Google ships an embeddable Preferred Sources button, with 600,000 sources already selected

Google announced on 20 August 2026 that publishers can embed a Preferred Sources button directly on their own pages. When a reader clicks it, that site is added as a Preferred Source on Google, and the reader is sent straight back to where they left off on the publisher's page rather than into Google's source preferences tool. Google said people have already selected more than 600,000 unique sources, and that preferred sites appear more easily in Top Stories, AI Overviews and AI Mode. The code for the button is published in Google Search Central documentation.

What the Preferred Sources button actually does

Preferred Sources is a reader-side setting on Google. Before Google's 20 August 2026 announcement, a person who wanted to mark a publisher as preferred had to leave the article, find Google's source preferences tool, search for the site by name and tick it. Almost nobody does that unprompted. The change Google described moves the action onto the publisher's own page: the publisher pastes a snippet from Google Search Central documentation into a template, the reader clicks the button that renders, and the preference is recorded.

The return behaviour is the detail worth reading twice. Google said the reader is sent back to where they left off on the publisher's page instead of being handed into the preferences tool. That is what makes the button usable as a real on-page element. A control that dumps a reader into a Google settings screen mid-article is a control almost no editor would place above the fold. One that records the preference and returns the reader to the paragraph they were on is something an editor can put in a sidebar, a newsletter footer or an article end-slug without losing the session.

Google also repeated the payoff it attaches to the setting: preferred sites appear more easily in Top Stories, AI Overviews and AI Mode. That is the sentence that makes this interesting to anyone working on visibility inside AI answers, because Google is connecting a reader-controlled preference to placement in generative surfaces. AI Overviews and AI Mode are named alongside Top Stories.

What the button does not do, stated precisely

This is a per-reader preference, not a ranking factor applied to everyone. Nothing in Google's 20 August 2026 announcement says that collecting preferred-source clicks lifts a domain in general search results, or that a site with many preferred-source selections outranks one with few for a reader who has selected neither. The effect Google described operates for the individual who made the selection, on that individual's Google surfaces.

That distinction matters because the number Google published invites the wrong reading. More than 600,000 unique sources have been selected, Google said. That is a count of distinct sources people have picked, and Google did not break it down by how many readers picked each one, how often preferred sources actually surface, or what share of a preferred publisher's Google traffic the setting accounts for. Treating the button as an SEO lever that scales with clicks is an inference the announcement does not support.

The honest framing is narrower and still useful: this is the first Google ranking-adjacent control a publisher can hand to a reader from its own page. Everything else a publisher does to influence Google visibility happens on the publisher's side of the glass, in markup, links, content and technical health. This one happens in the reader's account, and the publisher's only job is to make the request.

The three changes Google described on 20 August 2026

The announcement bundled three separate things, and they sit at different stages of availability. Separating them keeps the shipped part apart from the described-as-coming part.

Google published the announcement on The Keyword, its company blog, and the post is available in full there.

Who this realistically helps, and who it does not

The button converts loyalty that already exists. It does not create loyalty. A site whose readers arrive once from a search result, read one answer and never return has nobody to ask. The click has to come from a person who already knows the brand well enough to want more of it, which means the population that can be converted is roughly the site's returning-visitor base plus its newsletter, app and messaging audience.

On that reading, the publishers and brands with something to gain here are the ones with a repeat audience and a modest domain authority profile: trade titles, niche vertical publishers, local news operations, and brand blogs with a subscriber list. Those are exactly the properties that lose to large general-interest domains in open ranking and can now route around part of that competition for their own readers.

The ones with least to gain are sites whose traffic is overwhelmingly non-returning: programmatic content farms, thin affiliate pages, and any property whose content programme is built for one-time capture rather than repeat readership. For them the button is a widget that nobody clicks, and the underlying problem stays what it was.

Where to put it on a page

Google did not publish placement guidance in the 20 August announcement beyond publishing the code in Search Central documentation, so what follows is reasoning from the mechanism rather than a Google recommendation. The button asks for a commitment, which means it belongs where commitment is highest. That is the end of an article a reader finished, not the top of one they have not started. It belongs on the pages a returning reader actually lands on, which for most sites is the homepage, section fronts and the two or three evergreen pages that carry the brand.

The one placement that reliably underperforms for controls like this is the interstitial or the modal on first visit, because the reader has no basis yet for the preference and the request reads as an interruption. The same logic that governs newsletter capture applies. Ask after value has been delivered, not before.

Treat measurement carefully as well. The button does not report back a conversion the way a form does. What a publisher can measure is the click, and only if the surrounding markup is instrumented. Anyone running this at scale should decide in advance which event fires and where it lands in analytics, because retrofitting that later across a template set is the expensive version.

The Discover control, and what it means for anyone optimising for Discover

In the same 20 August 2026 announcement, Google said that in the coming days, Google app users will be able to tap the three-dot menu on any Discover card and describe in their own words the topics or links they want more or less of, with the feed adjusting on the fly and remembering the request. That is a shift from thumbs-up and thumbs-down signalling to stated intent in free text.

Discover traffic has always been a black box that rewards headline curiosity and image treatment. A reader who can now type what they do not want changes the risk profile of that playbook. Headline patterns tuned for the tap rather than the read give the reader a fast way to say less of this, and Google said the feed remembers the request. The cost of a bad tap goes up, and it is a per-reader cost that no aggregate metric will show cleanly.

Google did not publish how the natural-language input is interpreted, whether requests apply at the publisher level or the topic level, or how long a remembered request persists. Anyone rebuilding a Discover strategy on this should note that those are the three questions that decide the answer, and none of them was answered.

Audio briefings and the pilot-programme gate

Google also said that daily audio briefings in the Google News app on Android can now be curated by topic, with source attribution and links to full articles, drawing on publishers in Google's news AI pilot programme. Two constraints are stated in that sentence and both limit how broadly it applies. The surface is the Google News app on Android, and the pool is publishers already inside a pilot programme. Google did not describe eligibility criteria for that programme in this announcement, nor did it say when or whether the pool widens.

The attribution point is the substantive one. Google said the briefings carry source attribution and links to full articles, which is a different posture from a generative summary that answers without sending anyone anywhere. For publishers, an audio surface that names the source and links out is worth watching, but there is nothing here to act on unless you are already in the pilot.

What Google did not say

The gaps are as load-bearing as the announcement. Google gave no data on how much visibility a preferred source actually gains, in Top Stories or in AI Overviews or in AI Mode. It gave no market list, and it did not say whether the embeddable button applies outside the United States. It did not say whether the 600,000 figure is growing, over what period it accumulated, or how selections distribute across sources. It did not describe eligibility for the news AI pilot programme. It did not state placement guidance for the button.

None of that makes the announcement less real. It makes the sizing exercise impossible from the outside, which is the honest thing to say before anyone builds a forecast on it.

What this means for Thai marketers

Thai publishers and brand blogs sit in an unusual position here. A large share of their returning readership arrives through LINE and Facebook rather than through Google, which means the loyal audience exists but does not currently translate into any Google-side signal. The Preferred Sources button is a mechanism for converting that loyalty into Google visibility for those same readers, without competing head-on for the English-language authority signals that decide open rankings.

That is worth saying plainly because it inverts a familiar constraint. A Thai-language site building organic visibility in Thai search against international domains is fighting an authority battle it did not choose. A reader who has already followed the brand on LINE for two years and clicks the button is not part of that battle at all. The audience a Thai brand already owns becomes the input.

The caveat is the one Google left open. Nothing in the 20 August announcement says the button is available in Thailand or in Thai, and no market list was published. A Thai publisher should treat implementation as a test to run and observe, not as a rollout with a known result. Check whether the Search Central code renders for Thai-locale readers before promising anything internally.

FAQ

Is the Preferred Sources button live in Thailand?

Google did not say. The 20 August 2026 announcement published no market list and did not state whether the embeddable button applies outside the United States. The code is in Search Central documentation, so a publisher can test whether it renders for their audience, but there is no Google statement covering Thailand either way.

Does getting readers to click it make my site rank higher for everyone?

No. It is a per-reader preference, not a ranking factor applied to all searchers. Google said preferred sites appear more easily in Top Stories, AI Overviews and AI Mode, and that effect operates for the person who made the selection. A high count of preferred-source clicks is not described anywhere in the announcement as lifting a domain for readers who did not click.

Do I have to do anything?

No. This is optional and there is no penalty described for ignoring it. The button is code a publisher chooses to embed. Sites without a meaningful returning audience have little to gain from adding it, and nothing changes for them if they do not.

How much extra traffic will a preferred source actually get?

Google published no figure. The announcement states that preferred sites appear more easily in Top Stories, AI Overviews and AI Mode and that more than 600,000 unique sources have been selected, but it gives no data connecting a selection to a measurable change in impressions or clicks for the publisher.

What is the news AI pilot programme behind the audio briefings?

Google said the daily audio briefings in the Google News app on Android draw on publishers in its news AI pilot programme, and described nothing further about it in this announcement. No eligibility criteria, no participant list and no timeline for widening the pool were published.

The short version

Google handed publishers a small piece of its own interface and told them to put it on their pages. The Preferred Sources button will do nothing for a site with no returning readers and something real for a site with many, and the size of that something is a number Google chose not to publish. Test it where your loyal audience actually lands, instrument the click before you scale it, and keep the claim honest internally: this is per-reader preference, not a rank lift.

If you are working out where reader loyalty, Google Discover and AI answer surfaces intersect for your brand in Thailand, that is the kind of question the Relevant Audience team works through with clients every week.