🧠AI Foundations
Your progress0%
0 of 49 lessons
Reading12 min·Lesson 2 of 5

Verifying AI Output — Practical Methods

Knowing that AI can be wrong is one thing. Knowing exactly how to check — quickly and systematically — is the practical skill that makes you someone others trust with information.

Why Verification Is a Professional Skill

In Kenya's job market, employers increasingly value people who can think critically about information. Whether you are working in banking at Equity, managing communications at a NGO, or running your own business, your reputation depends on the accuracy of what you present. An AI-generated error that you repeat as fact can damage that reputation quickly. The good news: verification does not have to take long if you have a system.

🔎
The SIFT Method — a widely-taught verification framework: Stop before you share. Investigate the source. Find better coverage. Trace claims to their original context. This method was designed for social media misinformation but works equally well for AI output.

Method 1: Lateral Reading

Lateral reading means opening new browser tabs to check what others say about the same claim — rather than reading deeper into the same source. Professional fact-checkers at organisations like Africa Check use this constantly.

When an AI gives you a statistic — say, 'Kenya has 22 million mobile money users' — do not just accept it. Open a new tab and search: 'Kenya mobile money users 2024 source.' Look for the Communications Authority of Kenya annual report, a World Bank data point, or a credible news article. If multiple independent sources agree, the claim is likely solid. If you cannot find the source, treat the number with caution.

⚠️
Watch for 'hallucinations.' AI models sometimes generate citations that look real but do not exist — a journal article with a plausible title and author that was never published. If an AI gives you a specific citation, paste the title into Google Scholar or the publisher's website and confirm it actually exists before using it.

Method 2: Reverse Image and Quote Search

Sometimes AI outputs include specific quotes attributed to real people — politicians, business leaders, academics. Before repeating any quote, paste it into a Google search with quotation marks around it. If the quote is real, you will find the original interview, speech, or article. If nothing comes up, the AI may have fabricated the wording.

For images (especially in AI-generated content you find online), Google's reverse image search or TinEye can show you where an image first appeared and whether it is being used in a misleading context.

Method 3: Check the Date

This is the most overlooked step. AI training data has a cutoff, and the world changes fast. M-Pesa transaction limits, HELB loan amounts, KRA tax brackets, fuel levy rates — these change regularly. Whenever you use AI to research anything with a number or a policy detail, always check when that information was last updated. Look for the publication date on any article you use as verification.

📅
Practical habit: When you find a source that confirms an AI claim, note the date it was published. If it is more than 12 months old for a fast-changing topic (finance, technology, government policy), keep searching for something more recent.

Method 4: Use Trusted Kenyan and African Sources

Not all verification needs to involve international databases. Kenya has strong primary sources you should bookmark:

  • Kenya National Bureau of Statistics (KNBS) — population, economic, and employment data
  • Communications Authority of Kenya (CA) — telecoms and internet statistics
  • Central Bank of Kenya (CBK) — financial sector data, M-Pesa statistics
  • Africa Check (africacheck.org) — independent fact-checking of claims circulating in African media
  • Nation and Standard Digital — for recent Kenyan news and context
  • Kenya Law (kenyalaw.org) — for legislation and court decisions

Putting It Together: A Quick Verification Checklist

Claim type: Statistic or number
Find the original data source (government report, research institution). Check the date. Confirm the number matches.
Claim type: Quote from a real person
Search the exact quote in quotation marks. Find the original speech, article, or interview. Check context.
Claim type: Academic citation
Search the exact title on Google Scholar or the publisher site. Confirm the paper exists and says what the AI claims.
Claim type: Current policy or law
Check kenyalaw.org or the relevant government ministry website. Confirm the policy is current and not amended.
Remember: Verification is not about distrust — it is about professional standards. The most respected researchers, journalists, and analysts in Kenya verify everything before presenting it. This habit will set you apart in any workplace.

Next, we look at a harder challenge: deliberate misinformation and deepfakes — where the false content is intentional, and the techniques for spotting it require a slightly different approach.