How much of the news is charts

I went in believing that news front pages had become chart-led. Every big story seemed to arrive with a line going up or a map going red. So I measured it, badly but honestly: 31 sites, the first twelve thumbnails each, one afternoon. Two of the 334 images were charts.

What I actually did

On 5 September 2026 I fetched the homepage HTML of 35 English-language news and business sites with curl and a Safari user agent. From each page I pulled the og:image and the first content-looking image URLs in article or teaser markup, capped at twelve, and downloaded them. Then I looked at every thumbnail and labelled it by hand as chart, photo, illustration, or other. Four sites blocked the fetch (Washington Post, Reuters, Bloomberg, Politico). Three more returned fewer than twelve usable images (CNBC 3, Axios 4, The Economist 6).

The raw counts are the numbers below. The classification is mine, one pass, no second rater.

The thesis did not survive contact

I thought I would find a split: business desks heavy on charts, general news heavy on faces. Instead the business group (FT, WSJ, Forbes, MarketWatch, Business Insider, Fortune, Yahoo Finance, CNBC, The Economist) produced 87 thumbnails and zero charts. The FT gave me twelve photographs. The WSJ gave me nine photographs and three illustrations. The chart-heaviest sites in this sample, at one chart in twelve, were the New York Times and TechCrunch.

The median chart share across all 31 sites is zero. Not close to zero. Zero.

What the front pages are actually full of is photographs, 67 percent, then illustrations at 16 percent, then “other” at 16 percent. That last bucket is mostly logos, podcast art, magazine covers, and video title cards, which tells you something about what a homepage’s markup leads with before a human ever sees it.

Why the chart still wins when it appears

Here is the thing the count does not capture: where the chart sat. The NYT’s og:image, the picture that travels with every share of the homepage, was a two-line poll chart. It got the best seat on the page. TechCrunch’s, a timeline of a security incident, was the twelfth image I pulled, well down the markup. One chart in the lead slot is worth more than one in the footer, and the count treats them the same.

I think the reason is simple. A photograph promises a scene. A chart promises a fact. Readers can tell the difference in a quarter second, and the fact is the one they click. Editors know this, which is why the chart, when they have one, goes on top. But a chart takes hours to make and a photograph takes a wire subscription, and at eleven in the morning the wire wins on volume.

So my thesis was half right. Front pages are not chart-led. They are photo-led with a chart on top when someone bothered to make one.

What this means if you publish data

If you are the person making the chart, the news is good. You are not competing with other charts for attention. You are competing with a photograph of a man at a podium, and on a front page a clear line with a labelled endpoint beats the podium.

The bar is low because so few people clear it. The chart share on a typical front page is zero, which means one good chart is the most distinctive image on the page. That is true for a newspaper and it is true for an internal weekly update. The report with one honest chart in it is the one people remember.

How I would do it properly

Fetch with a real browser so the JavaScript-injected images show up. Sample the same 30 sites every day for a month, at the same hour, and keep the images. Have two people label independently and report agreement. Measure position, not just presence, because a chart in the lead slot is worth more than a chart in the footer. Then, and only then, say something about a trend.

What to do

• Before you trust any claim that “the news is all charts now”, ask how many front pages the person looked at. I looked at 31 and found two.
• If you make charts for a living, lead with them. On a page full of photographs, a chart is the only image that makes a promise.
• Never let a logo be your first image. Nine of 31 publishers did, and it says nothing.
• Treat this post as a method, not a result. The script is in the repo. Run it on a different day and tell me what you get.