AI Overviews: The end of advertising-driven publications?
And how can newsletters like Substack's survive in this new world? A possible solution is pointed out.
This is an English translation of the text published in our sister magazine, Emprender (link below).
AI Overviews: The end of advertising-driven publications?
And how can newsletters like Substack's survive in this new world? A possible solution is pointed out
By now, publishers and many creators should be aware of the news about AI Overviews (which, however, I summarise below), so first I'd like to examine what the outlook will be for publishers, and what Substack creators and writers can do.
On the one hand, if publishers and publications cannot make a living from advertising, it is possible that most will opt to build paywalls for users to pay for. This may lead to a world where:
Almost all private and non-subsidised (politically, etc) publications will have their content restricted with paywalls, of whatever kind.
Google will offer more and more information and direct answers, as people, faced with so many websites where you have to pay to read the content, choose to let Google's algorithms give them the content.
It would be like a fish that eats its own tail: the survival of the media would put users in the hands of Google, which, precisely in order to respond, would increasingly reduce the appearance of independent publications in its results. And with far fewer readers, they would have to raise their prices and further restrict content (to force payment).
Some publications will probably never be able to charge for their content. It's not easy (even Time Magazine and Quartz had to back down). Perhaps, to prevent them from disappearing, there should be some kind of intervention from governments with some power on the internet.
As for platforms like Substack, which in addition to being a newsletter system, are also a set of posts visible and discoverable -more or less- on the internet, the situation is less bleak, but also worrying.
I'll leave for another day some ways to combat what's coming down on us (like having a homepage and sections attractive to direct traffic, etc) and I'll focus on perhaps Substack's biggest weakness: fragmentation.
It is difficult for a person to subscribe and pay 5 dollars a month for each newsletter they like. Comparatively, a subscription to a large publication with hundreds of monthly articles that costs 9 dollars or euros/month is a seemingly more sensible purchase (for the avid reader, especially).
For all these reasons, I have long believed that there should be a federated system of "substackers", where a user pays 5 dollars and has access to 100 or 1,000 newsletters to choose from, in every language with sufficient presence. It is better to charge $1 a year from 100,000 users than $50 from 100 users. But it's not clear to me that hundreds of writers here at Substack think the same way I do.
Now, the news:
The absolute star of Google I/O 2024 was Gemini , the artificial intelligence developed in Mountain View, which will permeate all of the company's services, including Google Search. The main new feature of the combination of Gemini and Google Search is called AI Overviews , which has already been rolled out to all US users, with other countries soon to follow. The goal is to offer this feature to more than a billion users by the end of the year. What are these AI previews? Instead of the usual list of websites, Google will compile the answers to users' search intentions and present them in a more graphically intelligible way. It will be possible to see a simplified or even augmented version of the answers and, of course, we will be able to communicate with Google/Gemini to expand and enrich them. But what will happen to all those web pages that are the real source of these results? According to tests, links included in Ai Overviews should receive more clicks than if they had appeared in traditional search results, as Google is committed to directing traffic to publishers and creators. Google's words in this regard should be taken at face value, even if the premise does not bode well.
Soon, starting with English in the US, multilevel reasoning capabilities will also come to AI abstracts. We're talking about searches like "find the best yoga or pilates studios in Boston and show me the details of their introductory offerings and the walking time from Beacon Hill", a query so complex that if you tried to feed it to Google now, you'd only get a bunch of different links that address at most one of those needs. And as we have already seen, this ability to group different intentions will also be very useful when planning trips. The advantage of these complex searches is that changing any of the parameters will be immediate, allowing the result to be adjusted on the fly, as shown in the animation below.
Other examples given by Google are the creation of personalised diets with many restrictions (microwave, special diets, etc.). Frankly, this is a minefield, because when one enters the field of nutrition/health, the first advice one should get is to contact a specialist, because each of us needs extremely personalised guidance in this area. And this may be another problem with the omnipresence of AI: the belief that we can do everything with it.
This is the original article, in Spanish:
