Google opened a Preferred Sources button that publishers embed on their own pages, and more than 600,000 unique sources have now been selected

Google's August 20, 2026 announcement is that publishers can now embed an interactive Preferred Sources button directly on their pages, and that people have already selected more than 600,000 unique sources. A reader who clicks the button adds that site as a preferred source and is returned to the exact spot they were reading. Content from a selected site is more likely to appear in Top Stories, and for that reader it carries a preferred badge in AI Overviews and AI Mode. Per Google Search Central documentation updated the same day, the button requires one script tag and one empty div, and only domain-level and subdomain-level sites are eligible. ASAP lays out the implementation conditions and the numbers, then explains why this is an answer-engine lever rather than a search-ranking one, and where a large share of Korean blogs fail the eligibility test.

The button is one script tag and one empty div

The standard implementation in Google Search Central documentation is two lines of code: a script tag that asynchronously loads https://news.google.com/swg/js/v1/publisher.js, and an empty div carrying the google-add-preferred-source-btn attribute where the button should render. The documentation's last update is stamped 2026-08-20 UTC.

Two attributes control presentation. data-theme selects a light or dark button, and data-lang forces the button's language using a code from the list Google publishes. Left alone, the button localizes automatically.

Eligibility is restricted by URL structure. The documentation states that only domain-level and subdomain-level sites qualify, so example.com and code.example.com are eligible while a subdirectory such as example.com/blog is not.

The click flow is built to keep the reader in place. After adding the site, the reader returns to the same position on the publisher's page rather than landing in Google's settings, which removes one drop-off point compared with sending readers to Google and hoping they come back.

A selected site is treated differently in Top Stories and AI Overviews

Sites chosen as preferred sources receive two forms of preferential treatment. Google's documentation states that their content is more likely to appear in Top Stories and that it can be highlighted with a preferred badge in AI Mode and AI Overviews for the user who selected it.

Scope is per-user, and that defines the whole feature. Preferred Sources is not a site-wide ranking signal; it is a personalization setting that operates only on the results of the individual who set it. That is also why Google presented it inside a bundle of personalization features rather than as a publisher tool.

Preferred Sources reached AI surfaces on May 27, 2026. Google expanded the feature to AI Overviews and AI Mode that day and said selected sources are clearly labeled to stand out inside AI responses. The same announcement added a highly cited badge for original reporting that other articles reference, and an article carousel surfacing timely pieces on developing topics.

The path from 90,000 to 600,000 in eight months is the backdrop

Selected unique sources are traceable through three figures Google has published. The count stood at nearly 90,000 on December 16, 2025, passed 345,000 on May 27, 2026, and exceeded 600,000 on August 20, 2026, roughly a sixfold increase in eight months.

That trajectory places this button precisely. In December 2025, Preferred Sources was a setting users had to go find behind a star icon next to Top Stories. In May 2026 its reach expanded into AI Overviews and AI Mode. In August 2026 the act of selecting moved down onto the publisher's own page, so the expansion ran from surface area to enrollment path.

Google has repeated one conversion figure throughout. When someone picks a preferred source, they click through to that site twice as much on average, a claim that appears in both the December 2025 and May 2026 announcements.

That multiplier requires care in reading. It compares clicks from people who chose a source, and Google has not published the comparison baseline. Selection bias explains much of it, since readers pick outlets they already read often, so translating it into a promise that adding the button doubles traffic is a claim the source does not make.

Why this is an answer-engine lever rather than a ranking one

The variable this button touches is an explicit reader choice rather than content quality or links. Traditional search optimization works against Google's ranking algorithm; this button bypasses ranking and captures the user's setting directly, which makes it one of the few control points a publisher holds while AI Overviews and AI Mode absorb clicks.

The mechanics differ in a way that matters operationally. Ranking signals accumulate site-wide and pay off slowly, while a preferred source applies instantly and only to the one reader who clicked. It is a sum of individual relationships rather than a compounding asset, so outlets that already have returning readers have the most to recover.

Placement follows from that. Showing the button to a first-time visitor converts worse than putting it where newsletter subscribers and repeat readers pass, such as the end of an article, a newsletter confirmation page, or a post-subscription screen, where the reader has already chosen the outlet.

The lever does not expand total reach, however. Preferred Sources works only on the screens of users who selected the site, so reaching readers who do not yet know the outlet remains untouched. The accurate framing is a device that keeps an existing audience connected through AI surfaces, not one that recovers traffic lost to AI search.

The domain-level rule disqualifies a large share of Korean blogs

The restriction to domain-level and subdomain-level sites creates an immediate practical split in the Korean web. Outlets and personal sites on their own domain or subdomain can install the button, while platforms that assign each blog a subdirectory path leave their writers without an eligible unit.

Checking takes one glance at a URL. An address shaped like blog.example.com/post is a subdomain and qualifies; example.com/name/post is a subdirectory and does not. Posts on platforms such as Naver Blog or Brunch, which sit under a path on the platform domain, cannot become preferred sources per author, because the platform domain is the single source.

That reprices owning a domain. The advantages of a custom domain used to be framed around branding and portability, and eligibility for Google's personalization setting now joins the list, which gives platform-based writers one more concrete reason to move to their own domain or at least secure a subdomain.

For Korean publishers with their own domains, installation is available immediately. Two lines of code with no review or approval process means the deciding factor is not technical difficulty but where to place the button and how to explain it to readers.

The other two features in the same announcement are feed personalization

The two remaining features Google announced on August 20 cover Discover and Google News personalization. In Discover, readers can tap any three-dot menu in the feed to say, in their own words, which topics or links they want more or less of, arriving in the Google app in the coming days.

Google News gains curated audio. Users of the Google News app on Android can now choose specific topics to shape their audio briefings.

Placed side by side, the three features share a direction. Control over what gets shown shifts partly from the algorithm to the user, and what publishers get is a single button placed at the moment that transfer happens. The common thread is that publishers now petition readers rather than Google.

It only works when a user clicks, and nobody reports how many did

The sharpest constraint is that a publisher controls the presence of the button and nothing beyond it. Enrollment happens only through a reader's click, and Google's announcement shipped no reporting metric showing how many people have selected a given site as a preferred source.

Unmeasurability changes the nature of the decision. Without a conversion rate, placement cannot be optimized experimentally, and verification falls back on indirect signals such as Search Console impression and click trends. Instrumenting the button click in a publisher's own analytics is the minimum precaution.

Two questions stay open. One is how much a preferred-source selection actually changes citation frequency in AI Overviews and AI Mode. The other is whether those 600,000 sources concentrate among a few large outlets or spread widely. Google published neither number.

Two things are settled for publishers today. Installation costs essentially nothing at two lines of code, and eligibility is decided by URL structure. Because the size of the effect cannot yet be measured, this button belongs in the category of near-zero-cost moves made early rather than experiments carrying large expectations.

Source: ASAP analysis based on Google's official blog post "Personalize what you see on Google Search, News, Discover" (August 20, 2026), Google Search Central's Preferred Sources documentation (updated August 20, 2026), and Google's official blog posts of December 16, 2025 and May 27, 2026