62% of people now worry about what's real and fake online.

and trust in news just hit 37%, the lowest since 2015.

Both numbers are from the Reuters Institute's Digital News Report 2026, out in June. For student journalists that's actually an opening: verification is the skill everyone needs and few have practised. Here are the fact-checking tools worth learning, AI ones included, and how to use them without being fooled by them.

What's happening with fakes and fact-checking right now

The same report found:

→ 10% of people use AI chatbots for news every week, up from 7% → 16% of under-35s do → 52% of 18 to 24s say social media, video networks or chatbots are their main way of getting news

So your audience is increasingly meeting news through feeds and chatbots, which is exactly where fakes spread.

A real example from last week. On 24 July, Lead Stories debunked a viral clip claiming to show Donald Trump falling as he got into a car. How they did it:

  1. two AI detectors flagged it (one rated the clip 78% likely AI, another called the thumbnail "95% FAKE")
  2. Google News and Yahoo News searches found no credible outlet reporting the incident
  3. the account that posted it described itself as a humour account
  4. they archived the original post before it could vanish

Notice the detector was only ONE of four steps. That's the whole lesson.

The verification toolkit for student journalists

  1. Reverse image search. Google Lens and TinEye show where an image appeared before. Most "breaking" photos turn out to be old ones with new captions
  2. The InVID-WeVerify plugin. A free, open source browser extension built for journalists. It splits videos into keyframes (still frames you can reverse search), checks metadata and has a magnifier for zooming into details
  3. Watermark checks. You can now ask the Gemini app whether an image was made or edited with Google AI. It looks for SynthID, Google's invisible watermark, which has been added to over 20 billion pieces of content (Google). The catch: it only covers Google's own AI
  4. AI detectors (Hive, Sightengine and others). Useful as a clue, never a verdict. They give probabilities, not proof
  5. Archives. Save the original post on the Wayback Machine or archive.today before it's deleted, exactly like Lead Stories did
  6. Chatbots as research assistants. Fine for summarising background or suggesting where to look. Not fine as a source. We've seen what happens when lawyers treat them as one in our piece on hallucinated court citations

Why no single tool gives you a verdict

Even the best tools wobble.

In January The Intercept tested Gemini's watermark check on a doctored photo of an activist crying. It gave different answers on different attempts, at one point calling the image authentic.

bad workflow: run it through a detector → publish the score good workflow: detector + reverse search + source history + corroboration → then decide

A "no watermark found" result doesn't mean real. A "78% AI" result doesn't mean fake. Your job is the judgement in between.

What uni won't teach you about verification

→ journalism courses teach sourcing for interviews and documents, less for viral video → media law modules cover defamation, rarely how to prove a clip is fake → tools change monthly, and a syllabus written two years ago can't keep up

So you'll have to build this habit yourself (which, honestly, makes it a great thing to show off in applications).

The critical-thinking move: 5 questions before you share or publish

  1. Where did this first appear? Reverse search it and find the earliest version
  2. Who posted it, and what's their history? A humour account isn't a news source
  3. Who else is reporting it? No credible coverage of a big event is a red flag
  4. What do the tools disagree on? If detectors split, you don't know yet. Say so
  5. What would change my mind? Write down what evidence would prove you wrong, then go look for it

What to do this month

  1. Install InVID-WeVerify and run it on three viral videos from this week
  2. Debunk one thing yourself. Pick a claim going round your feed and write a 200-word fact-check using at least three methods
  3. Test a detector on purpose. Run a real photo you took and an AI image through the same tool, and note where it gets it wrong
  4. Read one fact-check a day from Lead Stories or Full Fact to learn the method, not just the verdict

Why does all this matter for your career? We argued in our piece on AI slop that a flood of fakes makes trusted journalism MORE valuable, not less. Verification is how you earn that trust. More in our journalism and media hub.

so TLDR: learn the tools, but never publish a detector score as a verdict. Stack the methods and make the call yourself.

What's the last viral clip you believed, that turned out to be fake?

Sources

Get the next issue by email

theVANTAGEd is written by students to bring critical thinking to AI across real industries. One issue every two weeks.