Google's Search Fix: A Follow Button With No Followers · Issue #210

Opening

Reader, on the 20th, Google handed publishers a button. Add a few lines of code to your site, and a button appears reading “Add this outlet as a preferred source.” Click it, and that publisher’s articles are supposed to show up more often in Google Search and AI Overviews.

For outlets whose traffic has been gutted by AI summaries, this sounds like welcome news. Several publications even reported it as “a lifeline Google threw to publishers.” I’ve actually written about this before—in an earlier issue, I covered how Google’s search overhaul has been draining traffic from small and mid-sized media outlets and websites across the board.

One Search Bar Is Rewriting the Food Chain · Issue No. 106 · INLEVEL9The ecosystem’s apex predator is restructuring its own ecosystem

letter.inlevel9.com

But looking at this button, something else came to mind first. This isn’t a search feature—it’s a follow button. Let me give you the conclusion up front: Google gave publishers a follow button, but it won’t tell them how many followers they actually have. And this design barely works for the very outlets that got hit hardest by the traffic collapse.

What Google Unveiled on August 20th

Let’s first look precisely at what was announced. Google revealed three things that day.

First, a Preferred Sources embed button. Readers could already go directly into Google’s settings page and designate outlets. What’s new is that publishers can now embed the button on their own sites, and when a reader clicks it, registration completes and they’re returned to the exact page they were reading. No drop-off in the process — that’s the key improvement.

Second, the Discover¹ feed in the Google app can now be adjusted using natural language. In the feed, you tap the three-dot menu and type something like “I want to see less of this topic,” and it’s reflected in real time.

Third, on Android’s Google News audio briefings, you can now directly select and configure the topics.

Line these three up and a single direction emerges. Google is embedding “you choose directly” in three places at once.

Google’s documentation also spells out what changes once you’re designated a Preferred Source. You show up more often in the Top Stories² section, and you get a “preferred” badge in AI Overviews and AI Mode³. There’s one thing worth flagging here: Google never says your ranking rises in regular search results. The official phrasing goes only as far as “more likely to appear in Top Stories.” And since you can only designate at the domain or subdomain level, outlets operating out of a subpath like example.com/blog aren’t eligible to begin with. You may have noticed — this is exactly why our own newsletter’s domain structure is testing the setup below.

INLEVEL9 Letter.com/issue/
letter.inlevel9.com/

Why Pull Out This Button Now, of All Times

Google hasn’t suddenly gotten friendly. There are numbers behind this.

Pew Research Center dug into the actual browsing records of 900 US adults — 68,879 Google searches, to be precise. When an AI summary appeared on search results, the click-through rate on regular links was 8%. Without a summary, it was 15%. That’s nearly a 50% drop. Clicks on the source links embedded inside the AI summary accounted for just 1% of total visits. Worth keeping in mind, though: this is US data from a single month, March 2025, and Google pushed back, arguing that “the sample queries aren’t representative of search as a whole.”

The traffic numbers are even starker. Chartbeat’s comparison of December 2024 and December 2025 shows that the decline in search-driven traffic split sharply by publisher size.

Small publishers (1,000–10,000 daily pageviews): down 60%
Mid-size publishers (10,000–100,000): down 47%
Large publishers (100,000+): down 22%

Total web traffic over the same period fell 6%. So this isn’t the web as a whole shrinking — it’s closer to an event where traffic got redistributed from small publishers to large ones. For reference, ChatGPT referrals grew over 200%, but still make up less than 1% of total traffic. AI chatbots didn’t take the traffic. It just vanished.

And one more thing. According to Chartbeat’s data aggregated across roughly 2,000 news and media sites worldwide, 68% of the traffic Google sends is from Discover, not search. Search accounts for only 32%.

These numbers are the real backdrop to today’s story. We still call Google a search engine, but from a publisher’s standpoint, Google is no longer a search box — it’s a feed. And a feed has no search terms. What it has instead is taste. A system that runs on taste doesn’t need keywords — it needs follow relationships. This new button is the missing part that fills that gap.

Let’s unpack that 600,000 figure

In its announcement, Google said “over 600,000 unique sources have been selected.” Back in the May announcement, the figure was 345,000 — that’s a 74% jump in just over three months. The growth itself is real.

But there are two things to watch out for when reading this number.

First, 600,000 is not a count of people. “Unique sources” counts the types of sites selected, not the people doing the selecting. If one person picks ten sites, that counts as ten. If ten thousand people pick the same site, that counts as one. So this metric tells you nothing about the scale of participation, nor the average number of followers per outlet. Google has never disclosed user numbers.

Second, the claim that people “click twice as much” isn’t causation. Google stated that people click on their preferred sources roughly twice as often on average. But the people who designated a source as “preferred” already liked that outlet to begin with. You can’t call it the button’s effect when the people who were going to click anyway are the ones clicking. This is a textbook case of what statisticians call selection bias⁴. Reading this as “adding a button doubles clicks” would be a mistake.

At this point, there’s a question I keep repeating to my students in data classes: “What did the person who built this number choose not to count?” What went uncounted here: the number of users, and any point of comparison.

The real problem is there’s no list

Now let’s get to the heart of today’s issue.

Say you install the button. Say 500 readers click it. So where can the publisher actually see that number, 500?

They can’t.

Digiday ran a piece last February interviewing people at BBC, Vox, Wired, The Guardian, and others, and it tackled this problem head-on. Here’s what publishers said they couldn’t verify: how many people came in through the preferred sources feature, what share of total traffic that represents, and how many people have designated their site as a preferred source. The head of search at one news outlet put it this way:

“We’ve never seen meaningful traffic from preferred sources. A big reason is that Google doesn’t give us any way to track it.”

What’s even more striking is the timeline. In August 2025, an SEO practitioner asked Rajan Patel, Google’s VP of Search Engineering, whether publishers could find out how many people had added their site as a preferred source. The answer was “we’re looking into it.” A year has passed since then, and that data still doesn’t exist. Instead, we got one more button.

To sum up the structure: the publisher installs a follow button on their own site, asks their own readers to click it, and the outcome of that relationship functions only inside Google’s algorithm. Google holds the list; the publisher does the labor.

The contrast with the social media era makes this clear. Even on a Facebook page, you could see your follower count. Even if reach dropped, you could at least run the diagnosis: “we have 100,000 followers but only 3% reach.” With preferred sources, you can’t even run that diagnosis. You can’t see the numerator or the denominator.

And think about who this design favors. It favors places that already have loyal readers willing to click the button. But the outlets that lost 60% of their traffic were the small ones. Sites with people willing to click lost only 22%; sites without such people lost 60%. It’s a medicine that works least where it’s needed most.

Oswald’s View

Honestly, I’m not a neutral observer here — I’m implicated in this problem myself.

Look at this newsletter’s recent numbers: web views run more than double email views. And over half of that web traffic comes from Google Discover. Put plainly, Discover delivers this newsletter to more people than my entire subscriber list does.

That should sound like good news. But every time I look at that number, it makes me a little uneasy. I didn’t build that Discover traffic. An algorithm picked this piece one day, and it won’t pick the next one. In practice, pieces of identical quality can pull 10,000 views one day and 500 the next. There’s almost no variable I control.

This is the failure pattern I’ve seen most often while building GTM strategy. When a distribution channel owns the customer relationship, one policy change from that channel can shake the whole business. A brand whose retailer won’t share customer data. A seller who exists only inside a marketplace. A commerce business whose revenue is tethered to platform reach. All the same structure. Everything’s fine until the quarter the policy changes — then it all collapses at once.

The preferred-source button is the latest version of this pattern. And this time, the platform has even outsourced the labor of building the relationship to the supplier. It amounts to: “Ask your readers to click it — we’re the only ones who see the results.”

So will I add the button? Yes. The cost is close to zero, and there’s no downside. The money and time it takes to embed one button are negligible, and if some readers click it, that’s pure upside. If they don’t, I lose nothing. This isn’t a hard bet to make.

But I won’t call it a solution. A preferred source is a device that nudges up your odds of exposure slightly — not a device that builds a relationship. A relationship only exists where there’s a list. That’s why, even though most of my traffic comes from Discover, I keep email delivery at the center of this newsletter. Email has no algorithm. If I send it, it arrives. That seemingly old-fashioned property is, right now, the most valuable asset I have.

Closing

To sum up, three things.

First, on August 20th Google rolled out a preferred-source embed button for publishers. It helps with exposure in Top Stories and AI answers, but nowhere does it promise a bump in general search rankings.

Second, the figure of 600,000 sources counts a type of site, not a number of people. The claim that a source gets “2x more clicks” also just counts people who already liked that outlet — it shouldn’t be read as the button’s effect.

Third, the biggest problem is that publishers can’t see their own follower count. Publishers do the labor of building the relationship, and Google holds the list of who’s in it.

Here’s one thing to try this week. If you run a site, check whether Google Search Console has a “Discover” performance report tab. This tab only appears for sites that have had Discover impressions within the last 16 months. If the tab is there, compare the share of traffic coming from Search versus Discover. If it isn’t, that means your site currently depends entirely on the search box, not Google’s feed. Either way, knowing that ratio beats building a content strategy blind to it.

💬 Have you ever seen the traffic sources for a site or channel you run suddenly flip? Tell me in the comments which channel collapsed and which one took its place. I’ll gather the examples for a follow-up piece on channel dependency.

💬 Tell me in the comments about a time your traffic sources flipped. I’ll fold it into the next issue. 📨 If you know a colleague wrestling with site or newsletter traffic, pass this along.

📎 References & Further Reading

Primary sources

Mrinalini Loew, “Personalize what you see on Google Search, News and Discover”, The Keyword (Google), August 20, 2026. ··· This is where today’s piece starts. Both the 600,000-source figure and all three features trace back here.
Google, “Guide to Preferred Sources in Google Search for Web Publishers”, Google Search Central. ··· Along with setup instructions for the button, this spells out exactly what Google does and doesn’t promise. The paragraph on placement alone is worth the read.
Sara Guaglione, “Without transparency, publishers can’t tell if Google’s Preferred Sources feature benefits them”, Digiday, February 26, 2026. ··· This is the core evidence behind today’s piece. It lays out in concrete terms what major outlets’ teams still can’t measure.
Athena Chapekis, Anna Lieb, “Google users are less likely to click on links when an AI summary appears in the results”, Pew Research Center, July 22, 2025. ··· This is the original source for the 8% vs. 15% figures. I’d recommend checking the methodology section for both the sample size — 68,879 searches from 900 U.S. adults — and Google’s rebuttal.
Sara Fischer, “Small publishers hit hardest by search traffic declines”, Axios, March 17, 2026. ··· This has the breakdown by outlet size: 60%, 47%, and 22% declines. The contrast with overall web traffic dropping just 6% is especially telling.

Background

Bron Maher, “Google Discover now makes up two-thirds of search traffic to news websites”, Press Gazette, August 7, 2025. ··· This Chartbeat data shows Discover accounts for 68% of Google traffic — hard evidence for the claim that the search engine has become a feed.
Barry Schwartz, “Google Looking Into Adding Analytics For Preferred Sources”, Search Engine Roundtable, August 14, 2025. ··· This is the original source for the “we’re looking into it” response. That this record is a year old is, in itself, informative for this piece.

Related past issues worth reading

One Search Box Is Reshaping the Whole Food Chain ··· Issue No. 106. It covered the bigger picture of Google reshaping the search food chain. Today’s piece is the latest scene in that same story.
The URL Berners-Lee Chose Died Within Four Months ··· This covered the era when links stopped being a path and became evidence. It shares the same root as today’s “the list” problem.

📝 Glossary

The author, Kwangseob Ahn, is a professor of business administration at Sejong University and lead consultant at INLEVEL9. He teaches statistics and data analysis, including business data management and business analytics, while leading GTM and AI strategy consulting in the field, designing the seam between technology and business. He has published academic research on a memory architecture for AI dialogue systems (HEMA) and runs Daily Arxiv, a daily curation of global AI papers. He holds a master's from Korea University's Graduate School of Technology Management and a KMBA. He is the author of Homo Brainless: The People Who Outsource Their Thinking.

Footnotes

Google Discover: A content feed that Google surfaces on its own within the Google app or Android home screen, no search query required. Unlike search, where users come looking, this is a case of the algorithm pushing content first. ↩
Top Stories: A card-format cluster of recent news that appears at the top of search results. It only shows up for time-sensitive queries, and it’s the spot where the preferred-source effect works hardest. ↩
AI Overviews / AI Mode: AI Overviews is the feature that generates a summary and attaches it to the top of search results; AI Mode turns the entire search experience into a conversational one. Both share the same trait: the answer can end before anyone clicks a link. ↩
Selection Bias: The error of misreading the cause of a result because the two groups being compared were made up of different kinds of people to begin with. It’s the same mistake as saying “joining a gym makes you healthy” just because gym members are healthier than average. ↩