Almost every lever a publisher has on Google is mediated by a system the publisher cannot address. Ranking, feed placement, whether an AI answer cites you: all of it is decided somewhere you do not have an account. Preferred Sources is the exception. It is a setting that lives with the reader, granted by the reader, and it is the only place in the Google stack where a publisher can ask a person directly for preferential treatment and have that request honoured. That structural oddity is why the feature attracts attention out of all proportion to any measured effect, and it is also why the interesting question is not how to install it but what you are actually buying.
What is Google Preferred Sources?
Preferred Sources is a Google Search personalisation control that lets a signed-in user nominate publications they want to see more of. Once a user selects a site, that site's content is more likely to appear in the Top Stories carousel for that user, marked with a "preferred" badge, and links to it can carry the same badge inside AI Mode and AI Overviews.
The selection is per-user, not per-publisher. It changes what one reader sees. It does not change your ranking, your indexing, or how any other reader's results are assembled. Google's own documentation frames the mechanism as audience building rather than optimisation, and states in terms that the promotional methods it documents are "not required to do them in order to appear as a preferred source". Eligibility is automatic for sites that appear in Google's source preferences tool; implementation is a distribution decision on top of it.
Availability differs by surface. Google's Search Central documentation says the feature works globally in Top Stories in every language where Google Search operates. In AI Mode and AI Overviews it works in all languages and locales where those two products themselves exist, which is a narrower footprint.
The feature has been shipping in stages for over a year. It began as a Search Labs experiment on 26 June 2025 for users in the United States and India, launched properly into Top Stories in those two markets on 12 August 2025, went global on 10 December 2025, gained a refreshed publisher guide with deeplinks and translated button assets on 30 April 2026, and extended into AI Overviews and AI Mode on 27 May 2026.
What changed on 20 August 2026
Until August, the mechanism sat entirely on Google's side. To designate a preferred source, a reader had to leave the article, find Google's source preferences tool, search for the publication by name, confirm the selection, and then find their own way back. Publishers could point at that tool with a link or a badge, but they were asking a reader mid-article to go and complete a form on a settings page belonging to someone else.
On 20 August 2026, in a post written by Mrinalini Loew, General Manager of the Google Search Ecosystem, Google released an embeddable button that closes the loop. The reader clicks it in place, the selection is made, and the reader is returned to the exact point on the page where they clicked. The same announcement disclosed that readers have now selected more than 600,000 unique sources, and bundled in two unrelated changes: free-text topic tuning inside Discover, and topic curation for the daily audio briefings in the Google News app on Android.
What the release changed is cost, not architecture. The ranking mechanics are identical. What was previously a promotional campaign to push readers onto a Google settings page is now two lines of markup and a conversion rate you control.
How to add the Preferred Sources button
Google documents three integration paths and recommends the first.
Standard JavaScript, the recommended path
Two lines. Load the library asynchronously, preferably in the head:
<script async src="https://news.google.com/swg/js/v1/publisher.js"></script>
Then place a div carrying the button attribute anywhere in the body you want the button to render:
<div google-add-preferred-source-btn></div>
The library scans the document for that attribute and renders a Google-styled badge, automatically localised to the reader's browser language. Two data attributes control presentation. data-theme accepts light or dark, defaulting to light. data-lang overrides the language against Google's published list of supported codes.
One detail worth noting for anyone auditing what they are loading: the script is served from news.google.com under the /swg/ path, which is the Subscribe with Google publisher library. The button ships on Google's reader-revenue infrastructure rather than on a standalone endpoint.
Advanced JavaScript, for custom design or custom triggers
If you want your own button design, or you want to fire the flow from an application milestone rather than a rendered badge, Google exposes the same capabilities two ways. Modern build environments can import the module from publisher.mjs, initialise it with a theme and language object, and bind addPreferredSource() to any click handler. Standard script tags load the same library with preferred-sources-control="manual" to suppress automatic rendering, then push initialisation and trigger binding into a global PREFERRED_SOURCE callback queue.
The documentation carries an explicit warning here. Omit the manual attribute and the library will immediately initialise any matching element on the page, which is a real risk if you are testing an attribute-tagged div in a template that ships to production.
Deeplink, if your CMS cannot run the script
The pre-August route still works and remains the fallback for anyone whose CMS will not execute the JavaScript. Link a reader to https://www.google.com/preferences/source?q=example.com with your own domain in the query parameter and they land on your entry in the source preferences tool. Google supplies official translated badge graphics for download, or you can design your own. The deeplink is also the only version that works outside your own pages, so it is the one to use in newsletters, social posts and app promotions.
The eligibility rule that catches consolidated publishers
Only domain-level and subdomain-level sites appear in the source preferences tool. Google's documentation makes the boundary explicit: https://www.example.com/ and https://code.example.com/ are both eligible, while the subdirectory https://www.example.com/blog is not.
This is more consequential than it first reads. A decade of SEO consolidation advice pushed large media groups to fold their vertical brands into folders on a single domain in order to concentrate authority. Those groups can now be selected only as a whole. A reader who wants the food coverage and not the political coverage has no way to express that at the source level, so the publisher is asking for an all-or-nothing commitment from a reader whose loyalty is usually to one section. Groups that kept their verticals on separate subdomains retain a granularity their consolidated competitors traded away, and they can run different button placements and different pitches per brand.
If you are a single-title publisher this does not affect you. If you run a portfolio on one domain, it is worth knowing before you write the promotional copy.
What Google has disclosed, and what it has not
Two numbers are public and both are Google's own.
The first is the count of unique sources selected. Google put it at nearly 90,000 on 10 December 2025, more than 175,000 by late February 2026, more than 345,000 on 27 May 2026 and more than 600,000 on 20 August 2026. That is roughly a sixfold rise across eight months, with the steepest section arriving after the feature reached AI Overviews and AI Mode in late May. Read it carefully, though: it counts unique sources selected, not selections made and not readers making them. A publication chosen by one reader and a publication chosen by a million each contribute one unit. The series describes the breadth of the long tail and says nothing about depth.
The second is the engagement claim. Google has stated since December 2025, through Robby Stein, VP of Product for Google Search, and Jaffer Zaidi, VP of Global News Partnerships, that when someone picks a preferred source they click through to that site roughly twice as much on average. The same doubling was restated in May 2026 and again by Liz Reid, VP and head of Google Search, in a June 2026 podcast appearance. Google has not published the baseline the doubling is measured against, nor whether the effect persists or decays after the initial selection. It is also, on its face, a claim with an obvious selection effect built in: readers who care enough to pin a publication were probably already clicking it more than average.
What Google has not disclosed is everything a publisher would need to plan with. There is no preferred-source dimension in Search Console. There is no Top Stories filter in Google's analytics. Google does not tell you how many readers have selected you, how many selections the median reader holds, how often a preferred badge appears per thousand queries, or what share of your Google traffic the mechanism accounts for.
Digiday reported in February 2026, six months after launch and drawing on six publishing executives, that the gap was already the defining complaint. One head of search at a news publisher, speaking anonymously, put it plainly: "We haven't seen any meaningful traffic from [Preferred Sources], and a big reason for that is Google doesn't give us a way to track it or report on it. It's just really hard to say if it's really moved the needle at all." Asked whether more analytics were coming, a Google spokesperson said the company had "heard overwhelmingly positive feedback from users, publishers and websites on Preferred Sources, and we continue to take into account input on what additional insights and updates would be helpful". That remains the position as of September 2026.
The result is a conversion mechanism with no conversion data attached to it. You can count clicks on the button using your own analytics, and you can observe nothing downstream of that click.
The three proxies that do exist
Absent a native metric, publishers have converged on three indirect measures. None is clean and all three are better than nothing.
Track button clicks on your own pages. This gives you a real numerator: how many readers engaged with the ask. It does not tell you how many completed the selection, though the August button removes most of the drop-off that used to sit between those two numbers, which makes the click a much better proxy than it was in February.
Track branded search volume and trend lines. Jessie Willms, SEO strategist at The Guardian, identified this as one of the few available signals: queries containing your masthead are the closest observable analogue to the loyalty Preferred Sources is meant to formalise, and they move for the same underlying reasons.
Track returning users and direct sessions in analytics. If a preferred-source strategy is working at all, it is working by converting occasional readers into habitual ones, and that shows up in returning-user cohorts before it shows up anywhere else.
Run all three against the date you shipped the button, accept that a December core update or a Discover reweighting can swamp the signal entirely, and treat the result as directional.
Who Preferred Sources actually helps
The honest answer is that it helps publishers who already have what it measures.
A reader has to know a publication exists before they can designate it. Selection is an act of confirmation, not discovery. That makes Preferred Sources a mechanism for converting existing brand recognition into surface priority, and it offers comparatively little to the newer or smaller sites whose discoverability has been eroding fastest. PPC Land made this point in a May 2026 analysis of the AI Mode expansion, and the August button does not change it: the release makes the act of confirmation cheaper without touching its precondition.
That is not an argument against using it. It is an argument about where it sits in a plan. Preferred Sources is a retention tool wearing the costume of an acquisition tool. If you have a masthead people recognise, a newsletter list, an app, or any other evidence of habitual readership, you have the raw material and the button is a cheap way to convert it. If your traffic is overwhelmingly first-time arrivals from search, you are being offered a lever that requires the thing you do not have.
It is also worth being clear about scale. Google Discover has accounted for around two-thirds of Google referrals to news sites since 2025, and referral volumes have fallen hard: Chartbeat figures cited through 2026 put page views from Google Search down 34% over the twelve months to December 2025, with Discover down 15% across the same window. Preferred Sources is a personalisation control operating inside Top Stories and the AI surfaces. It is not addressed to the part of the decline that is doing the damage.
Should publishers use it?
Yes, and with a specific expectation attached.
The case for shipping it is that the cost has collapsed. Two lines of markup, an afternoon of CMS work, no ongoing maintenance, no eligibility gate, no exclusivity, and no downside anyone has identified. Willms framed the calculation exactly right in Digiday: "It's not a ton of upside, because you can't measure the impact. But it's also, what is the downside to doing this?" When the cost is an afternoon and the mechanism is one of the only reader-granted signals in the stack, an unmeasurable upside is still worth having. BBC, Vox and Wired all ran reader-facing explainers when the feature launched, which is the right instinct: the button converts far better next to a sentence explaining what it does than sitting alone in a footer.
The case for keeping it in proportion is that you are being asked to invest promotional real estate in a loyalty primitive that Google owns, Google measures, Google reports nothing about, and Google can change or withdraw without notice. The pattern is familiar from the surrounding product line. Search profiles, launched in June 2026, shipped with documentation stating that changes to a profile do not affect ranking. The outbound link features added to AI Mode and AI Overviews in May 2026 were framed around user benefit with traffic impact left unquantified. Willms drew the obvious inference: "When Google doesn't make it easy to track something, it's probably because it's not necessarily a positive for publishers." That may be too cynical, but it is the correct prior when a company that reports impressions for everything else reports nothing here.
So the practical position is: ship it this week, promote it properly to the audience most likely to say yes, measure what you can, and put a zero in the revenue forecast next to it. It is a free option, and free options should be taken. It is not a plan.
Underneath all of this is a constraint the button cannot touch. Preferred Sources improves your odds with a reader who is present, signed in and willing to press something. A growing share of the demand for publisher content now arrives with no reader attached at all: retrieval agents and answer engines fetching pages to assemble a response somewhere else, where no badge is rendered and no preference is consulted. That traffic is invisible to the loyalty layer by construction, which is why publishers increasingly treat it as a separate problem with separate infrastructure. blankspace works on that side of the split, monetising agent retrievals at the CDN edge rather than competing for the human click. Both problems are real. Preferred Sources is a good, cheap answer to the first one, and it is not addressed to the second.
Frequently asked questions
What is Google Preferred Sources?
Preferred Sources is a Google Search feature that lets a signed-in reader nominate publications they want prioritised. Content from a selected site is more likely to appear in that reader's Top Stories carousel with a "preferred" badge, and links to it can carry the same badge in AI Overviews and AI Mode. The effect is per-reader and does not change the site's underlying ranking.
How do publishers add the Preferred Sources button?
Add Google's Preferred Sources library with a script tag pointing at news.google.com/swg/js/v1/publisher.js, then place a div carrying the google-add-preferred-source-btn attribute wherever the button should appear. Optional data-theme and data-lang attributes set the appearance and language. Publishers who cannot run the script can instead link readers to google.com/preferences/source with their domain as a query parameter.
Does Preferred Sources improve your Google ranking?
No. It changes what an individual reader who has selected you is shown, not how Google ranks your site generally. Google's documentation describes the promotional methods as audience building and states that implementing them is not required in order to appear as a preferred source. Treat it as a distribution mechanism for existing loyalty rather than an SEO tactic.
Can publishers measure the traffic Preferred Sources generates?
Not directly. There is no preferred-source dimension in Search Console and no Top Stories filter in Google's analytics, and Google does not disclose how many readers have selected a given site. The available proxies are clicks on the embedded button, branded search volume trends, and returning-user or direct-session cohorts measured against the date the button shipped.
Are subdirectories eligible for Preferred Sources?
No. Only domain-level and subdomain-level sites appear in Google's source preferences tool, so example.com and code.example.com are eligible while example.com/blog is not. Media groups that consolidated their vertical brands into folders on one domain can only be selected as a whole, while groups that kept separate subdomains can be selected brand by brand.