Not every source is worth citing. This sounds obvious until you're under deadline pressure and the first result that backs up your claim is a Medium post from 2019 with no author attribution. You cite it. Two weeks later a reader emails to say the statistic was fabricated. That happened to a team we know. Took them about 17 minutes to find the original erroneous post — and another 3 weeks to update every article that had cited it downstream.
The source evaluation problem isn't about being rigorous for its own sake. It's about not building your content's credibility on someone else's mistake.
The 4-Question Credibility Check
Before citing anything, ask these four questions. Who published it — is this an institution, a recognized publication, or a named expert with verifiable credentials? When was it published — is the information recent enough to still be accurate? What's the primary source — does this article cite something, or is it the original data? And can you find corroboration — does at least one other credible source make the same claim? If a source fails two or more of these, find a better one.
The corroboration question is the one most people skip. It takes 30 extra seconds and catches a surprising number of errors. A claim that appears in only one place, no matter how credible the source, is worth verifying against a second source before citing. Even well-regarded publications publish errors. That's not cynicism, it's how editorial processes work.
Source Tiers That Actually Matter
Not all sources carry the same weight, and it's worth being honest about this. Peer-reviewed research, government data, and official industry reports are the gold standard — use them when available, especially for statistics. Major news publications, established trade press, and recognized research firms are solid for context and trends. Expert blog posts with identified authors and cited data are fine for opinion and framing, but never for statistics. And then there's everything else: anonymous content, aggregator sites, press releases. Treat those as leads to a better source, not sources themselves. Would you cite a press release in a research paper? Same standard applies here.
There's a practical tier that often gets overlooked: official product documentation, government agency websites, and academic institution pages. These are underused in content marketing despite being extremely credible. A statistic from the Bureau of Labor Statistics is more trustworthy than the same figure quoted in a Forbes opinion piece — even though the Forbes article is more well-known. Go to the primary source when you can find it.
When Wikipedia Is Actually Fine
Wikipedia gets unfairly dismissed as a source. For general definitions and background context — what a term means, when a company was founded, the basic structure of a process — Wikipedia is usually accurate and well-maintained. The problem is that it's not citable for specific claims or statistics, because the information can change and often traces back to a source you should be citing directly.
Think of Wikipedia as a research starting point, not a destination. It's genuinely good for understanding context quickly and finding references. Almost every Wikipedia article has a reference section at the bottom. That section often contains the primary sources you actually want to cite. Use Wikipedia to find the source, then cite the source.
Evaluating Recent vs. Old Sources
Recency matters differently depending on the topic. For fast-moving fields — AI, crypto, social media platforms, regulatory environments — a 2-year-old source might be significantly outdated. For foundational research — how attention works, behavioral economics, basic biological processes — a 10-year-old peer-reviewed study is often still valid. The question isn't the age alone, but whether the underlying facts could have changed.
A specific pattern to watch: statistics about technology adoption, market share, or user behavior tend to become outdated quickly. A 2021 stat about how many people use AI writing tools looks very different in 2026. When citing figures in fast-moving areas, check whether a more recent study exists before using older data. Even if the general trend still holds, a reader who knows the field will notice an outdated number.
Where AI Helps and Where It Doesn't
AI is genuinely good at finding source candidates quickly. Citeya's automatically surfaces relevant sources ranked by domain authority and recency before generating the article — so you're not starting from scratch on source research. What AI can't do reliably is verify that the specific claim in a source is accurate. That's still a human job. Use AI to find sources, use your judgment to evaluate them.
One thing AI tools do well that's easy to overlook: they can often surface sources that a regular Google search would bury. Search engines optimize for popularity. AI source-finding can surface authoritative but less-linked government reports, academic papers, or industry association data that would otherwise take 20 minutes to find manually.
The Fast Verification Habit
For every statistic you cite, open the source, find the specific number, and check that what it says matches what you're saying in your article. Sounds basic. Almost nobody does it consistently. Statistics get mangled in transit — a study finding that '62% of marketers plan to increase AI investment' becomes '62% of marketers use AI' in the third retelling. That's the kind of error that gets screenshotted and shared in the 'bad AI content' thread on LinkedIn. Don't be that example.
Build the habit by adding it to your publishing checklist rather than trying to remember it each time. 'Every statistic verified against primary source' should be a checkbox the same way 'meta description filled in' is. The habit takes about 5 minutes per article once it's automatic. That's a small price for not having to issue corrections.
The source isn't just proof for your reader. It's proof for you that you actually said something true.