The Washington Post deploys AI summaries. I have questions.
In a staff town hall last week, Washington Post Publisher and CEO Will Lewis, along with a stage full of C-suite leaders, announced their plan to recover from a year in which the company lost $77 million (and led to the buyouts over over 200 newsroom staffers, myself included). The plan involves pulling on a number of levers, including the introduction of a variety of subscription options. However, the town hall topic most seemed to latch on to was the company’s future uses of artificial intelligence.
This week, as far as I’m aware of, the public got its first taste of one such use with the deployment of an optional AI-generated summary for a climate story on rewiring the United States’ electrical grid to meet the ever-increasing demand for power. (Quick aside, it’s a really fascinating story by Shannon Osaka, even if you’re not into the economics of infrastructure and its intersection with politics.)
Here’s what the article looks like when you click on the AI summary option:
The key takeaways appear right atop the lead. And in about 10 seconds I can read the highlights of an article that would have taken me about 2-3 minutes to read in its entirety.
Comparing the summary to the story itself, the bullets are pretty good and do touch on the top facts from the story. The third bullet actually combines two different paragraphs that fall fairly deep in the story, around the 13th paragraph.
In short, the feature did exactly what it’s supposed to do. It saved me time. And time is money, as they say. So will it translate to a resurgence in revenue for The Post? And how will summaries align with the paper’s ethics? And how will this sit with The Post’s writers? These are the three big questions I have as The Post pushes ahead with its AI initiatives.
Let’s take them one at a time. First, is this going to translate into a bunch of subscriptions for The Post? I doubt it. This isn’t a differentiator. Axios has been summarizing takeaways for years, even if they have to do it in the primitive fashion of having people type out the words themselves. What’s more, any publisher can add this feature and they likely will.
Is it a good feature? Does it help me save time? Yes and yes. It’s also not something that’s going to make me subscribe to The Post and it’s not something you can advertise against because … “Here are the top takeaways from this article on government policy, brought to you by Meta” doesn’t really strike the right chord, does it?
Given all that, it doesn’t really move the needle as far as Lewis’s third big “pillar” of his plan: Make money.
On the ethics side, while there’s going to be a lot of handwringing around this, I don’t think it’s at all out-of-bounds, so long as it’s edited. It’s a shorter version of a story written and edited by The Post’s journalists. All good.
The only issue I’d raise is whether it creates any issues with how journalists have railed, repeatedly, about how the problem with modern discourse is a lack of context. That’s why Twitter/X is so awful for discussing important, nuanced topics, right?
The character count for the summary of this story is 267. It could have been a Tweet. But, like a Tweet, it doesn’t include some of the context. Namely, the bullets don’t answer why utilities profit more from building new lines than upgrading them. And honestly, that’s kind of important to anyone who actually wants to address this problem. Would an addition like “because of current government regulations and project funding” have solved that? Maybe. I also feel like anyone who really wants to rail against the absence of context will never be satisfied until ever possible contextual stone on the planet has been overturned.
I do wonder if the AI summary option will be applied to all articles, however. Does this work for a humor column? Or, for example, would it be nearly as effective or appealing on this highly-visual story about how generative AI “views” attractive women? And how would that sit with the people who designed this elaborate story?
And that leads us to the final question: How will AI summaries sit with the staff? Well, I can tell you that this development was instantly noted by several former Post staffers with a bit of skepticism, along with questions of the utility when the information for the first two bullets is present in the top three paragraphs of the story. And also, they asked, what’s the value of having the reporter writing an 800-word story if they’re just going to distill it in three bullets?
To me, this is where journalists need to swallow a bitter pill. Hardly anyone reads full stories. I know, it’s hard. Sit down if you’re feeling lightheaded. It’ll pass. But it’s true. We have metrics that show precisely this and it’s been true for years. The vast majority of readers tune out after the first few grafs … that is if they’ve even clicked through after skimming the headline. (Or the Tweet.)
Editors at The Post, and throughout the known universe, have been trying to get writers to write shorter and tighter forever. There’s no space restrictions on the internet, they say. There is a significant restriction on modern attention spans, says everyone who isn’t the author. And ironically, that’s one area where I think the summaries could help.
One thing the summaries do is it gives the reader a very quick taste of what’s to come below. And I think that’s a good thing. Frankly, if I find the summary intriguing, I’m probably more likely to read the full article than I am to move on. I’ll be really interested to see if the stories with AI summaries (and this is the only one I’ve seen so far) actually increase the engagement time for readers.
But what are your thoughts? Is this a useful feature? Worth paying for? The latest sign of the end times? Please speak your mind in the comments. And in the meantime, have a great weekend.
Hi, I’m Mike. I’m a former editor for The Washington Post and ESPN. In 2024 I founded and now operate Launcher, LLC, a digital media consultancy operating out of Arlington, Va. Want to work together? Reach out on LinkedIn.
Read more
AI, AI everywhere
OpenAI was in the news again this week, and most certainly not in a way it would prefer. Instead of building momentum off its latest achievements in generative AI with the recent release of GPT 4-o, the company was instead defending itself against threatened legal action from