A recurring news export has an annoying failure mode: it works, but keeps handing you the same headlines.
For CleanScrape's news Actor, I wanted a small watchlist to answer two separate questions: which available articles match my topic, and which of those have I not received before?
I maintain Google News and Publisher Feed Scraper. It is a paid Actor that collects Google News searches and public publisher feeds into an article table. It does not generate summaries or read through paywalls.
Get a useful first export before adding a schedule
Open the electric-vehicle example. It looks for recent electric-vehicle news, with a maximum of 20 articles.
In the form, choose the Google News edition and Published within from the dropdowns. You do not need to remember a country code or type dates in an exact format.
For an API call, the same small starting point is:
{
"queries": ["electric vehicles"],
"edition": "US:en",
"timeRange": "week",
"maxArticles": 20,
"resolveUrls": true
}
Open Articles to inspect the headlines, publishers, dates and links. Any summary text is an available feed snippet, not the full article body.
Narrow the topic without a complicated search expression
For a more focused watch, you could keep headlines containing recall or battery fire and exclude opinion:
{
"queries": ["electric vehicles"],
"includeTerms": ["recall", "battery fire"],
"excludeTerms": ["opinion"],
"includeMode": "any",
"filterScope": "headline",
"timeRange": "week",
"maxArticles": 20
}
These are literal word or phrase filters, not an AI judgement about whether an article is important. They do not search full article bodies. No result can simply mean that none of the available entries matched.
The Actor can also read public RSS, Atom and JSON feeds. There is a separate BBC Technology feed example if you want to start with a known publisher instead of Google News.
One distinction is easy to miss: a Google search query does not filter a publisher feed added to the same run. The word filters apply across sources; the query itself does not.
Then stop exporting the same articles every time
Enable Return only previously undelivered articles and enter a stable Watch name. For the original broad search, that looks like:
{
"queries": ["electric vehicles"],
"edition": "US:en",
"timeRange": "week",
"maxArticles": 20,
"newOnly": true,
"watchName": "EV industry watch"
}
The first run exports eligible entries. Later runs using the same watch skip article identities already delivered to it. A repeat run with no new output can be doing exactly what you asked.
Keep the watch name, sources and filters consistent. Use a new watch name when changing the scope. Delivery history is retained for up to 90 days, subject to storage availability; this is not a permanent archive or a system for tracking edits inside articles.
To run automatically, save the configuration as an Apify task and add a schedule. Turning on the watch option does not create that schedule, send an email or post to Slack. Those are separate workflow steps.
Two things I would check before using the export
First, read Run report. Feed limits, date filtering and source errors can all reduce the output. Twenty is a maximum, not a guaranteed count. If each publisher must contribute, separate runs can be better than letting the first source fill a shared cap.
Second, check link status. Publisher-link lookup is best effort. If it cannot resolve a direct URL, the row retains its Google News link and reports that status. It is still a delivered, billable article. This is more useful than silently dropping the headline, but it matters if your next step requires direct publisher URLs.
At the current base price, 20 delivered articles cost $0.019, with no startup fee. The examples have a $0.05 spending cap. Previously delivered watch entries and filtered-out entries do not generate article charges.
The intended result is a manageable reading queue for your own research or reporting workflow. It is not a complete news archive, a verified factual briefing or a substitute for reading the source.
CleanScrape is independent of Google and the publishers. Public feed availability does not grant unrestricted republication rights. Questions or reproducible issues are welcome in the Actor's Issues tab or at contact.cleanscrape@gmail.com.
Make a reading brief from the saved articles
The Google News and feed research skill covers what happens after export. It helps a compatible assistant organise the available headlines and snippets into a sourced reading list, without presenting feed text as a full article or a verified account of events.
Use this article export only. Group the main topics, keep publisher links and dates, and flag unresolved links or missing context. Do not fetch full articles or start another run.
The instructions are free and do not create a schedule. Fresh collection and any assistant usage have their own costs. Keep the underlying sources handy when a detail matters.
Top comments (0)