Fact-Checking & Verification Intern

Fact checking is methodical work, and the method matters more than the conclusion.

In this task you will verify one widely circulated claim from start to finish, documenting every step so that a reader could repeat your work and reach the same result.

Everything is done online.

Select one claim currently circulating on social media or messaging platforms that makes a specific, checkable assertion.

Claims about numbers, dates, who said what, or what a document contains are ideal.

Avoid matters of pure opinion, which cannot be fact checked.

First, establish exactly what is being claimed.

Write the claim in one precise sentence, because vague claims cannot be verified.

Next, trace its origin.

Search for the earliest version you can find, note the date and account that posted it, and record how it changed as it spread.

Screenshot each significant version with its date.

Now verify against primary sources.

If the claim concerns a law, find the law.

If it concerns a statement, find the full recording or transcript rather than the clip.

If it concerns data, find the original dataset from the issuing authority.

Record what each primary source actually establishes.

Write a fact check report of 600 to 700 words.

State the claim, give your verdict as true, false, misleading, or unverifiable, then present the evidence in the order you gathered it.

Explain clearly why the claim spread, since that often matters as much as whether it is true.

Be rigorous about the misleading category, which is where most real claims land.

A claim can be literally accurate yet create a false impression through missing context, a selective date range, or a comparison that does not hold.

Where your claim falls into this territory, state precisely what is true about it and precisely what it leads a reader to believe wrongly.

Then write the correction in a form that works.

Research shows that simply repeating a false claim to deny it can reinforce it, so write a short corrective paragraph that leads with the accurate position rather than with the falsehood, and explain the choices you made in doing so.

Submit the report, the corrective paragraph, your full evidence trail with screenshots and dates, and every source link.

Most visual misinformation is not fabricated at all.

It is a real image or video from another place or another time, attached to a false description.

Learning to find the original is therefore the core verification skill, and it can be done entirely from a browser at home.

Collect five images or videos currently being shared online with specific news claims attached to them.

Record where you found each one, the date, and the exact claim being made about it.

For each item, work through a verification sequence and document every step.

Begin with reverse image search across multiple search engines, since they index different material.

For videos, take screenshots of distinct frames and reverse search those individually.

Examine the visual content itself for clues about location and time, such as signage, language on boards, vehicle registration formats, weather, clothing, construction, or shadows.

Check whether the item appears in any earlier published article or archive.

For each of the five, record your conclusion: genuine and correctly described, genuine but from a different event or date, digitally altered, or unverifiable with available tools.

State the specific evidence behind each conclusion, not merely the verdict.

Pay attention to the limits of these tools.

Reverse image search works poorly on heavily cropped images, on frames from video, and on anything recently created.

Where a check produced nothing, record whether that means the item is original or simply that the tool failed, since treating an absent result as proof of authenticity is a serious error.

Also consider synthetic media.

For at least one item, examine it for the signs associated with generated or manipulated imagery, such as inconsistent lighting across a frame, irregular detail in hands and text, or edges that do not behave correctly.

Record what you looked for and what you found.

Then write a short method note explaining which checks produced results most often, which produced nothing, and what you would do differently if you had to verify fifty items under deadline pressure.

Submit a verification sheet covering all five items with evidence and links, the screenshots you used, the synthetic media assessment, and the method note.

Verification is not only about individual claims.

Journalists constantly decide how much weight to give a source, and doing that consistently requires a framework rather than instinct.

This task is entirely research based and done from home.

Select eight websites or social media accounts that regularly publish news or news-like claims.

Deliberately choose a mix: established news organisations, newer digital outlets, aggregator sites, and openly partisan outlets.

First, build your assessment framework.

Define between five and seven criteria you will apply to all eight equally.

Reasonable criteria include ownership transparency, whether staff and editors are named, whether sources are cited and linked, whether a corrections policy exists and is used, how clearly opinion is separated from reporting, and the track record of accuracy.

For each criterion, define in advance what a strong, moderate, and weak result looks like, so your assessment does not simply reflect which outlets you already like.

Now apply the framework to all eight sources.

For every criterion and every source, record the specific evidence behind the rating with links.

Where a source does badly, note it with evidence.

Where a source you expected to do badly performs well, note that too.

Compile the results into a comparison table with the ratings and an overall assessment of each source.

Then test the framework against its own weakness.

A credibility framework that simply ranks established outlets highest tells you nothing you did not already assume.

Examine whether yours would correctly identify a well presented but unreliable source, and whether it would unfairly penalise a small outlet doing careful work with limited resources.

Revise any criterion that fails this test.

Next, consider the single story problem.

A source with a strong overall record can still be wrong on a particular story, and a weak source can be right.

Write a short section on how you would weigh source reputation against the specific evidence in front of you.

Finally, write a two page note on what the exercise revealed, including at least one case where your prior assumption about a source turned out to be wrong, and how you would use this framework under deadline pressure.

Submit the framework, the evidence table, the revision notes, the single story section, and the final note.

Individual fact checks correct individual claims, but misinformation moves in recognisable patterns.

Learning to see the pattern lets a journalist anticipate the next false claim rather than only reacting to the last one.

This analysis is done entirely from public sources online.

Choose one recurring misinformation theme.

Workable options include health remedies and cures, claims about government schemes or benefits, doctored or misattributed quotes, old images presented as current events, or manipulated statistics.

Collect at least ten documented examples of your chosen theme.

Published fact checks from established fact checking organisations are an excellent source, as they provide both the false claim and the verified position.

Record each example with its date, the exact claim, the format it took, where it circulated, and what the verified truth was.

Now analyse the collection for patterns.

Examine the format each example took, the emotional response it was designed to trigger, the kind of authority it falsely borrowed such as a doctor, an official body, or a news logo, the time at which it surged, and the specific gap in public knowledge it exploited.

Write an analytical report of about 1000 words presenting your findings.

Describe the pattern, support each observation with specific examples from your collection, and explain why this particular theme recurs.

Examine also why the theme persists despite repeated debunking.

Research what is known about why corrections often fail to reach the people who saw the original, and apply that to your specific theme.

Identify where in the spread cycle a correction would have to arrive to make any difference.

Then write a prediction section.

Based on your pattern analysis, describe the form you expect the next version of this claim to take and what event would trigger it.

This is the real value of pattern work, since anticipating a claim allows preparation rather than reaction.

Close with a practical section written for ordinary readers, listing the warning signs that should make someone pause before sharing a claim of this type.

Submit the example collection with sources, the analytical report, the persistence research, the prediction section, and the reader guidance.