Skip to content
GuideEvergreen guide

Read the press release from the bottom up

Every release is written by someone paid to make you believe one thing. Five checks show you what got left out.

By Marcus Vale · Editor-in-Chief5 min read
Read the press release from the bottom up

Quick answer: A press release is an argument dressed as an announcement. Read the last paragraph first, ask "out of what?" about every number, and look hard at the verbs. Then ask what a skeptical reporter would want to know that the release doesn't say.

Start at the bottom

Most people read a release top to bottom, which is the order the author designed. Flip it.

The last paragraph is usually the "About" boilerplate. It tells you who's talking: a think tank, an advocacy group, a drug maker, a police department, a city hall, a startup. Check the line that says who funded or commissioned the work. Check whether the "institute" has a product, a donor list or a lobbying arm.

This doesn't make the release false. A company that studies its own product may have found something real. It tells you what the sender wants from you, and that's the filter for everything above it.

The method is the same whether the sender is a gun-rights group, a gun-control group, a federal agency or a cereal company. Nobody gets a pass, and nobody gets treated as a villain for having a point of view. Everyone with a press office is selling something, if only their own competence.

Ask "out of what?"

Numbers in releases come without denominators. Find the denominator.

Take a made-up example. A company announces that complaints about its product "rose 100 percent." If that's two complaints going to four across a million units sold, the number is true and tells you nothing. Flip it. A company announces a "50 percent drop in incidents." If the count fell from four to two, same problem.

Three versions of this trick come up constantly:

  • Relative versus absolute. A risk that goes from 2 in 10,000 to 3 in 10,000 is a "50 percent increase." It's also one extra case per 10,000 people. Ask for both numbers.
  • The convenient baseline. "Lowest in a decade" depends on which year the decade starts. Move the start date two years and the claim can vanish. Look for the chart, or ask what the number was before the chosen window.
  • "Up to" and "as much as." "Saves you up to 40 percent" is a best case. The typical case is somewhere below it and the release won't say where.

If a release gives a percentage and no raw count, that's a flag. Sometimes the writer just got lazy. Often the raw count is small and unimpressive.

Read the verbs

The nouns and numbers get the attention. The verbs carry the weasel room.

"Linked to," "associated with" and "could" describe a correlation or a possibility. They don't describe a cause. "Supports," "highlights" and "underscores" mean the sender found a piece of data that fits the story already being told.

Then there's the announcement verb group: "launches," "commits to," "pledges," "unveils," "will invest." A pledge costs nothing on the day it's made. Ask three questions:

  1. Is the money new, or was it already budgeted and relabeled?
  2. Over how many years? "$50 million" over ten years is $5 million a year.
  3. Who checks whether it happens?

Governments do this all the time. So do corporations and nonprofits. Big institutions of every kind love a ribbon-cutting for something not yet built.

Compare that to "completed," "delivered," "reduced" and "paid." Those are verbs about things that already happened, and they can be checked.

Quotes are decoration

Every release has a quote from an executive or director saying they're "thrilled," "proud" or "deeply committed." That's not evidence. It's a line written by the communications team and approved by the person named.

What matters is who isn't quoted. If every voice in the release belongs to the sender or a partner of the sender, you've heard one side. An independent outside expert, someone with nothing to gain, is a sign the sender wasn't afraid of a second opinion. Their absence isn't proof of anything, but it tells you how much work is left to you.

The same logic is behind the questions an editor asks before a story runs. A good editor wants to know who said it, how they know and who disagrees. A release is built to avoid exactly those questions.

Find what's missing

This is the hard part, because you're looking for something that isn't on the page.

If the release cites a poll, find the exact wording of the question, the sample size, who was surveyed and the margin of error. If the question wording is missing, assume it helped the sender. A survey of "members" or "supporters" isn't a survey of the public. If it cites a study, find out whether the study is published, who paid for it and what the sample was. "A new study finds" with no link is a paragraph, not a source.

Timing is worth a look too. Bad news tends to go out late on a Friday, when fewer people are watching. Good news gets a Tuesday morning. A release that lands at 5:40 p.m. before a holiday weekend is telling you something about how its author rates it.

Then compare the release to the news story built on it. A lot of coverage is the release with the adjectives rewritten and a phone call added at the end. If the story has the same numbers, the same quotes and the same framing, the outlet did transcription, not reporting. It's worth knowing which outlets own their mistakes and which ones just pass along whatever arrives in the inbox.

And where a release leans on unnamed people, the rules for anonymous sources apply here too.

The ten-minute drill

Next time a release lands in front of you, run this:

  • Read the "About" paragraph. Who's the sender and who pays them?
  • Circle every number. Write "out of what?" next to each one.
  • Underline the verbs. Mark which are about the past and which are about the future.
  • Count the voices. How many are independent of the sender?
  • Write the one question you'd ask if you could call the author. Then look for the answer yourself: the study, the dataset, the budget line.

That's it. Ten minutes, a pencil and a little suspicion that has nothing to do with cynicism. Most releases survive the drill in part. A claim can be partly true, and the job is to sort the part from the rest.

One more rule, and it's the one people skip. Run the drill first on the releases you want to be true. Anyone can be sharp about the other side's spin. The test of the method is whether you'll use it on your own team's.

Marcus Vale

Editor-in-Chief

Twenty years in magazines, most of it deciding what to cut. Writes about work, discipline and the decisions that compound.

Watch

Worth an hour of your evening

More from Media Literacy.

From channels we rate. Plays on YouTube.

    EdutopiaMay 28

    A Media Literacy Tool to Assess News Credibility

    New ScientistSep 2

    How to spot deepfakes and AI-generated images

    LifehackerJan 31

    How to Spot a Deepfake Video | Lifehacker

    NOVA PBS OfficialOct 12

    How to Detect Deepfakes: The Science of Recognizing AI Generated Content