Article counts and media tone move on different clocks. Trends MCP returns both `news volume` and `news sentiment` on the same keyword so PR and research teams can spot rising coverage with falling tone before a story hardens into a crisis narrative.
Rising article counts with falling tone are among the earliest structured warnings that a brand story is turning hostile. Social listening suites often surface the pattern after mention volume crosses an alert threshold. Trends MCP exposes the same logic through two keyword sources documented at https://www.trendsmcp.ai/docs: news volume counts how many articles mention a phrase, and news sentiment scores whether that coverage skews positive or negative on a normalized zero to one hundred scale. Both ride the same Get Trends and Get Growth tools, which keeps provenance short when an assistant drafts a triage memo.
Divergence shows up when the two normalized series move in opposite directions across a shared window. On weekly charts, news volume climbing while news sentiment slides usually means outlets are publishing faster than tone can stabilize. Crisis prevention research often treats that shape as a leading signal because volume spikes can precede mainstream escalation by forty-eight to seventy-two hours when sentiment slope turns negative at the same time.
The inverse pattern also carries meaning. news sentiment recovering while news volume fades can mark the tail of a controversy: fewer new articles, and the remaining coverage skews neutral or supportive. Both sources rising together often signals positive momentum or a neutral news cycle gaining mass attention. Both falling together usually means the topic is aging out of the press cluster.
These reads describe media pipeline behavior on a fixed keyword. They do not replace legal review, customer support queues, or social mention text. Pair divergence charts with Reddit discussion growth or Google Search demand when public lookup matters; the lead-lag workflow on https://www.trendsmcp.ai/news-volume-google-search-lead-lag-mcp covers that pairing in depth.
Run Get Growth twice when tables must stay source-pure, or once when a combined response is enough. A crisis desk checking a corporate name after a product recall might open with short windows before widening to quarterly context.
{
"source": "news volume",
"keyword": "Acme Robotics",
"percent_growth": ["7D", "30D", "3M"]
}
{
"source": "news sentiment",
"keyword": "Acme Robotics",
"percent_growth": ["7D", "30D", "3M"]
}
A combined pull returns both result arrays in one response:
{
"source": "news volume, news sentiment",
"keyword": "Acme Robotics",
"percent_growth": ["7D", "30D", "3M"]
}
Each response row includes period, growth as a signed percentage, direction (increase or decrease), recent_date, baseline_date, recent_value, and baseline_value on the zero to one hundred scale. When volume_available is true on the news volume rows, recent_volume and baseline_volume expose raw article counts alongside normalized scores.
For charts, call Get Trends once per source. Weekly mode returns roughly five years of {date, value, volume} points. REST callers can request data_mode: "daily" on supported pipelines for the last thirty days; MCP hosts omit data_mode and receive weekly series by default.
Event windows anchor better with custom objects inside percent_growth:
{
"source": "news sentiment",
"keyword": "Acme Robotics",
"percent_growth": [
{
"name": "post-recall statement",
"recent": "2026-06-13",
"baseline": "2026-05-01"
}
]
}
Same-day headline lists belong in Get Top Trends with type set to Google News Top News and limit between ten and twenty-five. That feed shows what leads the cycle now without naming a keyword first.
Example MCP phrasing that routes correctly: "Using TrendsMCP, compare 7D and 30D growth on news volume and news sentiment for Acme Robotics and label any divergence quadrant." REST callers POST the same JSON bodies to https://api.trendsmcp.ai/api with Bearer authentication.
The matrix below maps common weekly shapes into PR and investor relations actions. Thresholds are heuristics for triage, not automatic escalation rules. Always name the keyword, preset window, and pull date in the brief.
30D news volume |
30D news sentiment |
Typical read | Suggested action |
|---|---|---|---|
| Increase above ~20% | Decrease below ~10% | Hostile coverage stacking | Open comms war room; draft holding statement; monitor 7D slope daily |
| Increase above ~20% | Increase above ~10% | Positive or neutral pile-on | Amplify favorable coverage; prepare FAQ for search follow-through |
| Decrease below ~10% | Decrease below ~10% | Story fading with sour tone | Track residual articles; avoid premature victory laps |
| Decrease below ~10% | Increase above ~10% | Recovery phase | Publish corrective proof points; shift to proactive outreach |
A practical Monday screen many brand teams run: flag any watchlist name where 30D growth on news volume exceeds twenty-five percent while 30D growth on news sentiment stays below minus ten percent. Confirm with two consecutive weekly Get Trends points so one noisy syndication week does not trigger a full response.
On a Trends MCP pull dated June 13, 2026, news volume for the keyword Tesla showed divergent windows across presets on the same entity. The 30D window posted a 27.9% decrease with recent_value 16.3 against a May 16 baseline of 22.6, while the 3M window posted a 33.6% increase against a March 14 baseline of 12.2. The short window cooled as the longer window still reflected an elevated quarter. A sentiment pull on the same keyword and dates would complete the divergence read; teams should run both sources before inferring tone from volume alone.
That pattern illustrates why multi-preset tables matter. A single 30D row can understate a quarter-long build, and a single 3M row can miss a recent cooldown. Crisis desks compare at least two horizons before locking narrative language.
Communications leads use divergence to separate "more articles" from "worse articles." Investor relations teams pair the matrix with earnings watchlists so adverse tone does not hide inside steady mention counts. Equity researchers treat sustained volume growth with sentiment decay on a policy keyword as context for later search and commerce signals elsewhere in the stack.
Teams already running the PR workflow on https://www.trendsmcp.ai/pr-crisis-trend-monitoring can add explicit divergence scoring to weekly standups. Broader brand health monitoring on https://www.trendsmcp.ai/brand-monitoring benefits from the same JSON when tone drift must be separated from raw mention spikes.
Source field tours that shorten onboarding live on https://www.trendsmcp.ai/news-sentiment-data for tone scoring and on https://www.trendsmcp.ai/news-volume-data for mention counts. When coverage runs ahead of public lookup, extend the review with the Google Search lead-lag guide on https://www.trendsmcp.ai/news-volume-google-search-lead-lag-mcp.
Trends MCP normalizes both sources to zero one hundred so curves share one chart. That aids comparison and hides raw unit differences between article counts and NLP-derived tone scores. Weekly aggregation smooths intraday shock stories; desks covering breaking wires should pair history with Get Top Trends rather than inferring hour-level moves from Monday snapshots alone.
Sentiment scores summarize headline and summary tone across aggregated English-language outlets. They can miss sarcasm, legal boilerplate, or mixed-audience framing inside one article cluster. Sparse entities return thin series or documented not_found responses. Upstream gaps surface as explicit error codes in the API reference.
Keyword mismatch is the most common false divergence. A formal legal name on news volume paired with a colloquialism on news sentiment can fabricate a gap that disappears when both calls use identical strings. Any conclusion should cite both source values, the keyword string, preset windows, and whether the pattern repeated across more than one cycle.
Free-tier budgets count one source per Get Trends or Get Growth call. A twenty-name watchlist reviewed across both sources therefore needs forty growth calls before any live feed pulls. Batch weekly reviews and reserve daily 7D checks for names that trip the twenty-five over minus-ten screen above.
FAQ