Real-Time News Tracking

Build an intelligent news feed

Traditional news feeds (RSS, Google Alerts) drown you in raw information. The Grok API lets you build a system that not only retrieves news but filters, contextualizes, and prioritizes it based on your business criteria.

You don’t want to know everything that’s happening. You want to know what matters for your business, why it’s important, and what you should do about it.

Configurable thematic alerts

Create custom alerts that go well beyond simple keywords:

def alerte_thematique(theme, contexte_entreprise):
    response = client.responses.create(
        model="grok-4.5",
        input=f"""Search for news from the last 24h
        on the theme: {theme}

        Context: {contexte_entreprise}

        For each relevant news item, indicate:
        1. The fact (one-sentence summary)
        2. Potential impact on our business
        3. Urgency (immediate / monitor / context)
        4. Recommended action

        Keep only news items truly
        relevant to our context.""",
        tools=[{"type": "web_search"}]
    )
    return response.output_text

alerte = alerte_thematique(
    theme="AI regulation in Europe",
    contexte_entreprise="SaaS company offering AI tools for French SMEs"
)

Real-time multi-source tracking

Combine web and X to get a complete real-time view of an ongoing event:

def suivi_evenement(evenement):
    response = client.responses.create(
        model="grok-4.5",
        input=f"""Real-time tracking of: {evenement}

        1. CONFIRMED FACTS (official web sources)
        - What is verified and published

        2. IMMEDIATE REACTIONS (X/Twitter)
        - What experts and influencers are saying

        3. TIMELINE
        - Sequence of known events

        4. UNCERTAINTIES
        - What is not yet confirmed

        Clearly distinguish facts from speculation.""",
        tools=[
            {"type": "web_search"},
            {"type": "x_search"}
        ]
    )
    return response.output_text

Generate a structured newsletter from sector news:

def newsletter_secteur(secteur, sources_fiables):
    response = client.responses.create(
        model="grok-4.5",
        input=f"""Write a professional newsletter on
        the {secteur} sector for the past week.

        Structure:
        - Editorial: the highlight of the week (3 sentences)
        - Top 5 news items (summary + source)
        - Figure of the week (notable data)
        - Watch list (emerging trend)

        Tone: professional, concise, factual.
        Length: 500 words maximum.""",
        tools=[{
            "type": "web_search",
            "allowed_domains": sources_fiables
        }]
    )
    return response.output_text

Crisis monitoring

In a crisis situation (bad buzz, technical incident, unexpected regulation), speed of information is critical:

def monitoring_crise(sujet_crise, notre_marque):
    response = client.responses.create(
        model="grok-4.5",
        input=f"""CRISIS MONITORING - {sujet_crise}

        1. Scope: what is the current media coverage?
        2. Virality on X: is the topic trending?
        3. Mentions of {notre_marque}: are we associated?
        4. Dominant narratives: which angles are being picked up?
        5. Influencers involved: who is amplifying the message?

        Prioritize accuracy over speed.
        Cite all sources.""",
        tools=[
            {"type": "web_search"},
            {
                "type": "x_search",
                "enable_image_understanding": True
            }
        ]
    )
    return response.output_text

Manage volume and relevance

The main challenge in news tracking is not collection but filtering. Here are a few techniques to improve relevance:

Define a precise business context in each prompt so the model filters with full knowledge
Prioritize by asking for ranking by impact or urgency
Differentiate sources: explicitly ask to separate verified facts from opinions
Limit volume: ask for a specific number of news items (top 5, top 10) to force prioritization

Hands-on practice

Configure news tracking on a professional topic of your choice. Run it for three consecutive days and evaluate the relevance and freshness of the results. Adjust the prompt and filters to optimize the signal-to-noise ratio.

Key takeaways

Contextualized thematic alerts are more useful than keyword alerts
Combining web + X gives a complete view (facts + reactions)
Structure prompts with business context to improve filtering
Crisis monitoring requires both sources with image analysis
Always ask for prioritization to avoid information overload
Automated newsletters are a proven format for internal distribution