Ask Skift Is How You Make AI a Media Product
AI is going to disrupt the media business. If we look at technological innovations over the past 120 years that have directly impacted media, it’s radio, television, the internet, and AI. The question is no longer if, but how quickly it’ll change our businesses.
On Wednesday, Google revealed how it’s putting more of its AI products into the actual search result pages. According to The Verge:
The future of Google Search is AI. But not in the way you think. The company synonymous with web search isn’t all in on chatbots (even though it’s building one, called Bard), and it’s not redesigning its homepage to look more like a ChatGPT-style messaging system. Instead, Google is putting AI front and center in the most valuable real estate on the internet: its existing search results.
To demonstrate, Liz Reid, Google’s VP of Search, flips open her laptop and starts typing into the Google search box. “Why is sourdough bread still so popular?” she writes and hits enter. Google’s normal search results load almost immediately. Above them, a rectangular orange section pulses and glows and shows the phrase “Generative AI is experimental.” A few seconds later, the glowing is replaced by an AI-generated summary: a few paragraphs detailing how good sourdough tastes, the upsides of its prebiotic abilities, and more. To the right, there are three links to sites with information that Reid says “corroborates” what’s in the summary.
This is everything we knew was coming. Google crawls every website, it has increasingly been able to determine whether content is unique or not, and now, it can contextualize answers to people’s questions without them needing to leave the platform.
Sure, it has three links to sites with information that Reid says corroborates what’s in the summary. And we don’t know how many people are going to actually engage with these links. But the reality is, what started as structured data in the smart snippets has evolved to providing answers to more in depth questions without the user needing to leave Google.
We have known for some time that the era of getting free traffic from major platforms was at an end. It has been written ad nauseam about BuzzFeed’s business failing in big part because Facebook reprioritized its algorithm away from viral content.
…
Every major platform either doesn’t find much value in publisher content—Facebook—or it has figured out a way to keep more of the traffic on its own property—AI-based search. In all of these cases, the publisher is going to see less traffic.
But with every bad, there is also a potential good. And for media companies that have invested heavily in their brands and are highly trusted by their readers, AI could be a major boon to publishers.
One of the earliest examples of this is the recently launched AI chatbot that Skift has released called Ask Skift. In an announcement, CEO Rafat Ali wrote:
We have “trained” Ask Skift on all the sum totality of Skift archives over the last 11 years, including daily stories, research reports, all of our clients’ trends reports, our specialized products – Airline Weekly, Daily Lodging Report, and Skift Meetings – and all the U.S. public travel companies’ financial SEC annual and quarterly reports. As soon as a new story or report is published, it goes straight into Ask Skift.
For people who are subscribers to Skift, they can ask three questions every month. And if a reader is a paying subscriber to one of its products, they will get access to unlimited questions. Further in the announcement, Ali explained why they built this:
Until now, we had a package-based relationship with our reader – a package we created and presented to you, in the form of a story, a research report or a conference. Now, with the new Generative AI tech, it is possible to have a Q&A based relationship with our users, at scale, in a way that just wasn’t feasible before. If we are the expert brand in the travel industry, we should have all the answers to your queries as we build this. For most part, we do, locked into various packages and this allows us to deconstruct our package and create custom answers to every query you have.
This might annoy many traditional reporters—the same who find Smart Brevity to be annoying—but it’s the right thinking. The way that we deliver stories to people is often long-form. And some reporters are great writers and it’s enjoyable to read. But in many respects, what matters more is the information within the story.
But just as important is the ease of finding that information. Day-to-day, a story might be the fastest way to get information, especially as news and analysis happen. But how do you get information that’s a decade old? Search works, but you still have to review everything and determine if it’s the right answer.
Enter Ask Skift. As the quote explains, the team gave the AI a lot of information to learn from. And so, a user asking AI questions and it then contextualizing information is faster than searching, which is faster than browsing. This is the best evolution for the reader.
I took a screenshot, but I would encourage you to visit yourself.
As a product for subscribers, this is a no-brainer. What I particularly like about this is it introduces functionality to a media company. Now you’re not just selling content; you’re selling an easy-to-access archive of topical information that surfaces content for readers to get more context. That makes the product stickier because readers not only need new information, but they also need to be able to easily access older stuff as well.
The issue, which Skift’s Ali spells out, is that migrating to GPT-4 is too expensive right now despite it being far better technology. According to OpenAI’s API page, it costs $0.002 for 1,000 tokens on GPT-3.5, which is equal to approximately 750 words. For GPT-4, though, that would cost anywhere from $0.03 to $0.12. Depending on how people engage, it can very quickly add up.
This makes it a no-brainer for subscriptions, like I said above, but harder if you’re running a free site. However, I could see a couple of ways for ad-based media companies to offer this without burning through API costs.
First, you could get a daily or advertiser underwriter where the user receives a message after they ask their question with a short prompt about an advertiser. Years ago, Quartz had a chat product where sponsors would be able to deliver an ad in a native format as people were engaging with the chat product. But because the ad is directly in feed, the interaction should, theoretically, be higher, which could allow for a higher CPM.
We could even make the advertisements more custom to the reader. In essence, we could have the AI learn about the advertiser and then deliver answers that are contextually related to what the reader asked. This means we’re not stuck delivering the same message to everyone.
Second, this would be a great way to promote sponsored gated content. The AI can understand the content and if it fits to the context of the answer, we could display this to readers. It becomes another avenue for providing leads to partners, which likely carries a higher revenue per user than a static advertisement.
The only way this works, though, is if you have an established brand. You need to be known in a specific industry or topic and have to be trusted. Otherwise, why will people come to you directly? Fortunately for those that have been building subscription businesses, trust and brand are also necessary traits, so the introduction of a chat-like product becomes more straightforward.
AI is here. And it’s going to disrupt things faster than any of us can anticipate. The way the Skift team has taken that disruptive tool and put it to work is smart. I am excited to see others try.
Thanks for reading today’s newsletter. If you have thoughts, hit reply or join the AMO Slack. I hope you have a great weekend!
