SEOTechnical SEOGoogleAI SearchAI VisibilityIndustry NewsAugust 23, 2026

Google lets readers pin your site as a preferred source, from a button on your page

Google shipped an embeddable Preferred Sources button on 20 August. Two lines of HTML let a reader mark your site as a favourite, which lifts you in Top Stories, AI Overviews and AI Mode.

TL;DR: Google search results are normally decided by Google. But there is a setting where a reader can name the websites they personally want to see more of — their “preferred sources”. Until now you had to dig through Google’s own settings to add one. On 20 August Google released a button that publishers can paste onto their own pages, so a reader can mark your site as a favourite while they are reading it, without leaving the page. Google says more than 600,000 unique sources have been picked so far, and that people are twice as likely to click a result once it is a preferred source. It is two lines of HTML.

First, what is a “preferred source”?

Google Search has a personalisation setting. A reader can tell Google: when a story is breaking, show me these publications near the top. Those picks are that reader’s preferred sources.

The picks don’t rewrite the whole results page. They bias specific slots. According to Google’s documentation for publishers, a preferred source is more likely to surface in Top Stories (the news carousel), in AI Overviews and in AI Mode. AI Overviews is the summarised answer Google writes at the top of the page; AI Mode is the separate chat-style tab where you hold a conversation with Search instead of scanning ten blue links. Both of them cite sources, and both are places where being one of the cited names is now partly a reader’s decision rather than purely a ranking outcome.

The setting has existed since May. What was missing was any sane way for a reader to find it.

What actually shipped

An embeddable button. You add a script tag and a div:

<script async src="https://news.google.com/swg/js/v1/publisher.js"></script>
<div google-add-preferred-source-btn></div>

The script goes in the <head>. The div goes wherever you want the button to appear. Google renders a styled, automatically localised button in the reader’s own language. You can force a language with data-lang, and switch to a dark version with data-theme="dark".

The behaviour matters more than the markup. When a reader taps it, your site is added to their preferred sources and they are returned to the article they were reading. No redirect to a Google settings screen, no lost session. That is the difference between a feature that exists and a feature people use.

If you cannot run JavaScript at all, Google documents a plain link fallback: https://www.google.com/preferences/source?q=example.com.

One eligibility limit is worth knowing before you plan anything around it. Preferred sources work at the domain and subdomain level only. A subdirectory (example.com/blog) cannot be a preferred source. If your editorial content lives in a folder under a corporate domain, the reader who presses the button subscribes to the whole domain rather than to your section.

The numbers, and what they don’t say

Google’s announcement gives two figures. More than 600,000 unique sources have been selected. And, as TechCrunch reported, Google says readers are twice as likely to click through to a preferred source when one appears.

Be careful with the second one. “Twice as likely to click a source you explicitly said you like” is close to a definition rather than a finding. Someone who went to the trouble of pinning your publication was already inclined to read you. The uplift is real, but it measures intent as much as placement, and Google has not published the methodology.

The 600,000 figure is more interesting. Spread across the entire web, it is not a large number. That is presumably why Google is now handing publishers a distribution mechanism instead of waiting for people to find a settings menu.

TechCrunch frames the launch as a response to AI-driven traffic losses, and that framing is fair. Whether a button materially offsets what AI answers took is an open question nobody has data on yet.

What to actually do

Treat it as a small, cheap, low-risk addition. It is not a ranking factor you can game and it will not rescue a site that has no readers.

Put it where loyalty already exists. The end of an article, the newsletter sign-up block, the “about this publication” footer. Not a homepage hero. The reader who presses it just finished something of yours and liked it, so meet them there.
Check it renders. It is third-party JavaScript from news.google.com. Content Security Policy headers, consent managers and script blockers will all happily suppress it, and it fails quietly. A button that isn’t there looks exactly like a button nobody pressed.
Set data-theme deliberately. Light is the default. On a dark-themed site the default is going to look wrong.
Check your domain structure first. If your content sits in a subdirectory, the button still works, but you are asking readers to prefer the parent domain. Decide whether that is what you want.
Don’t nag. There is no version of this that survives being a modal.

The same announcement also let readers tune their Discover feed in plain language, typing what they want more and less of, as Search Engine Land covered. That is the same idea pointed the other way, and worth watching if Discover sends you meaningful traffic. A reader can now ask for less of a topic directly, and no amount of optimisation gets you past that.

Where the automation fits

For one site this is a ten-minute job. For an agency with sixty client sites it is sixty deploys, sixty CSP allowlists to check, and sixty pages to re-open afterwards to confirm the button rendered instead of being swallowed by a consent script. That last part is the one that gets skipped, and it is the one that silently costs you the feature. Preferium exists for that shape of work: it finds and fixes issues across every page, ships the change, and a real browser re-checks the live page after each deploy, so “it rendered” is verified rather than assumed. And because preferred sources feed Top Stories, AI Overviews and AI Mode, the payoff shows up on surfaces the system already tracks separately from the four AI engines it measures on every plan. More on how the system works.

Key takeaways

Google shipped an embeddable Preferred Sources button on 20 August 2026 — two lines of HTML that let a reader mark your site as a favourite without leaving your page.
Preferred sources get lifted in Top Stories, AI Overviews and AI Mode, so part of AI-answer citation is now a reader setting, not only a ranking outcome.
Google reports 600,000+ unique sources selected and a 2× click-through lift on preferred sources. The second figure partly measures existing reader intent, not placement alone.
Domains and subdomains only. A /blog subdirectory cannot be a preferred source.
It is third-party JavaScript and fails silently. CSP, consent tools and blockers can suppress it, so verify it renders on the live page rather than assuming the deploy worked.
The same release lets readers tune Discover in plain language, including asking for less of a topic. There is no optimisation that beats an explicit opt-out.