The same muse ai keyword can fall on Google Search while it is still rising on YouTube. One comma get_growth call prints that split as uptrend_ratio 1/3.
Live data as of 2026-09-24
muse ai can fall on Google Search in the same week it is still rising on YouTube. A comma get_growth on sources google search, youtube, google news, keyword muse ai, windows 7D, 30D, and 1Y, pulled September 24, 2026, returned sources_successful 3, uptrend_count 1, downtrend_count 2, and uptrend_ratio 1/3. YouTube closed 31 on 2026-09-19, +34.78% from 23. Google Search closed 26, -39.53% from 43. Google News closed 12, -68.42% from 38, with data_quality low. One keyword. Three sources. Opposite 7D signs.
The finding that still matters next month is the lag: search printed its high the week Meta shipped Muse, then dropped, while YouTube kept rising. Dedicated series and the comma payload agree on the closes. Google Search get_time_series returned 261 weekly points from 2021-09-25 through 2026-09-19. The last three rows are 7 on 2026-09-05, 43 on 2026-09-12, 26 on 2026-09-19. That 43 is the peak in this pull. An earlier bump reached 18 on 2026-04-11, then 11 the following week.
YouTube get_time_series on the same keyword also returned 261 points from 2021-09-25 through 2026-09-19. The last three rows are 7 on 2026-09-05, 23 on 2026-09-12, 31 on 2026-09-19. YouTube's 31 is higher than search's 26 on the same date. Indexes are not comparable across sources as absolute demand. They are comparable as shape. Search spiked and retraced. YouTube stepped up twice.
| Source | Close 2026-09-19 | 7D from 2026-09-12 | 30D from 2026-08-22 | 1Y from 2025-09-20 | Points |
|---|---|---|---|---|---|
| youtube | 31 | +34.78% from 23 | +342.86% from 7 | +520.0% from 5 | 261 |
| google search | 26 | -39.53% from 43 | +333.33% from 6 | +550.0% from 4 | 261 |
| google news | 12 | -68.42% from 38 | -66.67% from 36 | 999999.0 from 0 | 261 |
30D is still up on both search and YouTube. That is why a 30D-only dashboard looks like a single launch story. 7D is where the sources disagree. News is down on both 7D and 30D. A coverage burst can finish before video search does.
Meta published Introducing Muse on 2026-09-08. The first full weekly close after that post is 2026-09-12. Search 43 and YouTube 23 sit on that date. The next weekly close, 2026-09-19, is search 26 and YouTube 31. Headline week and video week are not the same row.
One request can ask three sources. The handler returns per-source windows plus a trend_analysis block. This pull set sources_requested to ["google search", "youtube", "google news"], sources_processed 3, sources_failed 0, uptrend_percentage 33.3. volume_weighted_growth was null. has_volume_data was false. weight_coverage printed 0/3 sources. None of these three rows attached an independent volume field in the comma payload. Do not average 31, 26, and 12 into a blended Muse score.
{"mode": "get_growth", "source": "google search, youtube, google news", "keyword": "muse ai", "percent_growth": ["7D", "30D", "1Y"]}
{"mode": "get_time_series", "source": "google search", "keyword": "muse ai"}
{"mode": "get_time_series", "source": "youtube", "keyword": "muse ai"}
MCP matches: get_growth(keyword="muse ai", source="google search, youtube, google news", percent_growth=["7D", "30D", "1Y"]). Spaces after commas matched this run. get_time_series rejected a comma list in prior pages work with invalid_source and 16 names, no all. History is one source per call. Growth is where the multi-source string belongs.
get_growth still bills one request per source plus keyword. The comma form is three credits, not one, on this trio. Extra percent_growth windows inside that call stay included. A follow-up pair of get_time_series calls is two more credits. Five requests reconstruct the table and the two charts.
Source labels are lowercase. google search is not Google Trends. youtube is not YouTube Trending. Type YouTube Trending on get_top_trends is a live video board with no keyword. The YouTube search data page is the keyword series. The Google Search data page is the search series. Muse's live iOS rank is a third tool on Muse vs ChatGPT.
News closed 12 on 2026-09-19. 7D compared that 12 with 38 on 2026-09-12, -68.42%. 30D compared 12 with 36 on 2026-08-22, -66.67%. 1Y compared 12 with 0.0 on 2025-09-20 and printed 999999.0, direction increase. The payload also set data_quality low and warned that the underlying signal is over 90% zero values. Treat this trend with caution is the API's sentence, not a gloss.
Skip 999999.0 in a forecast. Quote the 7D pair 12 versus 38. A news series that is mostly zeros will flip that sentinel on any window whose baseline lands on a quiet week. Search and YouTube on muse ai did not print the sentinel on 7D, 30D, or 1Y because those baselines were 43, 6, 4 and 23, 7, 5.
muse from meta on google search is the same class of warning. Close 23 from 31, -25.81% 7D, then 30D and 1Y at 999999.0 from 0.0, data_quality low, over 90% zeros. Do not use that phrase as the YouTube comparison key. muse ai is the string that returned clean 1Y percentages on search and YouTube.
Invalid swaps fail loudly. get_top_trends type muse ai is HTTP 400 invalid_request with 24 feed names. get_growth source Google Play is HTTP 400 invalid_source with 16 keyword sources plus all. News coverage of an app launch is source google news. The store board is type App Store Top Free. They do not substitute.
30D on search is still +333.33% from 6. 30D on YouTube is still +342.86% from 7. 1Y is +550.0% and +520.0%. Those windows stay true if the next few weekly closes settle in the 20s. The 7D disagreement is the operational alert: a news and search retrace can print while YouTube is still adding. A desk that only watches Google Search will call the story over a week early.
Free plan is 100 requests per month. Starter is $19 for 1,000. A Monday job that stores the comma 7D/30D/1Y row, plus two series snapshots when 7D signs disagree, stays inside a handful of credits. Cursor JSON for the HTTP MCP is on how to track Muse from Meta. Nightly jobs belong on REST at https://api.trendsmcp.ai/api, not on an IDE that is closed.
The 0-100 scale on each source is relative to that source's own history for that keyword. YouTube 31 next to search 26 is not "YouTube is bigger." It is "YouTube is higher in its own recent range than search is in its own, and the 7D signs disagree." Keep the ratio. Keep the dates. Keep the 1/3.
FAQ