Finding and evaluating credible sources
A report built on five strong, well-chosen sources is worth more than one built on twenty mediocre ones. Knowing where to actually look, and how to quickly judge whether a source deserves your trust, saves enormous amounts of time and keeps weak claims out of your work before they ever get in.
Where to actually look
- Google Scholar — a good general starting point for academic literature across most fields; useful for its "cited by" links, which let you trace how influential a paper has been and find newer work that builds on it.
- Field-specific databases — IEEE Xplore and the ACM Digital Library for computer science; PubMed for medicine and life sciences; and your university library's own subscription databases, which often have access to paywalled journals you can't reach otherwise. Ask a librarian if you're unsure which is right for your field — genuinely, this is what they're there for.
- Preprint servers (arXiv, bioRxiv, SSRN) — useful for very recent work, but be aware these papers usually haven't been peer reviewed yet, so treat their claims with a bit more caution.
- Citation chasing — once you find one strong, relevant paper, look at both its reference list (what it built on) and its citing papers (what's built on it since). This is often a faster way to map a research area than starting from scratch with a fresh search each time.
Why is this important?
A weak or biased source doesn't just make one sentence of your work wrong — it can quietly undermine the whole argument built on top of it, and it's exactly the kind of thing a supervisor or reviewer will notice and question first. Learning to filter fast means you spend your reading time on sources that are actually worth citing.
Quick credibility checks
- Who wrote it, and where does it appear? A peer-reviewed journal article or a paper from a reputable conference has already been vetted by other experts before publication. A blog post or a company's marketing page hasn't — that doesn't automatically make it wrong, but it changes how much weight it should carry.
- How recent is it, and does that matter for your topic? A ten-year-old machine learning paper might be badly outdated; a ten-year-old paper on a stable mathematical result might be perfectly fine.
- Is it been cited by others, and how? A highly cited paper has generally been scrutinised by the field. Check briefly whether it's cited approvingly or as an example of a flawed approach — citation count alone doesn't tell you which.
- Does the author (or funder) have an obvious conflict of interest? A study on a product's effectiveness funded by the company that makes it deserves a slightly more sceptical read, not automatic rejection.
- Can you verify the claim elsewhere? A surprising or important claim resting on a single source is riskier to build on than one supported by multiple independent sources.
A word on AI-generated summaries and search results
Tools that summarise or find papers for you can be a useful starting point, but they can also misrepresent a paper's actual findings or cite work that doesn't say what it's claimed to say. Always trace a claim back to the original source before relying on it in your own writing — a summary is a pointer to the real thing, not a substitute for reading it.
Tips
- Set up alerts (Google Scholar has this built in) for key search terms in your project's area, so new relevant work surfaces automatically instead of requiring repeated manual searching.
- Keep a running list of sources you've evaluated and rejected, along with a one-line reason why. It stops you from accidentally re-evaluating the same weak source twice, and it's a useful record if a supervisor asks why a particular source wasn't included.
- Use a reference manager from the very start of your search, not just once you've settled on your final source list — it's much easier to keep organised than to reconstruct after the fact.