Popular Sources and AI Mode Create Filter Bubbles – A New Problem to Find
Barry Adams recently argued that Google is creating an ecosystem of audience loyalty. His view on the mechanics is that Popular Sources, Search Profiles, and Linking Subscriptions provide publishers with new tools to stay visible to trusted readers.
His piece tells businesses with loyal audiences how to keep them. The most difficult question is what happens to those who are not on the current list?
Google’s trust features allow people to choose the sources they want to appear more often. That creates a new discovery problem for sites that need awareness before they can gain likes.
What Popular Sources Do
Popular Sources allow people to select the publishers they want to see the most in search results.
Google has launched this feature in the US and India for top news. It went global in all supported languages in April. Then in May, Google brought Popular Sources to AI Preview and AI mode.
In Top Stories, selected sources appear more often, or in the “From your sources” section. In AI Overview and AI Mode, links from those sources are labeled with a badge so people can see them. Google says that searchers are twice as likely to click on a popular source, and more than 345,000 different sources have been selected so far.
Barry’s article covers the mechanics in depth.
When Likes Turn into Distributions
The features support the same goal.
Search Profiles, which launched in June in the US, offers many fans a dedicated Search page. A follow button can reveal more of that source’s content in Discover. And Subscription Linking allows paying readers to link publisher subscriptions to Google Accounts, so paid subscription content can be featured in Search, Discover, and other Google products.
Each feature rewards publishers that people already know. That’s a reasonable choice, but it means the layer of discovery becomes smaller for publishers who haven’t built that audience yet.
This is not like the old algorithmic filter bubbles. Popular Sources are different because people deliberately choose websites that they want more of.
That changes the behavior of the argument. You can’t fault an algorithm for the decisions people have made on purpose. But the structural effect is the same. This is a filter bubble problem on a user-directed form.
Profit builds in all aspects. Search Profiles require 100,000+ followers on YouTube, Instagram, or X, or 300,000 on TikTok. Linking subscriptions require an existing subscriber base. Each feature is easy to open with a stable audience.
What Adds Custom Quizzes
Questions add another layer of personalization on top of selected sources.
Robbie Stein from Google gave an example of how people search in AI Mode. Instead of “restaurants in Nashville,” people wrote questions like “restaurants in Nashville but your friend has allergies, and we have dogs, and we want to stay outside.” That single query gives Google more context about the user than a typical search ever did.
Put on top of source preferences and Google’s Human Intelligence feature, which connects Gmail and photo data in AI Mode, and the photo becomes individualized.
An iPullRank study published in May found a 46 percent increase in brand mentions for accounts connected to Personal Intelligence. Products that have been consumed increased from 23.9% to 66.8% of correct responses for AI mode, with Gmail showing the strongest impact. The test consists of three accounts within 17 days with only Personal Intelligence selected.
The combined effect is the search experience that two people asking the same question might not share. The query, sources, and background data of opt-in users are personalized, giving Google more individual context than a traditional keyword search.
How Content Creators Are Trying to Hack
The problem is getting into awareness before there is a preference. For publishers outside of the user’s select set, visibility must come from places Google’s preference layer does not fully control.
Another option becomes a source cited by preferred sources. If sources the user trusts refer to your work, your content can still be accessed. That means building a presence in podcasts, industry publications, original research, ChatGPT, social media, and peer recommendations, places where people meet new sources and where AI systems can find, cite, or display your work.
Another is to use the tools that Google provides. Search Central’s articles include a deeplink format and downloadable button assets that you can add to your site and other calls to action. A deeplink takes people directly to the source preferences tool with a pre-populated publisher URL. Google says the buttons are designed to sit alongside social follow notifications and newsletter signups.
Writing personal inquiries is a third option to look at. People who use AI Mode provide Google with detailed context about themselves. Content with original information and depth beyond AI summaries may perform better in conversational search.
We’ve put together the growth of AI-driven citation patterns in all of these areas, and that pattern is pointing in the same direction. Publishers cited in AI’s responses tend to have strong brand recognition across all channels, not just the highest level of traditional.
None of these methods are guaranteed. Google hasn’t disclosed how much weight Popular Sources carries compared to other signals, and adoption numbers are still early. But it’s the options that show how the feature works.
What we don’t know
Google reported 345,000 unique sources selected, but has not said how many people have activated Popular Sources.
If detection is low, the structure’s impact on detection is limited. If adoption grows around AI Mode, which Sundar Picahi said in May this year has already exceeded 1 billion active users, the effect could be very large.
Digiday reported in February that publishers cannot measure the effect of Popular Sources on their traffic. There is no Search Console filter for it, so you can’t see how many people have added your site as a popular source.
Google says popular sources see a 2x click-through rate, but there’s no way to verify that number on your site. In the AI overview and AI Mode, Google currently labels preferred sources with a badge rather than increasing its rank. How much that changes, and when, is an open question.
Looking Forward
Whether this creates meaningful barriers to discovery depends on adoption and how Google weighs these attributes against content quality and relevance. For businesses and search professionals, these features are already important. The question is how people become a source of choice before preference-based distribution becomes a big part of how search works.
Additional resources:
Featured Image: Konstantin Faraktinov/Shutterstock
