News sentiment and volume data for AI

Trends MCP returns news sentiment as a positive/negative tone series and news volume as a coverage-quantity series. Both sit on a 0-100 index with about 5 years of weekly history. Polarity is computed from headlines and summaries, then rescaled from a -1 to +1 score so it matches other sources. The keyword is a topic, company, or common ticker, not an outlet name. The same MCP connection covers 25+ sources, including Reddit and Wikipedia. The free plan includes 100 requests per month. Starter is $19/month.

Volume and sentiment are different jobs

News volume answers "how loud is this." A product launch, a crash, and a celebrity wedding can all raise volume. News sentiment answers "which way does the copy lean." High volume plus falling sentiment is the classic crisis shape. High volume plus rising sentiment is the classic hit shape. Low volume makes sentiment noisy because a handful of wires can swing the average.

Calling both sources in one get_growth request is the intended pattern: source='news sentiment, news volume'. Reading only sentiment hides a story that everyone stopped covering. Reading only volume hides a story that everyone covered angrily.

get_trends on news sentiment draws the tone history. get_top_trends with type Google News RSS needs no keyword; it is the live headline feed for "what is on the wire now."

Keyword choices that resolve

Company names and tickers (tesla, NVDA) work when they dominate English coverage. Ambiguous words (apple, meta, target) need a more specific phrase or a confirmation call on Wikipedia pageviews. People need the name the wires use, including a last name when the first name is common.

This source is not outlet filtering. source='news sentiment' plus keyword New York Times will try to score coverage of that newspaper as a topic, not restrict the corpus to that newspaper. News volume data is the sibling series when quantity is the only question.

Languages outside the English aggregation will under-read. Local political stories that never hit national wires may be invisible even when they dominate a city.

Native news APIs versus a polarity index

Google News RSS, publisher APIs, and GDELT-style event databases return headlines, URLs, and sometimes themes. Few of them hand an assistant a five-year 0-100 tone curve aligned with Amazon and Google Search. Building that still means NLP, storage, and a second set of API keys.

Trends MCP does not replace media monitoring suites that alert on brand mentions, clip TV, or count share of voice by outlet. Meltwater, Brandwatch, and similar platforms are built for comms teams. Teams should check current vendor pricing for those contracts. Trends MCP supplies topic-level tone and volume for research assistants, not a newsroom command center.

The official X, Meta, and Reddit APIs are social, not newsroom, objects. Social volume belongs on those sources. News sentiment is editorial copy.

Who uses news polarity series

Investor relations and equity research watch tone into earnings, then check whether Google Trends moved with the press or against it. Public-affairs teams watch regulation keywords. Product marketers watch launch week to see if reviews stayed in the product section or migrated to the scandal section. Academics treat the series as a coarse media-attention index, not as a substitute for hand-coded content analysis.

Customer-support tone and app-review stars are different corpora. Those should not be mixed into this source.

Earnings week is the usual stress test. Volume rises for almost every liquid ticker. Sentiment tells whether the copy clustered on beat-and-raise language or on guidance cuts, lawsuits, and executive exits. A print that is "mixed" in the newspapers can still show as a small index move if positive and negative wires cancel. That cancellation is information: the story is contested, not ignored. Daily mode for the print window, then weekly mode for the quarter, keeps both facts.

Political and celebrity keywords swing harder than industrial issuers because the corpus is larger and more opinionated. A 20-point drop on a quiet manufacturer is a bigger operational alert than a 20-point drop on a polarizing public figure.

Limits and caveats

Headline NLP misses sarcasm, coded language, and mixed features that are positive on growth and negative on governance in the same article. Wire duplication can overweight a single press release. Paywalled body text is not fully visible to a headline model. The 0-100 index is relative to that keyword's window; a quiet company at 90 is not "more beloved" than a famous one at 50 during a rough month.

Daily mode exists for 30-day windows around a known event. Weekly mode is the default for the 5-year archive.

Quota is 100 requests per month on the free plan, $19/month on Starter. Full-text clipping of every URL is a media-monitoring job.

Comparison: media tone data

Trends MCP is one MCP connection covering 25+ sources. News sentiment and news volume share the 0-100 index and long weekly history with search, Wikipedia, and Reddit. Google News RSS is available through get_top_trends without a keyword.

Publisher APIs and RSS remain the path for full articles and outlet-specific apps.

Enterprise media intelligence platforms add mention alerts, influencer graphs, and TV. Teams should check current vendor pricing. Trends MCP is the assistant-native polarity and volume series, not a replacement for a PR operations suite.