Insights endpoints
Insights endpoints query the same corpus as the article endpoints, but they return an answer instead of a feed: a ranking, a number, or a verdict. Use them when downloading articles would be the wrong shape of response — for dashboards, sample-size estimation, monitoring, and verification.
Which one do I need?
| You want | Endpoint | Cost | Guide |
|---|---|---|---|
| A ranking of who or what is being talked about | GET/POST /v1/news/trends | 1 point | Trends |
| Only the number of matching articles | GET /v1/news/count | 1 point | Count articles |
| A verdict on a factual claim, with evidence | POST /v1/fact-check | 5 points | Fact check |
Trends — aggregate instead of list
/v1/news/trends takes a field and returns its most frequent values with counts and shares, optionally with growth rate, a trending score and a comparison against a previous period. It answers "who is being talked about right now" without downloading a single article:
curl -H "X-API-Key: YOUR_API_KEY" \
"https://api.apitube.io/v1/news/trends?field=entity.id&language.code=en"Every filter from /v1/news/everything applies, so the aggregation runs over exactly the slice you care about — one industry, one country, one date range.
Count — the total without the payload
/v1/news/count accepts the same filters but runs a single count(*), which is much cheaper and faster than paging through results. Parameters that shape a result set (page, per_page, sort.*, export, fl) have no effect on it.
It is the right call before a large export, when sizing a sample, or when a dashboard needs a number rather than a list.
Fact check — a verdict grounded in articles
/v1/fact-check verifies claims against the live corpus: it retrieves recent articles matching each claim and grounds the verdict in them, rather than relying on a language model's memory. For every claim it returns a verdict on an 8-level scale, a calibrated confidence score, a short explanation, and the supporting or refuting evidence articles. You can check a single statement, a full article's text, or an article already in the corpus by its id.
Paid feature
Fact Check is not available on the Free plan. Each request costs 5 points and the endpoint has its own rate limit of 10 requests per minute, stricter than the global one, because every request runs language-model inference.
Next steps
- Turn facets into a dashboard: Build a faceted news dashboard
- Watch one company's coverage over time: Monitor a company
- Check what a request costs you: Balance and usage