Search for real-time news
On this page
News at your fingertips
A language model left to its training data knows nothing about events that occurred after that data was collected. Web search removes this limitation: ChatGPT can inform you about an event that happened today because it reads the pages discussing it at the moment you ask. This lesson shows you how to turn this capability into a usable daily monitoring routine, rather than a one-off curiosity.
The mechanism is simple to describe. ChatGPT detects that a question concerns current events, launches searches across news sites, press agencies, and specialized sources, synthesizes the most recent and relevant information, and then cites the selected sources. Their date is the piece of information most often neglected, yet it decides everything, and ChatGPT does not always specify it: open the source to verify it. On a story developing by the hour, a source from yesterday and a source from this morning do not tell the same story.
Explicitly frame the time
You can let ChatGPT decide, but with news, it is better to remove all ambiguity. Phrases like “What are the latest news on…”, “What happened today regarding…”, “What is the current situation with…”, or “Have there been recent developments on…” anchor the request in the present and rule out background analyses published two years ago. A sales director preparing for a meeting will get a very different result depending on whether they ask for general information about a client or the latest news about that client from the past month.
Time zones deserve the same attention. “Today” only makes sense relative to a location: if you are tracking a decision expected in Washington or a market close in Asia, specify it, or the answer may cover a day that is not yours.
Uses that structure monitoring
Sector monitoring is the most common case: “What are this week’s news in the cybersecurity sector in Europe?”, “Summarize the GAFAM announcements on artificial intelligence this week”, or “What are the latest regulatory changes in France regarding data protection?”. The formulation always combines three elements—a sector, a zone, and a time window—and this combination prevents the answer from drifting into the generic.
Competitive monitoring follows the same logic, applied to named actors. “What is the latest news on [company name]?” covers basic tracking; “Have there been any fundraising rounds in the [your sector] sector this month?” and “What new products have launched in [your market] recently?” broaden the scope to market movement. A product team that asks these last two questions at the start of every month builds a map of movements in minutes that they would otherwise discover by chance. Note that monitoring never fits in a single query: re-running the same question at the end of the day often reveals what was missing in the morning, simply because pages were published in the interim.
Tracking an event unfolding over several days—a conference, a crisis, a negotiation—requires a different approach. You do not re-run a general search; you ask for the delta: “Where does [the event] stand today?”, “What are the latest developments in [the situation]?”, “Summarize what has happened since yesterday regarding [the subject]”. The time savings come precisely from not having to reread what you already know. This is also the usage most exposed to error: on an ongoing event, the first available information is frequently incomplete, sometimes refuted a few hours later. Treat the first response as a provisional status and request an update before drawing a conclusion you will have to stand by.
Your turn
Your morning routine. Try the following briefing for a week, adapted to your own domains: “Give me a briefing of the 5 most important news items today in the following domains: artificial intelligence, French economy, European tech. For each news item, give me a 2-sentence summary and the source link.”
Then choose a single topic from the list and dig deeper: “Give me more details on [chosen topic], including viewpoints from different media outlets.” The value of this second step is seeing the differences in coverage that the briefing smoothed over. After five days, you will know if your three domains are the right ones, and above all, you will have measured the difference between reading the news and following it.
Once the prompt is refined, the ChatGPT documentation suggests scheduling a task in the same conversation: the briefing then arrives every morning without you having to re-launch it.
Official sources: Searching the web with ChatGPT · Prompting (ChatGPT documentation)
Key takeaways
- ChatGPT with web search accesses real-time news
- Use explicit time phrasing to target recent information
- Build a monitoring routine by requesting structured briefings
- For ongoing events, re-run your searches regularly
- Always verify breaking news with cited sources, as news evolves quickly