How to Fact-Check 'Record High' Headlines With AI (Nominal vs. Real)

A movie that grosses $900 million today can still be a smaller hit than one that made$600 million in 1997. Run both through an inflation calculator and the order can flip. That one piece of arithmetic quietly breaks a huge share of the "record high" headlines you will read this year: record box office, record home prices, record holiday spending. Most are technically accurate, because prices rise every year and any statistic measured in dollars sets records on autopilot. The headline is not lying; it is letting you assume "record" means "unprecedented in a meaningful way." Checking that assumption used to take a spreadsheet and an afternoon. With a chatbot pointing you at the right check and two free government tools doing the verifying, it takes minutes.

The three tricks behind most 'records'

Nominal versus real. Nominal is the raw dollar figure; real is inflation-adjusted so years are comparable. That is the whole movie trick from the top of this post. Any dollar record that does not mention inflation adjustment should be treated as unverified.

The base effect. Growth rates depend on the starting point. If last year was a slump, this year's "surge" may just be recovery to normal. The classic case: almost every economic statistic posted "record growth" in 2021, measured against locked-down 2020.

The convenient window. "Highest since 2020" means one specific thing: it was higher before 2020. Watch for the word "since"; it is doing heavy lifting.

The 3-step routine

Steps 1 and 2 use this prompt. Paste the headline or the claim sentence:

Fact-check the "record high" claim below. Do NOT invent any numbers.
If something cannot be verified from the text alone, write "needs source check."

[paste the claim]

1. What statistic is this, which agency produces it, and where is the
   original series (FRED, BLS, Census)?
2. Is the figure nominal or inflation-adjusted? If unstated, which is
   more likely and why does it matter here?
3. Was the comparison base (last year / the cited start year) unusually
   low or high?
4. What exact window does "record" cover? Is "since [year]" hiding
   higher readings before that?
5. Verdict in one line: solid record, misleading framing, or cannot
   tell without the source.

Feed it a made-up example headline, "Holiday retail sales hit all-time high," and the answer comes back in this shape (illustrative excerpt):

Seeing "needs source check" repeatedly is a feature, not a failure. It is the AI declining to invent what it does not know, and it hands you a short list of exactly what to verify.

Step 3: the two free tools that settle it

The BLS inflation calculator (bls.gov/data/inflation_calculator.htm) converts a dollar amount from one year into another year's dollars using CPI. Thirty seconds, no account. This single tool kills or confirms most "record dollar amount" claims.

FRED (fred.stlouisfed.org), run by the Federal Reserve Bank of St. Louis, is the free charting front-end for hundreds of thousands of official US data series. Search the statistic, chart the full history rather than the headline's window, and check whether an inflation-adjusted ("real") version of the series exists. If the real series also sits at a peak, the record is genuine. That is the whole test.

FAQ

Why are 'record high' numbers in the news often misleading?

Three common reasons. The figure is nominal, meaning not adjusted for inflation, so almost any dollar amount sets records over time. The comparison base was unusually low, which inflates the growth rate (the base effect). Or the window was cherry-picked, like "highest since 2020," which quietly excludes higher readings before that. None of these require anyone to lie; they just require you to check the basis.

What is the difference between nominal and real values?

Nominal is the dollar amount as recorded at the time. Real is that amount adjusted for inflation so different years are comparable. Box office revenue, home prices, and wages all set nominal records almost by default because prices rise. A record only means something if it holds in real terms, which you can check in seconds with the BLS inflation calculator.

Can I trust AI to fact-check statistics?

Use it as the first pass, never the last. Chatbots invent plausible-sounding numbers and sources, and they can serve stale data as current. The reliable pattern is to make the AI identify what to verify (nominal or real, the base year, the window) and then confirm the actual numbers yourself on FRED or the original agency release.

Where do I check the original US numbers for free?

Two places cover most cases. FRED (fred.stlouisfed.org), run by the St. Louis Fed, charts hundreds of thousands of official data series and often has an inflation-adjusted version of the same series. The BLS inflation calculator (bls.gov) converts any dollar amount between years using CPI. Both are free and require no account.

My honest take

I run this reflex on every "record" headline now, and the usual outcome is anticlimax: it is nominal, or the base year was a crater, and the story is smaller than the font size. Once in a while the real series is genuinely at a peak, and that is when I actually pay attention. Either way, the win is the same: I start from what the data says instead of what the headline wanted me to feel. Fact-checking is not about dunking on journalists. It is about not being the person who repeats the wrong number at dinner.