STUDboardsPlay · Research · Self · Fit · Money · Learn · Study · Plan · Token STUDtokenOptional

STUDlearn · Learning

Making sense of online information

Also known as

evaluating information · fact checking · media literacy · online research · is this true · checking sources · misinformation · information overload · AI image · AI generated photo · synthetic image

Named sources. May be wrong or incomplete. Not medical, legal, financial, or other professional advice.

Open in the directory ↗

Check what supports an online claim and decide what to do with it.

An adult comparing information on a phone and laptop at a library desk

What would help today?

Understand a screenshot or short clip
Check the original
Find the full notice or recording; check its date, question, qualifications, and who it applies to.
Two separate questions
Does the screenshot look authentic, and have the facts actually been checked?

A screenshot can accurately copy a sentence while leaving out the question, date or qualification that changes its meaning. Media educator Renee Hobbs distinguishes how authentic a message looks from whether its facts have been checked. Those are separate questions. Before using a clipped announcement to make a plan, look for the full notice or recording and check who it applies to. An event advertised last year might have a different date now. This ordinary example illustrates context checking; a polished image is neither proof of truth nor proof of fakery.

Sources 2
Make sense of several sites saying the same thing
Trace the source
Follow the links to see whether the pages report separate observations or repeat one underlying announcement.
Count the underlying reports
Three pages quoting one venue notice still give you one notice to check.

Several pages may be repeating one announcement rather than reporting several separate observations. Claire Wardle and Hossein Derakhshan distinguish the people who create, produce and distribute a message. The person sharing it with you need not be its originator. Follow the links far enough to see whether the pages rely on the same underlying report. For example, three event roundups quoting one venue notice still give you one notice to check. Repetition can help you find information, but the number of copies does not add new evidence by itself.

Sources 2
Assess a recommendation from an expert or creator
Name the exact benefit
Separate an experience such as “I enjoyed it” from a factual promise such as “it guarantees employment.”
Match the evidence
Check support for that specific promise, with expertise and financial interests as context.

A useful recommendation can combine firsthand experience, specialist knowledge and an interest in selling something. Alexios Mantzarlis’s fact-checking guidance separates a checkable factual assertion from opinion and asks how close the evidence is to the thing being claimed. Start with the precise benefit you care about. “I enjoyed this workshop” is different from “this workshop guarantees employment.” You can value the experience while asking for stronger evidence for the guarantee. Relevant expertise and financial interests provide context; neither a famous name nor sponsorship settles the claim by itself.

Sources 2
Finish a useful check without opening endless tabs
Define the check
Write the one thing you need to decide and what would answer it.
Finish usefully
Use the source closest to that decision and save the answer with any unresolved detail.

Research can wander when each page introduces another question. Mike Caulfield’s SIFT guidance includes stopping to recover your purpose, not simply searching for longer. Write down the one thing you need to decide and what would answer it. If you are checking whether a community workshop is happening, the organizer’s current notice is more relevant than a long argument about the topic of the workshop. Save the result and any unresolved detail. That is a practical adaptation, not a universal two-minute rule or a reason to rush a consequential decision.

Sources 2

Good to know. This is a reading companion, not a professional assessment of a particular claim. Use it to organize questions, not to diagnose a condition, make a financial commitment or settle a legal dispute. The amount of checking should fit the consequence of being mistaken. For consequential decisions, take the specific claim and supporting material to a suitably qualified person. Keep private information out of public searches and requests for help. Missing links do not, by themselves, make someone’s reported experience false.

What the research found

Where experts still disagree

  • Choose whether the source or the specific claim needs checking. A source-first approach investigates who is speaking; a claim-first approach looks directly for evidence about the assertion. They answer different questions. Lisa Oswald and colleagues compared them in a randomized study of 2,666 German adults. Immediate gains were small, and some were too uncertain to conclude that the lessons helped. At two weeks, scores had improved in all groups, including the comparison group; the study could no longer clearly separate the groups’ results. That does not prove the lessons were forgotten or establish a universal winner. Choose the entry point that addresses your uncertainty.
    Sources 1
  • Pair a search with a reason to trust what it returns. Searching might uncover better evidence, or it might supply more copies of the same weak claim. Aslett and colleagues tested these competing possibilities in five experiments. In the first, people prompted to search were more likely to believe selected false or misleading news than people not encouraged to search. Across the research, this problem was concentrated among people receiving low-quality results. That association does not prove exactly why it happened or test whether trained source-checking would work better. Search systems also change. Look for support whose origin and relevance you can explain, not just apparent agreement.
    Sources 1
  • Balance questioning with reasons for confidence. One approach emphasizes skepticism about doubtful stories; another teaches recognition of trustworthy news; a mixed approach does both. Altay and colleagues tested all three in a US online experiment that recruited 3,919 participants. Each improved headline discernment compared with no tips, and their effects were more alike than different. Mixed tips both increased acceptance of true headlines and reduced acceptance of false ones. These immediate ratings and sharing intentions do not establish permanent behavior or justify trusting every outlet. A constructive check should be able to conclude either “this is supported” or “this needs more evidence.”
    Sources 1
  • Use preparation and correction at the moment they help. Prebunking teaches a misleading technique before someone meets a new example; debunking corrects a claim after exposure. Roozenbeek and van der Linden explain the value of advance preparation, while correction researchers emphasize repairing an existing misunderstanding. Tay and colleagues directly compared pre-emptive and retroactive interventions in an experiment involving fair-trade misinformation and 735 participants. Both generally reduced reliance on misinformation. That does not establish equivalence or a winner across every message and audience. Practising a technique and correcting a particular error can serve different needs.
    Sources 2

Just talk

  • Let relevant evidence earn your confidence. A credential, logo or professional-looking page is not a guarantee that every statement is accurate. The Commons Library’s source guide treats expertise as relevant context alongside the actual method and claim. Check whether the expertise fits this question. You can respect someone’s work without treating every subject they discuss as their specialty.
    Sources 1
  • Separate repeated wording from new support. Seeing a claim again does not confirm it. Fazio’s familiarity research helps explain why repetition can feel reassuring, even without better evidence. Ask what has been added: another direct observation, a different analysis, or simply another copy. A familiar statement may be true, but familiarity is not the reason that makes it true.
    Sources 1
  • Use a stopping point that fits the decision. Two links or two minutes do not turn every question into a completed investigation. Caulfield’s method returns attention to purpose and context, not a fixed score. You might settle a routine event date quickly while leaving a complex benefit claim unresolved. Record that distinction instead of forcing every result into a confident yes or no.
    Sources 1
  • Use media history as one part of the check. Content Credentials can provide information about an image’s recorded origin or editing history. The C2PA explainer explicitly distinguishes that history from whether the depicted claim is factual. Credentials do not certify the truth of a caption, and their absence is not proof of a fake. Keep checking what the image is being used to establish.
    Sources 1

What to try

  • Choose one claim and the evidence that would answer it. Write a single checkable sentence from something you want to use or share. Mantzarlis distinguishes factual claims from opinions: “the workshop starts at six” can be checked against a current schedule; “it looks enjoyable” expresses a judgment. Next to your sentence, name the evidence you need and where you expect to find it. Keep the original wording attached so the question does not drift while you look. This everyday exercise adapts professional fact-checking guidance; it is not an automatic verdict.
    Sources 1
  • Compare the shared extract with the original. In a lesson coauthored by Peter Adams and Suzannah Gonzales, students compare a circulating broadcast screenshot with the original recording. Try the same kind of comparison with an ordinary public notice or quotation: look at the surrounding passage, date and attribution. Note exactly what stays the same and what changes. Finding an older original can clarify context without establishing anyone’s intent.
    Sources 2
  • Keep a short conclusion and the unresolved detail. Caulfield’s SIFT guidance brings a search back to its purpose and original context. Apply that idea by writing what the evidence supports and what you still need to ask. For example: “The organizer lists Saturday; the shared image is from last year; accessibility arrangements are not stated.” Save the current notice beside that conclusion. You now have a specific follow-up question rather than another collection of tabs.
    Sources 1
  • Inspect one shared image for details, text, and an outside source. Sean Guo, Briony Swire-Thompson and Xiaoqing Hu compared specific tips for spotting AI-generated images with general fake-news tips and no tips among 1,012 U.S. adults. Specific tips—look for abnormal details such as extra fingers, incoherent or misspelled text, and a sharp subject with a blurry background—improved discernment between real headlines and AI-image headlines more than general tips. Both kinds of tips also lowered belief in real headlines, so extra skepticism is not the same as better judgment. C2PA Content Credentials and Mike Caulfield’s SIFT method remain useful companions: a visual clue is a reason to check, not a verdict. Open the original post, look for a current official notice, and treat missing credentials as unknown rather than proof of a fake.
    Sources 5

How to keep it

  • Keep the question beside the result. A source may answer a neighboring question rather than yours. The Commons Library emphasizes relevance and currency. Save the date, the exact claim and the particular passage you used. For a recurring event, an old timetable can explain its history without telling you this weekend’s hours.
    Sources 1
  • Distinguish the original message from its journey. Wardle and Derakhshan separate creation, production and distribution. That makes a useful question when a friend forwards a post: where did the underlying information begin? Preserve a helpful firsthand observation while checking stronger general claims separately. An unavailable original leaves uncertainty; it does not establish falsehood or bad motives.
    Sources 1
  • Match the confidence to the scope. A clear account of what happened to one person can be useful without proving what will happen to everyone. Mantzarlis’s evidence questions include proximity and rigor: how was the information obtained, and what conclusion can it carry? Keep the specific observation, then label the wider inference as an inference rather than quietly upgrading it to a universal rule.
    Sources 1
  • Give the recipient enough context to use what you share. Before forwarding a useful find, include the detail that matters to the other person: which date, location or group it applies to. Pennycook and Rand’s accuracy-prompt work distinguishes sharing judgments from truth itself. Checking your message’s implication is a practical adaptation; it is still worth opening the source rather than relying on a confident caption.
    Sources 1
  • Make a correction useful enough to replace the old account. Stephan Lewandowsky and fellow authors of the Debunking Handbook explain that removing an error can leave a gap in someone’s understanding. State the accurate information and the reason for the correction. For a changed venue, send the new location and the current notice, not just “that was wrong.” This ordinary example does not guarantee agreement or erase every earlier impression.
    Sources 2

Sayings people repeat

The longer notes

  • Explain why an outside source changes your judgment. Breakstone and Wineburg’s course included demonstrations of how evaluators reasoned about sources. Try making that reasoning visible in an everyday case. Suppose a course advertises recognition by an organization, and you find that organization’s own list of recognized courses. A matching entry supports the recognition claim; a testimonial about enjoying the course answers a different question. Neither tells you whether its timetable suits you. Write which claim the outside record answers and what remains open. That is the useful work of looking beyond the advert, rather than simply collecting another link.
    Sources 1
  • Choose a source for the part it can actually know. A venue knows its current opening arrangements; an attendee can describe the experience of visiting; a researcher may examine broader patterns. These are different contributions. The Commons Library recommends relevant expertise, stated methods and appropriate scope. Ask what the source can directly support, then seek another kind of evidence for the unanswered part. A useful answer can combine official records with well-described personal experience.
    Sources 1
  • Read the condition attached to a quotation. Adams and Gonzales’s broadcast-comparison lesson gives a useful way to examine a clipped statement. Imagine a meeting report quoting “the room will be available,” while the full answer continues “if the repairs finish by Friday.” The quotation may copy the words correctly, but the longer answer describes a condition rather than a confirmed booking. Check whether that condition has since been met. The original recording resolves what was said; it does not, by itself, establish today’s availability or explain why someone shared the shorter version.
    Sources 1
  • Separate a commercial purpose from the claim’s evidence. Amazeen’s work shows why advertising styled as editorial content can be difficult to recognize. A sponsorship label tells you something important about purpose, but it does not test the promised benefit. Read what is actually claimed, who benefits, and what independent evidence is supplied. A practical product demonstration can still be informative; the stronger the promise, the more important it is to distinguish the demonstration from the conclusion drawn from it.
    Sources 1
  • Check what a media credential actually records. Provenance is an account of where an asset came from and how it changed. C2PA’s versioned explainer describes credentials that can help establish aspects of that history, not a universal truth detector. A genuine photograph of a real event can still be paired with an incorrect place or date. Conversely, a missing credential may simply mean no such record was attached or preserved. Ask which part of your claim the available history supports.
    Sources 1
  • Learn the reasoning behind a manipulative technique. Jon Roozenbeek and Sander van der Linden describe prebunking as more than a general warning to be suspicious. Their explanation combines advance warning, a weakened example of a persuasive technique and a way to answer it. As a simple illustration, a choice presented as “this option or nothing” deserves a check for other available options. The example teaches a question to ask; it does not make every two-option choice false or provide permanent immunity from persuasion.
    Sources 2
  • Give a corrected fact an explanation it can fit into. Briony Swire-Thompson and Emily Vraga coauthored the Debunking Handbook, which explains why an old account can keep influencing later reasoning. Imagine hearing that a workshop was canceled because nobody enrolled. Learning that many people enrolled removes one claim but leaves the cancellation unexplained, so the old story may still come to mind. If the organizer confirms a building repair caused the cancellation, include that explanation and its source. This example illustrates how an accurate account fills the gap; inventing a convenient reason would create a new problem.
    Sources 3
  • Keep track of what changes your judgment. Oswald and colleagues treated searching for information and judging it accurately as separate outcomes. Your own notes can make a similar distinction. After a check, record your starting conclusion, the detail that changed it and the question still open. You might discover that ten extra pages added nothing, while a date on the original notice settled your practical question. This is a reflection exercise, not a test score. A useful record names the evidence that mattered rather than treating confidence or time spent searching as proof of accuracy.
    Sources 1
  • Use specific image checks alongside provenance and outside sources. Guo, Swire-Thompson and Hu’s preregistered experiment used twenty real headlines with real photos and twenty false headlines with AI-generated images, without sources or engagement counts. Specific tips improved discernment and, in mixed models that accounted for headline and person differences, reduced belief in the AI-image headlines more than general tips. The authors warn that generators improve: some already produce readable text, and only half of their AI images showed the features named in the tips. Renee Hobbs separates how authentic a message looks from whether its facts have been checked, and C2PA credentials record file history rather than caption truth. Combine a closer look at the picture with a source check, a reverse search when useful, and the original context. Visual tips help; they do not replace lateral reading or a qualified check for a high-stakes claim.
    Sources 5

Who this is drawing from

  • Sam Wineburg. Stanford professor emeritus and Digital Inquiry Group cofounder. His source-evaluation research and public teaching inform lateral reading. Student and instructional findings do not establish a universal truth detector.
    Sources 1
  • Joel Breakstone. Digital Inquiry Group executive director and education researcher. His online-course work connects explanations with practice and demonstrations. Using a checking strategy more often does not necessarily mean using it correctly.
    Sources 1
  • Mike Caulfield. Creator of SIFT and a public educator on web research. His method helps a reader recover purpose and context. It is practical guidance, not an independently validated time limit for every question.
    Sources 1
  • Lisa Fazio. Vanderbilt psychologist researching learning and belief. Her work explains how familiarity can influence perceived truth even when knowledge is available. It does not say knowledge is useless or estimate an individual’s susceptibility.
    Sources 1
  • Gordon Pennycook. Cornell psychologist studying judgment and misinformation. His accuracy-prompt research separates attention to accuracy from automatic belief. Sharing judgments remain different from verified real-world behavior.
    Sources 1
  • David Rand. Cornell researcher in behavioral and computational social science. His coauthored accuracy studies examine what people choose to share. Disclosed advising relationships with technology companies are context, not evidence for or against a particular finding.
    Sources 1
  • Sander van der Linden. Cambridge social psychologist and public educator on persuasion. His prebunking explanation teaches a technique and the reasoning needed to answer it. Immunity is a metaphor, not a guarantee of permanent protection.
    Sources 1
  • Stephan Lewandowsky. University of Bristol cognitive psychologist and Debunking Handbook coauthor. His contribution here is explaining why a correction should also supply an accurate account. A research-informed correction cannot promise another person’s agreement.
    Sources 1
  • Briony Swire-Thompson. Northeastern researcher studying correction and belief. Her contribution to the Debunking Handbook helps distinguish remembering a correction from fully revising later reasoning. It is not a clinical or interpersonal treatment protocol.
    Sources 1
  • Emily Vraga. University of Minnesota communication researcher and Debunking Handbook coauthor. Her work brings correction into communication settings. Message, platform and audience differences limit all-purpose prescriptions.
    Sources 1
  • Michelle Amazeen. Boston University communication researcher. Her public account of native-advertising studies helps separate editorial appearance from commercial purpose. Detecting sponsorship does not settle whether a product claim is accurate.
    Sources 1
  • Claire Wardle. Cornell communication scholar and public information-literacy educator. Her work with Hossein Derakhshan distinguishes a message’s creation, production and circulation. That framework does not license guessing a sharer’s motives.
    Sources 1
  • Peter Adams. News Literacy Project research and design leader. A lesson he coauthored asks readers to compare a circulating screenshot with the original broadcast. It is a concrete teaching example, not evidence that a brief check works on every image.
    Sources 1
  • Alexios Mantzarlis. Cornell Tech SETS director and former International Fact-Checking Network leader. His UNESCO chapter distinguishes checkable claims from opinions and examines the quality of supporting evidence. It offers professional method, not an automatic verdict.
    Sources 1
  • Renee Hobbs. University of Rhode Island professor and Media Education Lab founder. Her framework distinguishes selected representations, authenticity and factual checking. It is a way to question a message, not a visual AI-detection test.
    Sources 1

Related

Good to know

  • Good to know. This is a reading companion, not a professional assessment of a particular claim. Use it to organize questions, not to diagnose a condition, make a financial commitment or settle a legal dispute. The amount of checking should fit the consequence of being mistaken. For consequential decisions, take the specific claim and supporting material to a suitably qualified person. Keep private information out of public searches and requests for help. Missing links do not, by themselves, make someone’s reported experience false.
    Sources 2
  • Not advice. Named sources. Honest paraphrase of the finding. Not medical, legal, or financial advice.

All STUDlearn topics →