What is confirmation bias, and how does it shape the news you pick?

Summary: Confirmation bias is the habit of seeking, trusting, and remembering information that fits what you already believe, while checking harder on information that doesn't. In the news, it decides which sources you trust and which headlines you open before you've read a word. A 2009 meta-analysis measured the pull toward agreeable information at a moderate effect size (d = 0.36), and the same research found the one motive that reverses it.

The view arrives before the article does

We like to picture ourselves as readers who take in the news and then decide what to make of it. The order usually runs the other way. The opinion is already in the room when the tab opens, and the reading mostly confirms what was there.

Confirmation bias works at three points, and none of them feels like bias from the inside. At selection, you reach for the source or headline that sounds right. At evaluation, you accept the agreeable claim on sight and audit the disagreeable one line by line. At memory, you keep the piece that fit and lose the piece that didn't. Each step feels like good judgment, because each step uses the same standard: Does this match what I already know?

That's why more knowledge rarely cures the bias: A sharper reader finds sharper reasons to dismiss the inconvenient report.

What the research measured

In 2009, a team led by William Hart and Dolores Albarracín pooled decades of selective exposure experiments in a meta-analysis published in Psychological Bulletin. Across the studies, people chose information that agreed with them over information that challenged them with a moderate effect size, d = 0.36. A moderate effect means the tilt shows up reliably across people and topics without being the only thing that decides what they read.

The same analysis found what moves the dial. The pull toward agreeable material ran stronger when the belief was tied to a person's values and held with conviction, and when the person scored higher on closed-mindedness. It weakened when the person felt confident in the belief (a settled mind can afford a challenge), when the available information was low quality, and when the person wanted an accurate answer for a decision in front of them. In that last case the preference reversed: People sought out the inconvenient material because they needed it to be right. That reversal is the one lever in the analysis that points the other way.

What confirmation bias looks like in the news

The clearest footprint sits in trust, because trust is selection done in advance. Once you've decided a source is reliable, you don't have to judge its articles one at a time; The judgment of the source does the sorting.

In a survey of 12,043 U.S. adults in late 2019, the Pew Research Center asked about trust in 30 news sources. Republicans and Republican leaners distrusted more than they trusted 20 of the 30. Democrats and Democratic leaners trusted more than they distrusted 22 of the 30. Thirteen sources earned trust from at least 33% of Democrats; Two earned that from at least 33% of Republicans. Not one of the 30 was trusted by more than half of all adults. The widest gap sat at a single national newspaper, trusted by 66% of liberal Democrats and 10% of conservative Republicans, a 56 point spread over the same pages. Compared with Pew's 2014 survey, Republicans' distrust had grown for 15 of the 20 sources asked about in both years, while Democrats' trust and distrust moved far less.

Read those numbers as a map of confirmation bias at work on both sides, from different starting points. Each group has a short list of trusted sources and a long list of distrusted ones, and the lists barely overlap. The same article, carrying the same facts, gets read as reporting by one group and as spin by the other.

The pattern is global. In the Reuters Institute's 2025 Digital News Report, covering 48 markets, overall trust in news sat at 40%, unchanged for three years, and 58% of people said they worry about telling true from false online. In the United States that worry reached 73%. We'd bet low trust and high worry push readers toward the sources that already feel safe, which is the bias feeding itself.

Sometimes the side you trust is right

Trusting sources that agree with you isn't always bias. Some sources have earned trust with years of accurate reporting, and some have earned distrust the same way. A reader who has watched a source get things right and keeps reading it is exercising judgment, and the reader who drops a source after repeated errors is doing the same.

We concede the hard part: Confirmation bias and good judgment produce the same reading list, and from the inside you can't tell which one built yours. The feeling of having judged carefully is identical in both cases. No outside referee is coming to score your trusts for you.

The research does offer one honest test. Hart's analysis found that the bias reversed when people needed an accurate answer for a goal in front of them. So the question to ask of your own reading is whether you're reading to be right or reading to feel right. The first gives you a reason to open the report you expect to dislike. The second gives you a reason to close it.

A reading method that works against the grain

The method below works with a newspaper, a radio bulletin, a feed reader, or a stack of printouts. None of it needs a product.

1. Decide the question before you pick the source. Write it as a sentence you could answer with a fact, example: "How many people did the policy affect last year?" A question gives you an accuracy goal, which is the one motive the research found that reverses the bias.

2. Read the article you expect to dislike first, and read it to summarize, not to rebut. Then say its main claim in a sentence the author would accept. If you can't, you haven't read it yet.

3. Separate the fact layer from the frame. What happened, said by whom, with what numbers, is the fact layer. What it means is the frame. The guide to media bias lays out seven kinds of frame; Agreement between sources should happen on the fact layer, and that is where to settle your view.

4. Keep a short list of claims you changed your mind on in the last year. An empty list after a year of daily reading is information about the reader, not about the world.

5. Weight sources by whether they show their sourcing. Our post on unbiased sources gives the checks: Named sources, linked documents, and corrections printed in the open.

Confirmation bias has a sibling, negativity bias, which keeps the bad news and drops the good. Together they build a reader who is sure of the worst. The method above corrects for the first; Reading to the fact layer corrects for a good part of the second.

Where a plain brief fits

The method is the reason we built informed.now the way we did: A short daily news brief, delivered by text, written to the fact layer, with no feed learning what you agree with and serving more of it. Every reader gets the same facts, and what each reader makes of them stays their own. It's one illustration; A front page read in full, including the pieces you'd rather skip, does a version of the same job.

The view will keep arriving first

You can't stop the opinion from showing up before the article. Decades of pooled experiments say it will, at d = 0.36, in you and in us. What you can do is notice the view standing on the doorstep, name the question you actually need answered, and read the inconvenient report anyway. The reader who makes a habit of that ends up with fewer certainties than before, and better supported ones.