Checking AI Facts Before You Share Them

AI is confident. That's both its strength and its danger. It will give you a detailed, authoritative-sounding answer whether or not it actually knows the answer. It won't say "I'm not sure" the way a person would. It just generates something that sounds right.

This is called a hallucination. The AI isn't lying. It's not trying to deceive you. It's more like a very articulate person who fills in gaps in their knowledge with a plausible-sounding guess. And you need to know how to spot it and fix it before you share anything.

Why AI Gets Things Wrong

There are a few reasons:

Outdated information. Most AI models have a knowledge cutoff. ChatGPT knows things up to April 2024 (at the time of this writing). If you ask about something recent, it might not know or might guess wrong.

Confusion between similar things. AI can confuse facts if multiple things are similar. Ask it about a famous person and it might mix up details with someone who has a similar name or biography.

Made-up sources. The AI might cite a study or book that doesn't exist. It sounds real. It has a title, authors, a year. It's made up.

Interpretation vs. fact. AI sometimes presents opinions or interpretation as facts. It's not intentional. It's a limitation of how it was trained.

The One Rule: Always Check Important Facts

Here's the simple rule: If you're sharing it, verify it first. Especially if it's:

• A statistic or number
• A historical date or event
• A quote from someone
• Advice about health, finance, or law
• A claim about someone else's beliefs or actions
• Anything you're not 100% sure about

Quick fact-checking takes two minutes. Publishing something false takes longer to fix.

How to Check Facts Fast

Step 1: Use Perplexity for Recent Information

Perplexity searches the web and cites sources. If you ask it the same question you asked ChatGPT, it will give you the same answer plus the sources where it came from.

Compare the two:

• ChatGPT: "The first iPhone was released in June 2007 and started at $499."
• Perplexity: "The first iPhone was released on June 29, 2007 and started at $499 for the 4GB model and $599 for the 8GB model. Source: Apple press release, 2007."

Perplexity gives you sources you can check. If the source is real and reputable, you can trust the fact.

Step 2: Google It

For any fact you're unsure about, Google it. Spend 30 seconds checking. If multiple reliable sources say the same thing, it's probably true.

Good sources:

• Government websites (.gov)
• Educational institutions (.edu)
• Reputable news outlets (NPR, BBC, Reuters)
• Official company websites
• Subject-matter experts in peer-reviewed publications

Skippy sources:

• Random blogs
• Social media posts
• Wikipedia (useful starting point but not always accurate)
• Anything that quotes other sketchy sources

Step 3: Check the Quote

If the AI quoted someone, verify the quote is real and in context.

Copy and paste it into Google with quotation marks. If it's a real quote, you'll find the original. If you can't find it, it's probably made up.

Step 4: Check the Date

When did this happen? Is the AI's claim recent or old? If it's about a current event and the AI is giving you outdated information, that's wrong.

Example: "What is the current CEO of Twitter?" If ChatGPT says "Jack Dorsey," that's outdated. Use Perplexity for current facts.

Red Flags That AI Might Be Wrong

Look for these warning signs:

• The AI sounds VERY confident, almost too confident
• It cites a source but you can't find that source
• The numbers are suspiciously round ("exactly 50% of people") when real data is usually messier
• It presents opinion as fact ("Everyone knows that...")
• It mentions a study but gives no way to find it
• It contradicts what you've read from reputable sources
• The detail is oddly specific but you've never heard it before

The Three Categories of AI Output

Understand which category your information falls into:

Category 1: Likely Accurate

• Common knowledge (the Earth is round)
• Widely known facts (the first president of the US was George Washington)
• Information from the AI's training data that's old and stable (Shakespeare wrote Hamlet in the early 1600s)

Action: You can probably use this without extensive checking. But if you're publishing, still do a quick verify.

Category 2: Requires Checking

• Recent events (anything in the last few months)
• Specific statistics or numbers
• Quotes from people
• Information about individuals
• Medical, legal, or financial advice

Action: Verify with Perplexity or Google before sharing or acting on it.

Category 3: Should Not Be Trusted Alone

• Personal medical advice (get a doctor's opinion)
• Legal guidance (consult a lawyer)
• Financial investment advice (consult a financial advisor)
• Anything life-changing

Action: AI can give you information to understand the topic, but get professional verification before making big decisions.

The Fact-Checking Workflow

Here's your process:

Step 1: Ask the AI your question. Get the answer.

Step 2: Identify the claims. Highlight the specific facts the AI is making.

Step 3: Assess the category. Is this likely accurate, requires checking, or too important to trust to AI?

Step 4: Check if needed. Use Perplexity, Google, or consult an expert.

Step 5: Share confidently. Now you know you're sharing something true.

Real Example

You ask ChatGPT: "What was the impact of the COVID-19 pandemic on small business? Give me statistics."

ChatGPT responds: "45% of small businesses closed permanently during the pandemic. The average business lost 60% of revenue in 2020. Government support programs helped keep many afloat."

What you do: These are specific statistics. You need to check them.

You ask Perplexity: "What percentage of small businesses closed during COVID-19?"

Perplexity responds: "According to census data, approximately 3.3 million small businesses closed during the pandemic (2020-2021). This represents a higher rate of closure than normal years. Source: U.S. Census Bureau, 2021."

Now you compare: ChatGPT said 45%. Perplexity cites census data. The Perplexity answer has a source you can check. That's more trustworthy.

You Google it: You find news articles from reputable outlets confirming a higher-than-usual closure rate during COVID, though the exact percentage varies. The overall trend is confirmed.

You now know: The AI's point (small businesses were hit hard) is true, but the specific statistic (45%) might not be exactly right. You can say "Many small businesses closed" and point to the census source, rather than the 45% number.

When You Find an AI Error

Don't just move on. Tell the AI: "Actually, that's incorrect. According to [source], the correct fact is [correct fact]."

The AI will adjust and give you better information going forward. This helps you and makes the next person's experience better.

The Habit

Make fact-checking automatic. Every time you get an answer from AI and you're thinking about sharing it, spend 30 seconds verifying. Check one source. Make sure it matches.

This one habit prevents embarrassment, builds trust with your audience, and keeps you from spreading misinformation.

It's not because AI is evil or dumb. It's because you're responsible for what you share. And that responsibility is worth 30 seconds of checking.