# What makes a good news summary? > Most news summaries fail by keeping too much, not too little. A 5-question test tells you in seconds whether a summary deserves your trust. Source: https://informed.now/blog/what-makes-a-good-news-summary By Kira Shishkin · November 8, 2026 · News Curation **Summary:** A good news summary is mostly subtraction. It keeps 5 things (what happened, who says so, what's new, what's still unknown, and where to read more) and drops everything else. Most summaries fail by keeping too much and by losing the source along the way. The usual way to judge a summary is to ask what it covers. The better question is what it leaves out. A news event arrives wrapped in reaction, speculation, background, and the writer's prediction. A summary that keeps all of that at shorter length is a compressed article, and a compressed article still costs you most of the reading time a summary was supposed to save. ### Why subtraction is the whole job People scan news summaries rather than read them. The Nielsen Norman Group pooled 5 eyetracking studies run over 13 years, with more than 500 participants, and reported in 2020 that reading all or even most of a page's text is extremely rare [online](https://www.nngroup.com/articles/how-people-read-online/). Readers jump between the first words of lines and paragraphs, hunting for the sentence that answers their question. Every sentence a summary keeps beyond that answer is a sentence the eye has to jump over. Subtraction is also what readers are asking for. In the Reuters Institute's 2025 Digital News [Report](https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2025/dnr-executive-summary), covering 48 markets, 40% of people said they sometimes or often avoid the news, up from 29% in 2017. Avoidance is what [information overload](https://informed.now/blog/information-overload) looks like after a few years. The same report found 27% of people interested in AI that summarizes articles, the most wanted of any AI news feature it tested. Those are 2 faces of one request: Give me less, and make the less count. ### The 5 things a summary has to keep Cutting is easy, and cutting the right things is the whole craft. A summary that passes keeps 5 items, and nothing else. 1. **What happened.** The event itself, in the past tense, with the number or the decision that makes it news. "The central bank held rates at 4.5%" is a summary. "Markets digest the central bank's latest move" is a headline about a summary. 2. **Who says so.** The source of the claim, named: A court filing, a company statement, a named official, or a study. A fact with no owner can't be checked, and a summary that drops the owner has removed the one thing that lets you decide how much to trust the rest. 3. **What's new.** The piece that wasn't known yesterday, separated from the background that was. Most of the length in a typical news article is background. A summary can assume you've either read that background or don't need it today. 4. **What's still unknown.** The honest gap. Whether the bill will pass, how many people were affected, or what caused the failure. Summaries that fill the gap with a guess read as more complete than summaries that admit the gap, and they are less true. 5. **Where to read more.** A link or a name. The summary's job is to tell you whether to go deeper, and with whom. Each item fits in a sentence. 5 sentences is a summary you can read in 20 seconds and trust for the rest of the day. ### Where summaries fail most Almost every summary gets the first item right. The second item is where they break. In October 2025 the European Broadcasting Union published a [study](https://www.ebu.ch/news/2025/10/ai-s-systemic-distortion-of-news-is-consistent-across-languages-and-territories-international-study-by-public-service-broadcaste) in which journalists from 22 public service media organizations in 18 countries graded more than 3,000 answers that AI assistants gave to news questions. 45% of the answers had at least one significant issue. The single biggest category was sourcing: 31% of answers had missing, misleading, or incorrect attributions, compared with 20% that had major accuracy problems. The assistants kept the facts more often than they kept the source. Human summaries lose the source the same way, just more quietly than the assistants do. A newsletter writes "officials say" without naming the office. A radio bulletin says "a new study finds" and never names the journal. The reader is left holding a claim with no handle on it. When the Reuters Institute asked people in 2025 what AI would do to news, they expected it to make news less trustworthy, by a net 18 [points](https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2025/dnr-executive-summary). Dropped sourcing is the mechanism behind that expectation. A summary earns trust by showing its work in one clause, and loses it by skipping that clause. ### Is a summary just a worse article? The strongest case against summaries is that they strip away what makes news worth reading: The context that explains why a vote mattered, the detail that shows what a policy does to one town, and the quote that reveals what an official believed. Compress those out and you have a fact sheet. Fact sheets rarely change minds, and they don't help a reader notice when a simple account is hiding a complicated one. That case is right about any news you care about. For the 2 or 3 topics a week that touch your life, your work, or your vote, a summary is the wrong format. Read the full article, and then a second one from a different outlet. The case is wrong about the other 40 items the day throws at you. Nobody reads 40 full articles, so the real comparison is between a summary and a headline, because the headline is what people read when they skip the article. Measured against the headline, a 5-sentence summary with a named source is a large gain in context. And a summary that keeps item 5, the pointer to the full piece, hands the reader the way back in whenever an item turns out to matter. ### A test you can run on any summary Take the next news summary you see, from any newsletter, podcast intro, push alert, or the bolded paragraph at the top of an article, and read it with 5 questions: - What happened, in one past-tense sentence? - Who says so, by name? - What here is new since yesterday? - What does the writer admit is unknown? - Where would I go to read the whole thing? A summary that answers 4 or 5 is doing its job. One that answers 2 or fewer is a compressed article or a headline wearing a longer coat, and either way the time you spend on it would be better spent on the full piece. The test works in reverse too. When you pass news along to a friend or a group chat, write the 5 lines yourself. The hardest line is the second one. Naming the source forces you to check whether you know it, and that check is where a lot of forwarded news dies quietly. [informed.now](https://informed.now) sends 6 news summaries every weekday morning, picked with one transparent filter, and each summary names its [source](https://informed.now/blog/news-sources-we-cite) so a reader can check it. The filter decides what gets cut, which is most of the day's news. The 6 summaries are the subtraction, and the sources are what we refuse to subtract. A [news digest](https://informed.now/blog/what-is-a-news-digest) from any outlet, run through the 5 questions, does the same work without us. ### What's left after the cutting A summary is a promise about what you can skip. The promise holds only if what remains carries its own evidence: The event, the name behind it, the new part, the gap, and the door back to the full account. The article is still there when you want it, and a good summary is how you know when you do.