How to verify information and detect misinformation with AI

Before using AI to verify, understand its limits:

• Hallucinations: Claude, ChatGPT and other models sometimes invent data, dates, or quotes that sound real but aren’t.
• Knowledge cutoff: most have a knowledge limit (for example, August 2025). For recent news, they can’t be trusted alone.
• No real-time browsing (unless search is enabled, like Perplexity or ChatGPT with browsing).

Golden rule: use AI to flag suspicious claims and suggest what to look up, not to confirm facts directly.

Perplexity AI is an AI search engine that shows the source for each part of its answer, making it ideal for quick fact-checking.

1. Go to https://perplexity.ai (create a free account or use it without one).
2. Type the claim you want to verify as a direct question:

Is it true that [claim]? Give me the sources.

3. Perplexity gives you an answer with numbered citations. Click each citation to go to the original source.
4. If the source is a recognized outlet or official institution, the claim has backing. If it’s an anonymous blog or social media post, investigate further.

Tip: enable “Focus: Academic” to search only academic papers, or “Focus: News” for recent news. Confirm prices and features on the official Perplexity page.

To go deeper on using Perplexity, check the tutorial How to use Perplexity AI to search the internet.

Claude is excellent at reading a long text and flagging which parts need verification. You don’t have to read it all; just ask Claude to do it for you.

1. Copy the text (article, WhatsApp message, social post) you want to analyze.
2. Go to https://claude.ai and paste the text with this prompt:

Analyze this text as a fact-checker. Flag:
1. Which claims need a source to be verified.
2. Which data points seem exaggerated or hard to confirm.
3. Whether the tone is objective or appears biased.
Be specific and quote the exact phrases from the text.

[Paste the text here]

3. Claude returns an analysis with suspicious parts highlighted.
4. Use those hints to search on Google, Perplexity, or the original source.

Remember: Claude can be wrong. If it says something is false, verify it yourself with a primary source.

You can also analyze PDF documents with this method: How to analyze PDFs and long documents with AI.

Fake or out-of-context images are one of the most common forms of misinformation. These tools help you trace their origin:

TinEye (https://tineye.com):

1. Go to TinEye and upload the image or paste its URL.
2. TinEye shows you where that image appeared before. If the “new” image shows up in results from years ago, it’s likely being reused out of context.

Google Images:

1. In Google, click the camera icon in the search bar.
2. Upload the image or paste the URL.
3. Google shows you pages where that image appears.

How to interpret results:

• The image appears in a different context than what’s being shared: likely misinformation.
• The image is new and appears only on one site: research that site before sharing.
• The image appears in recognized outlets with the same description: a sign it’s legitimate.

Confirm availability and free limits on the official TinEye and Google Images pages.

The SIFT method is a 4-step protocol created by researchers to verify information quickly, without special tools. Apply it before sharing anything on social media or group chats:

S — Stop: Before sharing, pause. Do you feel urgency or a strong emotion (fear, outrage, exaggerated joy)? That’s a warning sign: misinformation is designed to trigger those reactions.

I — Investigate the source: Who is writing this? Is it an outlet with a track record, an official institution, or an anonymous profile? Search the outlet or author name on Google to see their reputation. Wikipedia has a criticism section that helps.

F — Find better coverage: Are other recognized outlets reporting the same thing? If only one site has the story and no serious outlet confirms it, be skeptical. Use Perplexity or Google News to search.

T — Trace claims: Where does the original claim come from? Many viral posts misquote a study or take a quote out of context. Find the original study or press release and read it yourself.

The SIFT method doesn’t need AI, but you can combine it with the previous steps for a deeper analysis.

Copy these prompts directly into Claude or ChatGPT to verify information:

To analyze text:

Act as a professional fact-checker. Read the following text and flag:
- Claims that need a source.
- Data points that seem inaccurate or exaggerated.
- Signs of bias or emotional manipulation.
Quote the exact phrases from the text.

To verify a news story:

Search whether this news story is real: [describe the news].
Which recognized outlets confirm it?
Is there any official source that supports or contradicts it?

To evaluate a source:

What do you know about the outlet [outlet name]?
Does it have a history of errors or misinformation?
Is it considered reliable by fact-checking organizations?

These prompts work in Claude and ChatGPT. Results may vary; always cross-check with external sources.

Problem	Fix
Claude says something is false but I’m not sure	Search Google News or Perplexity with the same claim and compare
Perplexity doesn’t find recent results	Switch Focus to “News” or search directly on Google News
TinEye doesn’t find the image	Try Google Images or Bing’s reverse image search tool
The text is too long for Claude	Paste it in parts or ask it to focus only on the first three paragraphs
I don’t know if a source is reliable	Search the outlet name on Wikipedia or AllSides.com to see its bias rating