Today I ran an early test across three AI systems to see how up to date they are on current events, and more importantly, how they respond to a high-level prompt with no additional guidance.
The prompt was intentionally simple:
“Summarise latest news on the Middle East war and the Russia–Ukraine war.”
The goal wasn’t to trick the models.
It was to see how each one interprets ambiguity when asked to condense something as complex and fast-moving as global conflict.
What stood out wasn’t just the information, it was the thinking style behind the summaries.
ChatGPT interprets the request through continuity.
It doesn’t just summarise, it contextualises. Sometimes too much. It pulls in historical framing and inferred intent, which is powerful when you want depth, but can get in the way when you’re simply trying to scan what happened today. The result often feels structured, almost like a narrative brief or a PowerPoint without slides.
Where it stands out is in what it adds unintentionally – connecting dots, surfacing patterns, and nudging you toward themes you didn’t explicitly ask for. It leans toward analysis over digestion.
Gemini treats the prompt more literally.
It prioritises clarity, speed, and clean summarisation – closer to how you’d expect a news digest to read. The outputs are tighter, with clearer citation structures, and particularly strong on “what’s current right now.” This likely reflects Google’s strength and experience in indexing and ranking real-time information.
The trade-off is depth. It can feel slightly too compressed, but it compensates by guiding you – prompting what to explore next. It leans toward efficient digestion with light guidance. Perhaps it it worked exactly as expected, I did ask it to summarise.
Claude approaches the prompt with curiosity.
It expands the scope beyond the obvious headlines, sometimes pulling in secondary systems or less visible dynamics others miss. This makes it particularly strong when the goal is to understand what’s happening beneath the surface. This read like an executive summary if I were assessing business risk related to geopolitics and supply chain impact.
But it’s inconsistent. In the same response, it can deliver sharp, insightful analysis in one half, and then default to generic or misaligned summaries in the other – giving broader stats since the start of the war than recent updates. It leans toward depth-first exploration, but with variable precision.
What this test really shows:
The same vague prompt produces three different interpretations of “summarise”:
ChatGPT → Contextualise and expand
Gemini → Condense and clarify
Claude → Explore and enrich
So the question isn’t which one is better, it’s what kind of thinking you want applied to ambiguity.
If you want patterns and perspective, ChatGPT
If you want clean, current snapshots, Gemini
If you want hidden layers and deeper systems, Claude
These tests are still early. I’m figuring out how to personally benchmark these across different use cases, and with how frequently these models update, the baseline is constantly shifting.
I’m keen to see where this experimentation leads, but one thing is clear. If I were researching something, I would likely use all three. Even the free tiers help condense data well enough to explore the breadth of a topic and give you enough citations to evaluate the relevance yourself.