Free lesson

Evaluating sources: how to check a claim before you repeat it

Evaluating sources means checking who is behind a claim, how current it is, what evidence supports it and what bias might shape it, then confirming it elsewhere before you use it. This lesson applies those checks to a statistic an AI assistant handed you, because AI citations need the same scrutiny as any web page, and sometimes more.

For you or your whole team

Start with a useful first step.

7-day free trial. No credit card needed.

You will continue on the Tosheo.ai app to create your account and review the trial terms.

What this lesson builds

Skills worth putting to work.

Lateral reading

Checking a publisher or author from independent sources rather than their own pages.

Date and method checks

Finding when data was collected and how, not just when it was posted.

Bias awareness

Asking who benefits from a claim and seeking confirmation from neutral sources.

Tracing to the original

Following a claim back to the primary source and the exact sentence.

AI citation checking

Confirming that a cited source exists and says what the AI claims.

Example task / fictional source material

Check the Lumen Foods statistic

Ask an approved AI tool for a statistic on a topic in your own field, with sources. Apply the six steps and the three AI citation checks to one claim. Write a short verdict: use as is, use with changes, or discard, with your evidence.

Review criterion

Existence checked

The cited source is searched independently, not only through the AI's link.

Review criterion

Lateral reading

At least one independent source about the publisher is consulted.

Review criterion

Date and method

The fieldwork date, sample and method are stated or marked as missing.

Review criterion

Clear verdict

The decision follows from the evidence and the final wording matches what the source says.

Copy and adapt

Prompts to try

Replace the parts in square brackets with your own details, and check every output before you use it.

Claim breakdown

Break this paragraph into individual factual claims. For each, say what kind of source would be needed to confirm it, and what search I should run to find that source. Paragraph: [paragraph]

Citation quote check

For the claim [claim], you cited [source]. Quote the exact sentence from that source that supports it and say where in the document it appears. If you cannot quote it, say so plainly.

Source questions

I am evaluating a source published by [publisher] on [topic]. List the questions I should ask about its authority, date, evidence and possible bias, and what independent searches would help answer each.

Why do AI answers make evaluating sources more important?

AI answers make evaluating sources more important because they present claims fluently and with references that look authoritative, even when the reference does not exist or does not say what the answer claims. A confident tone is not evidence.

Picture a fictional 15-person snack brand, Lumen Foods. The marketing lead, Ade, needs a headline statistic for a retailer pitch. He asks an AI assistant and gets: '68% of UK shoppers now choose low-sugar snacks (Global Snack Trends Report 2024, Retail Insight Institute).' A second search turns up a blog post repeating '7 in 10 shoppers prefer low sugar', with no link. The pitch is on Friday, and both numbers look usable.

How do you evaluate a source step by step?

Evaluate a source by stopping before you use it, checking who is behind it from outside the source, looking for better coverage of the same claim, and tracing the claim to its original. Library guides teach this as the SIFT method, created by digital literacy expert Mike Caulfield: Stop, Investigate the source, Find better coverage, Trace claims to the original context.

1. Stop

Notice when a claim is exactly what you hoped to find. That is when you are least likely to check it.

2. Check authority by reading laterally

Leave the source and search for what others say about the publisher or author. Lateral reading, opening new tabs to see what independent sources say, tells you more than an About page written by the source itself.

3. Check the date

Find when the data was collected, not just when the article was posted. A 2024 article may quote 2017 figures.

4. Check the evidence

Look for sample size, who was asked, the exact question and the method. A statistic without a method is a rumour with a number attached.

5. Check for bias

Ask who benefits if you believe the claim. A low-sugar statistic published by a low-sugar brand needs extra confirmation.

6. Trace to the original

Follow citations back until you reach the primary source, and read the sentence the number came from.

How do you check citations from an AI tool?

Check AI citations in three steps: confirm the source exists, confirm it says what the AI claims, and confirm it is the original rather than a summary of a summary. Library AI guides warn that AI tools can fabricate information, including references that look real.

Search for the exact report title and publisher yourself rather than clicking only the link the AI provides. If you find it, use the search function in the document to locate the figure. If the tool can browse or search the web, ask it to quote the exact sentence and page, then compare that quote with the document. If you cannot find the source, treat the claim as unverified, however specific it looks.

What does weak and careful source evaluation look like?

Weak approach: Ade puts '68% of UK shoppers choose low-sugar snacks' on slide two with the AI's citation in small print. In the meeting, the retailer's buyer asks for the report. Ade searches and finds no organisation called the Retail Insight Institute and no report by that title. The credibility of the whole pitch drops.

Careful approach: Ade searches the report title and finds nothing, so he discards the 68% figure. He traces the '7 in 10' blog claim and finds it came from a survey published by a sweetener company, with a small online panel and a leading question. He then looks for better coverage and finds a government food survey with a stated method, which reports a more modest trend and dates its fieldwork clearly. His slide now says what that survey shows, names the publisher and year, and links the table. The claim is smaller but it survives the buyer's question.

What are the common mistakes when evaluating sources?

The most common mistake is judging a source by how professional it looks, which is easy to fake. Others: trusting an AI citation because it has a title and year, checking only the source's own About page, using a recent article that quotes old data, stopping at a secondary report instead of finding the original, and treating many blogs repeating one number as independent confirmation when they all copy the same origin.

Where does evaluating sources fit in the Tosheo.ai catalogue?

Evaluating sources sits in the Critical Thinking category, in the critical thinking course, next to lessons on spotting weak arguments and checking AI output. On Tosheo.ai, the hands-on task gives you a claim with a mix of real and fabricated citations, and the rubric feedback shows which checks you skipped and what they would have revealed.

Sources and scope

Digital Inquiry Group, Civic Online Reasoning, teaching lateral reading (checked 19 October 2026) ↗

Scottsdale Community College Library, the SIFT method (checked 19 October 2026) ↗

University of Montevallo Library, fact-checking AI output and citations (checked 19 October 2026) ↗

Before you decide

A few good questions.

What are the main criteria for evaluating sources?

The main criteria are authority (who is behind it), date (when the information was gathered), evidence (what supports the claim and how it was collected) and bias (who benefits if you believe it). Confirm important claims with an independent source.

What is lateral reading?

Lateral reading means leaving a source and searching for what independent sources say about it and its publisher. It is usually a faster and more reliable judge of credibility than reading the source's own description of itself.

What is the SIFT method?

SIFT is a four-step method created by Mike Caulfield: Stop, Investigate the source, Find better coverage, and Trace claims, quotes and media to the original context. Many university libraries teach it.

Can I trust the sources an AI chatbot gives me?

Not without checking. AI tools can produce references that do not exist or that do not say what the answer claims. Search for each source yourself and find the exact sentence before you rely on it.

Is a source reliable if many websites repeat it?

Not necessarily. Many pages often copy one original. Trace the claim back to where it first appeared and judge that original source.

Give your team a path, not another playlist.

Explore role-specific lessons, practical tasks and the manager view.

Start free trial7-day free trial · No credit card needed

Evaluating sources: how to check a claim before you repeat it

Start with a useful first step.

Skills worth putting to work.

Lateral reading

Date and method checks

Bias awareness

Tracing to the original

AI citation checking

Check the Lumen Foods statistic

Existence checked

Lateral reading

Date and method

Clear verdict

Prompts to try

Claim breakdown

Citation quote check

Source questions

Why do AI answers make evaluating sources more important?

How do you evaluate a source step by step?

1. Stop

2. Check authority by reading laterally

3. Check the date

4. Check the evidence

5. Check for bias

6. Trace to the original

How do you check citations from an AI tool?

What does weak and careful source evaluation look like?

What are the common mistakes when evaluating sources?

Where does evaluating sources fit in the Tosheo.ai catalogue?

Sources and scope

A few good questions.

Keep exploring.

Critical thinking course for working professionals

Why AI tools invent facts, and what to do

AI for research at work

Lesson: correlation versus causation

All free AI lessons

Give your team a path, not another playlist.