Compare sources and cross-check information
On this page
A single source does not make a truth
Cross-checking information is the core skill of researchers, journalists, and analysts—and also the most labor-intensive. You have to open multiple documents, keep track of what each one says, and spot where they diverge. ChatGPT significantly lightens this load by letting you run the comparison within a single conversation. But you must ask for it explicitly: a simple question invites a simple answer, and a simple answer almost always masks disagreements.
The need for cross-checking stems from the very nature of sources. A media outlet has an editorial line that shapes what it chooses to report and how it presents it. A company communicates to serve its interests; that doesn’t make it a liar, but it does make it a stakeholder. A study may rest on questionable methodology or biased funding. And social media amplifies opinions without any fact-checking filter. None of these flaws is fatal on its own; it’s their overlap that lets you separate facts from interpretations, spot consensus, and identify areas where uncertainty remains.
Ask for multiple points of view
Your first move is to turn your question into a comparison. “What do French AND international sources say about [topic]?” forces the response to separate two sets instead of blending them. “Compare media coverage of [event] across Le Monde, Reuters, and the Financial Times” goes further by naming specific outlets, making divergence visible title by title. However, ChatGPT cannot read every page: the help center cites paywalls, technical issues, and sites that block indexing bots. So verify that each named outlet actually appears among the cited sources. And “What are the arguments FOR and AGAINST [topic] according to different sources?” imposes a contradictory structure that a spontaneous synthesis would have smoothed over. A manager preparing a decision has every reason to frame the prompt this way: they’ll see where opinions split, which is precisely the information they need.
One precaution accompanies this move: check who owns the cited sources. Three links with three different names may belong to the same media group, in which case the apparent plurality means nothing. A comparison between two titles from the same group will return perfect agreement that proves absolutely nothing.
Track the origin of a figure
Statistics circulate while distorting, and a datum repeated ten times ends up seeming established even though no one has verified its origin. Three questions are enough to break this effect: “Is the figure of [X] confirmed by multiple sources? Which ones?” tests convergence; “What is the original source of this statistic on [topic]?” traces the chain back to the document that produced it; “Are there studies that contradict this figure?” actively seeks dissonance. When a study is cited, push one step further by asking for its methodology and sample size—a survey of a few dozen respondents cannot support the conclusions people attribute to it.
Popularity signals don’t help you here. An article shared ten thousand times is no more accurate than a report read by three hundred people; it’s just better written or better promoted. Faced with a comment, systematically look for the document it comments on: an official report, a study, or raw data beats ten articles that paraphrase them, and it’s often by opening the original that you discover the nuance the entire chain of reposts lost.
Compare offerings, map a consensus
For products or services, ChatGPT excels at structured comparisons—provided you give it a timeframe and demand references: “Compare [product A] and [product B] based on tests and reviews published in 2026. Provide sources.” Audience distinctions matter too, since they judge by different criteria: “What are the pros and cons of [service] according to users AND experts?” reveals the gap between everyday experience and technical evaluation.
On debated topics, map the landscape rather than hunting for a single answer. “On [topic X], is there a scientific consensus? What are the minority positions?” and “Do experts agree on [claim]? Cite sources that converge and those that diverge.” return a dominant position and its challenges, which is far more honest information than a smooth assertion.
Your turn
The same topic from two sides. Pick a controversial current event and proceed in three steps. First, ask what major French media say about the topic, then what English-language international media say, and finish with a synthesis: “Comparing French and international sources, what are the points of consensus and disagreement on [your topic]?”
Observe how your understanding shifts between the first and third steps. In most cases, the central fact doesn’t change—what changes is what surrounds it: the causes advanced, the actors highlighted, the relative importance assigned to the event. That gap is completely invisible as long as you stick to a single family of sources, which is why cross-checking is worth the five minutes it costs.
Official sources: Searching the web with ChatGPT · Does ChatGPT tell the truth? (OpenAI Help Center)
Key takeaways
- Cross-checking sources is essential to distinguish facts from opinions
- Explicitly ask ChatGPT to compare multiple points of view
- Verify figures by tracing them back to the primary source
- Identify areas of consensus and disagreement
- Always prioritize primary sources (studies, official reports, raw data) over secondary sources