Fact Checking AI Output Before It Leaves Your Desk
A practical routine for checking names, numbers, quotes and claims in AI assisted work before you send it.
A taste of a lesson
The AI wrote that 'over 70% of small businesses now use AI'. It sounds plausible. How do I check it fast?
Plausible is exactly when to be careful. First, search for the figure with a phrase like 'small businesses use AI survey' and look for the original survey, not articles repeating it. If you find it, check who ran it, when, how many businesses, which countries, and what 'use AI' meant. If you cannot find any source in a few minutes, cut the number or replace it with one you can cite. Try it now: what is the first source you find, and who published it?
Written by the teacher as an example. In your lesson the tutor answers your own questions, and like any AI it can be wrong.
What you will be able to do
- Extract the factual claims from any draft into a checkable list
- Prioritise checks by stakes when time is short
- Verify claims with primary sources and lateral reading
- Recognise patterns that suggest a claim or reference was invented
Lesson plan
- 1 Who owns the facts Accept that the sender is responsible for every claim in AI assisted work. Start
- 2 Pulling out the claims Turn a draft into a numbered list of checkable claims. Start
- 3 Primary sources first Verify important claims at their origin. Start
- 4 Lateral reading Judge sources by what others say about them, not by how they look. Start
- 5 Red flags of invention Recognise signs that a claim or reference may be fabricated. Start
- 6 A checklist for your team Build a short, shared verification routine. Start
Try asking
About this tutor
When AI helps you write a report, a briefing or a client email, you are still the one responsible for every fact in it. This tutor teaches a verification routine for busy people: pulling the factual claims out of a draft, deciding which ones matter most, checking them against primary sources, and reading sideways to judge whether a source is trustworthy. You practise on realistic drafts containing planted errors, learn the patterns that suggest a claim was invented, and build a short checklist for your team. Useful at any level, whether you are new to assistants or already rely on them daily.
Reviews
4.7
3 ratingsSample
- Kenji O.Sample
Lateral reading was new to me and very useful. Lessons are short, which suits a busy week.
- Nadia F.Sample
Our team adopted the checklist. The point that two models can agree on the same error stuck with everyone.
- Bethany C.Sample
The planted error drafts were humbling. I missed a wrong job title and a misquoted figure. Now I extract claims before every briefing.
About the teacher
Careful, practical use of AI assistants for everyday office work: email, meetings, files and team habits
9 tutors 332 lessons taught Sample
I help office teams use chat assistants for the work that fills a normal week: email, meeting notes, summaries, documents and translation. I spent a long stretch in administration and internal communications, so I know how much of a job is reading, condensing and replying. My teaching is hands on and cautious in equal measure. We practise on realistic tasks,...
See Chloe's profile and tutorsMore like this
Other tutors on the same or nearby topics.