Assessing Source Credibility
From "Looks Legit" to "Here's Why I Trust It"
Last lesson you learned to leave the page: stop, investigate the source, find better coverage, trace to the original. That habit keeps handing you names, domains, and bylines, and this lesson gives you a way to rate them instead of guessing.
By the end you should be able to:
- Name the four credibility axes: expertise, independence, recency, agenda.
- Ask the right question for each axis in under a minute.
- Separate relevant expertise from a credential that only sounds impressive.
- Spot conflicts of interest and the incentives behind them.
- Turn the four axes into one repeatable verdict you can say out loud.
Four Axes Beat One Gut Feeling
A logo, a clean design, and a confident tone tell you nothing. What tells you something is a small set of questions you ask the same way every time.
| Axis | Question to ask | Weak signal | Stronger signal |
|---|---|---|---|
| Expertise | Does this person or outlet know this specific subject? | Credential from another field | Named, checkable, on-topic background |
| Independence | Does anyone benefit from this conclusion being true? | Pays for the result, or sells the product | No stake, or stake clearly disclosed |
| Recency | Is the claim still current? | Undated page, recycled old material | Dated, recent, updated with new evidence |
| Agenda | What is this source trying to get me to do? | Sells, shocks, or rallies | Informs, shows evidence, invites correction |
One axis rarely decides the verdict. You weigh them together and name the weakest link.
Expertise: Relevant Beats Impressive
The expertise question is not "is this person smart?" It is "does their knowledge cover this exact claim?" A PhD in literature is not evidence about battery chemistry. A famous surgeon is not evidence about earthquake forecasting. Credentials are shortcuts, so check the shortcut.
Two habits fix most expertise errors:
- Check domain match. Write the claim in one line, then write the source's actual field in one line. If they do not overlap, expertise is low.
- Verify the credential laterally. Search the person's name away from their own page. Look for an institution profile, a publication list, or a news mention that does not come from the person themselves.
Tip: Watch for the "titled stranger" pattern: a quote attributed to "Dr. Someone" or "a leading researcher" with no institution and no link. Unverifiable expertise counts as no expertise.
Try it. A post about a new bridge design quotes "Dr. M. Okafor, PhD" saying the design is unsafe. The profile link goes to a personal blog. The PhD is in musicology.
Click to reveal guidance
Expertise is low: the field does not match the claim, and the credential is self-reported with no institution. Note that this does not make the bridge safe. It means this quote cannot carry the claim. Your next step is a structural engineer or the published design review.
Independence: Follow the Money and the Incentive
Independence means the source has nothing material riding on the conclusion. Sources rarely have zero interests, so think in degrees, not yes or no. Ask: who paid for this, who employs the author, who profits if I believe it, and does the page disclose that?
A conflict of interest does not make a claim false. It raises the bar for how much evidence you should demand. The best sources disclose their interests openly and still show their work.
Try it. Two documents support the same claim about a new factory coating: the manufacturer's own test report, and a lab study the manufacturer also funded, published with its methods and data visible. Which would you weight more, and what would you check?
Click to reveal guidance
The published lab study, but only slightly, and only after checking who funded the lab, whether the manufacturer could veto publication, whether the methods are complete enough to repeat, and whether any unaffiliated lab has tried to replicate it. The first document fails independence outright. The second fails independence partially but partly compensates with transparency.
Recency: Every Claim Has an Expiry Date
Recency matters in proportion to how fast the subject changes. Software versions, product specs, prices, and scientific consensus all move. Historical dates and basic arithmetic do not. An older source is not wrong by default, but it may be superseded.
Three things to check: is there a publication or update date, does the date actually match the content (an "updated" banner on unchanged text is decoration), and has newer evidence appeared since. If the page is undated, treat recency as unknown rather than recent.
Agenda: What Does This Source Want You to Do
Every source has a purpose, and having one is not a sin. The question is whether the purpose is served by giving you accurate information. An informational source shows its evidence and would tell you what would change its mind. A sales or outrage source needs you to react before you check.
Practical signals: small "sponsored" labels, press releases republished with a news-style byline, affiliate links inside recommendations, headlines that describe your emotions rather than the finding.
Try it. A review of running shoes ends with a link to buy each pair, and the page carries a faint "we may earn a commission" line.
Click to reveal guidance
Agenda is commercial and disclosed, so independence is partial rather than zero. The review may still be useful, but treat the ranking as a sales document: check whether cheaper or unsponsored options were tested at all, and look for a second review that earns nothing from your purchase.
Running the 60-Second Credibility Pass
You do not need a long report. You need four short lines and one decision.
flowchart TD
A["📄 Claim and source"] --> B["🧠 Expertise<br/>does their knowledge match the claim?"]
B --> C["💰 Independence<br/>does anyone gain if I believe it?"]
C --> D["📅 Recency<br/>is it still current?"]
D --> E["🎯 Agenda<br/>what does it want me to do?"]
E --> F["⚖️ Verdict<br/>rate each axis, name the weakest"]
F --> G["🔍 Next step<br/>what evidence would change my mind?"]
Rate each axis high, medium, or low, and write one line of evidence for each rating. Then name the weakest axis and the specific evidence that would fix it. That last step keeps you skeptical without sliding into "everything is fake."
The expertise axis asks whether the source's knowledge is _____ to this specific claim, not whether the credential sounds impressive. A source is _____ when it has no financial or reputational stake in the conclusion. Recency matters most in fields where knowledge _____. Agenda means the _____ behind publishing something. A conflict of interest does not prove a claim false; it _____. You weigh the four axes _____, because one weak axis rarely settles the verdict alone.
Notes From the Verification Desk
- Rate the source, not the vibe. Pretty design and confident writing are cheap. Four named axes are not.
- Common mistake: treating independence as binary. Almost everyone has interests. Look for disclosure and for evidence strong enough to survive the interest.
- Common mistake: letting one failing axis end the hunt. A funded study can still be well designed. Say what is weak, then say what would repair it.
- Rule of thumb: if you cannot state in one sentence what would change your mind, you have not finished checking.
- Relevance beats prestige every single time. The right expert is the boring one who studies this exact thing.
What to Carry Into the Next Hunt
Four axes, one pass: expertise, independence, recency, agenda, scored with a line of evidence each, plus the weakest link named. That verdict is a tool, not a personality, so keep it pointed at claims rather than at people.
The next lesson takes the last step of the trail: once a credible source points at an original study, report, or dataset, how do you trace the claim back to that primary source and check that it says what the summary claims?