and the same close date, then disagree on direction. get_time_series for chatgpt on source google search returned 261 weekly points from 2021-10-09 (value 0) through 2026-10-03 (value 71). First non-zero week was 2022-12-10 at 4. Peak 100 sat on 2025-09-20. No row carried a volume key. The series field is source google search.
get_time_series for chatgpt on source youtube returned the same 261 weekly dates. Open 0 on 2021-10-09. First non-zero week was also 2022-12-10, at 5. Close 54 on 2026-10-03. Peak 100 sat earlier, on 2025-04-05. No volume key. The series field is source youtube.
Growth windows on that keyword line up by calendar and split by sign. Google Search 7D matched 1W: 71.0 from 66.0, plus 7.58 percent, baseline 2026-09-26. 14D matched 2W: 71.0 from 67.0, plus 5.97 percent. 30D matched 1M: 71.0 from 86.0, minus 17.44 percent. 3M was minus 6.58 percent from 76.0. YTD was plus 18.33 percent from 60.0 on 2026-01-03. YouTube 7D matched 1W at 0.0 percent, 54.0 from 54.0, direction flat. 14D matched 2W at 0.0 percent from 54.0. 30D matched 1M: 54.0 from 44.0, plus 22.73 percent. 3M was plus 45.95 percent from 37.0. YTD was plus 184.21 percent from 19.0. Same dates. Opposite year-scale story.
| Keyword | Google close | Google 7D | Google 12M | YouTube close | YouTube 7D | YouTube 12M | Points |
|---|---|---|---|---|---|---|---|
| chatgpt | 71.0 on 2026-10-03 | +7.58% from 66.0 | -22.83% from 92.0 | 54.0 on 2026-10-03 | 0.0% from 54.0 | +157.14% from 21.0 | 261 / 261 |
| nfl | 44.0 on 2026-10-03 | +15.79% from 38.0 | -16.98% from 53.0 | 77.0 on 2026-10-03 | +1.32% from 76.0 | +10.0% from 70.0 | 261 / 261 |
| tesla | 41.0 on 2026-10-03 | -16.33% from 49.0 | -25.45% from 55.0 | 64.0 on 2026-10-03 | +16.36% from 55.0 | +20.75% from 53.0 | 261 / 261 |
| nike | 54.0 on 2026-10-03 | -15.62% from 64.0 | -8.47% from 59.0 | 49.0 on 2026-10-03 | +11.36% from 44.0 | +8.89% from 45.0 | 261 / 261 |
nfl rose on both 7D windows and still disagreed on 12M. Google Search 12M was minus 16.98 percent. YouTube 12M was plus 10.0 percent. tesla 7D was minus 16.33 percent on Google Search and plus 16.36 percent on YouTube. Quote both indexes. Do not average them.
nfl peaks are not the same week. Google Search hit 100 on 2024-01-20 at series volume 805194805. YouTube hit 100 on 2023-09-16 and again on 2025-09-13, with no volume on any row. Opening rows also differ: Google Search nfl opened 43 / 346233766 on 2021-10-09. YouTube opened 66.
Source names case-fold. Google Search and YouTube on get_growth for chatgpt returned the same 12M rows as the lowercase labels. Feed labels do not belong here. Source Google Trends returned HTTP 400 invalid_source, message 'google trends' is not a valid source., and listed 16 keyword sources plus all. Source YouTube Trending did the same, echoed as 'youtube trending'. The handler lowercased the rejected type string. It did not map it to source google search or source youtube.
Google Search volume is present on some keywords and missing on others. When it is present, get_growth flags it. The nfl 7D row set volume_available true, volume_estimated true, recent_volume 354000000.0, and volume_growth null. The omitted-reason string was volume is derived from the trend value for this source, not an independent measurement. The matching get_time_series close was 354285714, not 354000000.0. Growth rounded the series volume. Quote the series row when the exact integer matters.
YouTube nfl 7D had no volume fields at all. Index 77.0 from 76.0, plus 1.32 percent, 261 points. tesla Google Search printed estimated volume 6610000.0. YouTube tesla printed none. nike Google Search printed 19800000.0. YouTube nike printed none. chatgpt printed volume on neither source.
{"mode": "get_growth", "source": "google search, youtube", "keyword": "chatgpt", "percent_growth": ["7D", "12M"]}
That comma call on October 8, 2026 returned sources_successful 2. trend_analysis printed uptrend_count 1, flat_count 1, and uptrend_ratio 1/2. volume_weighted_growth was null with has_volume_data false, sources_with_volume 0, weight_coverage 0/2 sources, and total_volume 0. Neither side contributed a weight. The 7D Google Search plus 7.58 percent and the YouTube flat 0.0 percent stayed as separate rows.
nfl on google search, youtube with 7D and 12M did print a blend, and that blend is a trap. sources_successful 2. uptrend_ratio 2/2. volume_weighted_growth was 15.79 with weight_coverage 1/2 sources and total_volume 354000000.0. The only weighted row was Google Search, at 100 percent of that volume, using Google Search 7D index growth 15.79. YouTube 7D plus 1.32 percent never entered. Google Search volume_growth was null, so the mixer reused the index. Read volume_estimated and weight_coverage before charting volume_weighted_growth as a two-platform number.
get_time_series rejects the comma. Source google search, youtube returned HTTP 400 invalid_source and listed 16 keyword sources with no all. One source per series call. Two sources belong on get-growth.
Keyword format is the second durable split. Both sources treat the string as a search phrase. Neither strips a leading hash. On October 8, 2026, #chatgpt on Google Search closed 25.0 on 2026-10-03, 7D plus 78.57 percent from 14.0, 12M minus 28.57 percent from 35.0, 261 points. Plain chatgpt closed 71.0, 7D plus 7.58 percent, 12M minus 22.83 percent. Those are two Google Search series.
#chatgpt on YouTube closed 26.0, 7D minus 50.0 percent from 52.0, 12M plus 188.89 percent from 9.0, 261 points. Plain chatgpt closed 54.0, 7D 0.0 percent, 12M plus 157.14 percent. Two YouTube series. If the intent is the phrase people type without a hash, send chatgpt. If the intent is the hashed spelling, send #chatgpt and expect a different curve.
Case is not a second series. Keyword ChatGPT on Google Search echoed search_term ChatGPT and matched the lowercase 12M row: 71.0 from 92.0, minus 22.83 percent, 261 points. Keyword ChatGPT on YouTube matched 54.0 from 21.0, plus 157.14 percent. Store the lowercase string in configs. Do not treat title case as a join key.
Two strings, two tools. REST uses mode get_growth or get_top_trends. MCP uses the matching tool names. Google Search history lives on the Google search data page. YouTube search history lives on YouTube search data. Keyword Google Search versus the live Google Trends board is Google Search vs Google Trends.
{"mode": "get_growth", "source": "google search, youtube", "keyword": "chatgpt", "percent_growth": ["7D", "12M"]}
{"mode": "get_time_series", "source": "google search", "keyword": "chatgpt"}
{"mode": "get_time_series", "source": "youtube", "keyword": "chatgpt"}
{"mode": "get_top_trends", "type": "Google Trends"}
{"mode": "get_top_trends", "type": "YouTube Trending"}
The MCP equivalents are get_growth(keyword="chatgpt", source="google search, youtube", percent_growth=["7D", "12M"]), get_time_series(keyword="chatgpt", source="google search"), get_time_series(keyword="chatgpt", source="youtube"), get_top_trends(type="Google Trends"), and get_top_trends(type="YouTube Trending"). REST 1W matched 7D on Google Search chatgpt. REST 1M matched 30D. The same aliases matched on YouTube.
Omitting limit on type Google Trends returned limit 25, count 25, offset 0, and as_of_ts 2026-10-08T06:01:39.639196+00:00. Rank 1 was gypsy rose. Rank 2 was hurricane isaias. Rank 3 was russian plague. Those rows are search phrases. They are not YouTube titles. Omitting limit on type YouTube Trending returned the same stamp, limit 25, count 25, offset 0. Rank 1 was Lil Baby, Bally Baby, Fatt Smaxk - Coast 2 Coast. Rank is live video order, not a sort of the 2026-10-03 YouTube index.
REST type google search (the keyword source string, lowercase) returned HTTP 200, echoed type Google Trends, and matched gypsy rose at rank 1. REST type youtube trending echoed YouTube Trending and matched the Lil Baby title. llms.txt still describes feed labels as case-sensitive. The live handler accepted those lowercase strings on October 8, 2026. Prefer the canonical spelling in stored configs.
get_top_trends with type youtube returned HTTP 400, error invalid_request, message Unknown type 'youtube', plus 24 valid feed labels that include Google Trends and YouTube Trending. Source youtube is not a feed type. Type Google Trends is not a growth source. Type google search is the trap: it aliases the live board instead of erroring.
country GB with limit 3 on Google Trends returned HTTP 200, echoed country GB, and stamped as_of_ts 2026-10-08T06:12:50.955969+00:00. Rank 1 was richard e grant. Rank 2 was weather tomorrow. Rank 3 was world space week 2026. That stamp is later than the default board's 06:01:39 snapshot, and rank 1 is a different phrase. The same country GB call on YouTube Trending returned HTTP 200, stamped as_of_ts 2026-10-08T06:12:52.725238+00:00, with a different rank 1 than the omitted-country board. Country is supported on both feeds. Omit country for the default board. Do not join a GB list to the omitted-country snapshot as one chart.
REST omitting type is a different 400. POST https://api.trendsmcp.ai/api with mode get_top_trends and limit 5, no type, returned HTTP 200 with envelope statusCode 400, error invalid_request, message The 'type' parameter is required, then the same 24-name list. llms.txt still names that miss missing_parameter. Read statusCode inside the JSON, not only the HTTP line.
get_top_trends MCP pricing is 1 credit per 10 requested rows, max 10 credits per call. REST counts 1 request per call. Free boards still cap at the top 10. The 25-row pulls on this page are the paid default, which is also what the handler returns when limit is omitted on a paid key. Keyword follow-up stays on source google search or source youtube. The live boards stay on their type strings.
FAQ