The AskNews platform Historical News Search
Put past events in context
Search archived reporting, including licensed news sources, for the period that matters to your research. Bring dated articles into historical analysis, forecasting research and AI evaluations.
The news archive, with explicit time boundaries.
Point-in-time retrieval
Visible to the run
Reporting published on or before your cut-off.
Withheld
Everything after it, including the outcome.
end_timestamp — the evaluation cut-off
Research a period, not just a topic
AskNews Historical News Search brings archived reporting into the same retrieval workflow as current news. Choose a question and a date range to investigate how an event developed, compare coverage or assemble context for a historical evaluation.
Backtest without time leakage
AskNews can return only the reporting that existed at a chosen point in time — what customers call the time machine. If future information leaks into a historical test, the backtest is corrupted and the model looks better than it is. Bounding retrieval by date is what keeps an evaluation honest.
Set the boundary explicitly
Choose start and end timestamps, and the date field you filter on: pub_date filters by publication date, crawl_date by when AskNews ingested the article and is the default. The two answer different questions.
Why forecasting teams run on AskNews
- Point-in-time retrieval lets a backtest see only what was knowable then, which is the difference between a real result and a leaked one.
- The archive reaches back to 2023 across licensed and paywalled sources, so a historical question is not limited to whatever is still free on the open web.
- Dates and source links stay attached to every record, so a published result can be re-checked against the reporting it used.
What comes back
- Archived news context as prompt-ready text, structured article records, or both.
- Source URLs and publication dates for checking the reporting behind your analysis.
- Available article enrichments, including summaries, entities, categories and sentiment.
Proof
Forecasters pick AskNews when the result is scored
Metaculus runs quarterly AI forecasting tournaments where bots are scored against real outcomes — the setting where leaked future data shows up as a result nobody can reproduce.
The majority of winners of the Q3 forecasting tournament used AskNews. That's why our staff converted all of our in-house bots from Perplexity to AskNews for the Q4 tournament.
MCP, CLI and SDK quickstart
Query the archive through News Search
The archive is the same News Search interface with a time boundary: pass start_timestamp and end_timestamp, and choose time_filter deliberately. The window itself selects the archive. Run it from the CLI, the SDKs or the API.
Build vs buy
The pipeline is never just a pipeline
Teams that build their own news ingestion budget for a crawler and a parser. What arrives is bot detection, paywalls they cannot lawfully pass, per-publisher markup that changes without notice, deduplication, language coverage, and managing an archive for low-latency retrieval. It is a product, and it competes for the engineers you hired to build your own.
Read: Why your team shouldn't build its own news ingestion pipelineBuild
- Crawler + bot detection
- Per-publisher parsers
- Dedupe + language coverage
- Archive consistency
- Ongoing maintenance
Buy
ask.news.search_news(
query="...",
return_type="string",
)Licensed sources, enriched and prompt-ready, with the links and dates already attached.
Ground your next evaluation in the archive. Ground your next backtest in the archive. Ground your next forecast in the archive. Ground your next simulation in the archive
Start with a defined period and inspect the records before drawing conclusions.