Skip to content
Back to all insights

When AI makes things up: 5 ways to fact-check AI-generated content

AI once told me that 15–20 percent of the facts it generates are incorrect. When I asked for a source, it admitted to making the number up entirely. This exact problem threatens the quality of content creation, but it is an easy one to solve.

Written by

Paula-Maija Wallin

Published

March 27, 2026

03.png

Are there 1, 18, or 38 lakes named Pyhäjärvi in Finland? The last number is the correct one (I checked), but AI has confidently offered all three as the answer.

AI hallucinates. It invents facts that sound plausible but are entirely fabricated. This happens because language models don't actually "know" things; they simply predict the most likely next word. This can lead to false answers, fake sources, and events that never happened.

The responsibility for content quality always rests with the human. However, when used correctly, AI is an exceptional assistant.

Five easy ways to fact-check AI content

Because AI doesn't understand what is true—it only seeks linguistically logical answers—you must approach its output with a critical eye. Be especially vigilant with facts and sources, particularly dates and numbers.

You should always verify the accuracy of AI-generated content. Here are five easy ways to do it:

1. Fact-check against reliable sources

Before presenting AI-generated information as fact in your own content, run a separate search for any claims, technical terms, and definitions. Rely on official websites, established media outlets, and research papers, and ideally cross-reference multiple sources.

You can also ask most AI tools to provide links to back up their answers or to explicitly state if they are uncertain.

2. Verify the origins of statistics

Always cast a critical eye on percentages, dates, years, and numbers provided by AI. Track down the original source—such as a research report or a national statistics database—and ensure the figure actually makes sense in context.

3. Double-check personal details

Verify people's names, titles, work histories, and achievements using LinkedIn profiles or company websites. Exercise extra caution when dealing with lesser-known individuals.

4. Scrutinize quotes and citations

Use a search engine to track down alleged quotes and check their original context. If a quote sounds a little too perfect, it is probably a hallucination. Stay sharp.

When processing your own interview transcripts, always double-check the quotes or explicitly instruct the AI to use the verbatim text. Otherwise, the AI will often take artistic liberties, combining and altering quotes in ways that break the context and distort the original meaning.

5. Trust your common sense and expertise

When working with AI, lean on your own professional judgment to assess whether the content makes sense. If something seems off, it probably isn't true. Compare the claims against your existing knowledge and consult industry experts if you are unsure.

Remember: AI is a tool, not a source of truth

AI-assisted content creation is highly effective—as long as you remember its limitations. Don't trust it blindly. Verify, question, and rely on your own expertise.