Google Handed the Ranking Button to Readers: Preferred Sources Tripled to 600,000 in Three Months — and That "2x Click Rate" Is Selection Bias

Q: My blog lives at example.com/blog — is there a workaround?

Under current rules it is ineligible; Google documentation states only domain and subdomain level sites qualify. The viable path is migrating to a subdomain such as blog.example.com, but that is a full URL migration requiring 301 mapping and internal link rework, so it is not worth doing for this button alone.

Q: Can this recover the traffic AI Overviews took?

Not most of it. Pew Research Center tracked 68,879 searches and found AI summaries cut traditional result clicks from 15% to 8%, while Preferred Sources only operates on news and content-discovery surfaces. It reads like compensation, not repair.

On 20 August 2026, Google published a cross-surface personalization update spanning Search, Discover and News (Personalize what you see on Google Search, News, Discover). The most structurally significant item was not conversational Discover tuning — it was the Preferred Source button handed to publishers. Add two lines of HTML, and a reader can add your entire domain to the Preferred Sources in their Google account with one tap, then land back exactly where they were reading. Google disclosed a number alongside it: more than 600,000 unique sources have now been selected, up from 200,000 in May 2026. Three months, three times.

This lands after a click collapse Google itself engineered. Pew Research Center tracked 68,879 Google searches from 900 US adults and found users clicked a traditional result on just 8% of searches where an AI summary appeared, versus 15% where it did not. Clicks on links inside the summary itself: 1%. Over the same period AI Mode visits went from 126 million in May 2025 to 279 million in May 2026. The rollout cadence is no accident — US and India in mid-August 2025, global multi-language expansion at the end of April 2026, publisher-side triggering in August 2026. Google is not returning the click to the algorithm. It is returning it to a deliberate declaration by the reader.

Rivals answered the same problem differently. Perplexity chose revenue share: after the Publishers' Program it launched Comet Plus with an initial $42.5M publisher pool on an 80/20 split. OpenAI chose contracts: Press Ranger and OtterlyAI analysed 129.3 million citations across seven AI platforms and found publishers with an OpenAI licensing deal earn 48% more citations per page on ChatGPT, rising to 112% for exclusive signatories. The tooling layer is consolidating: Profound raised a $96M Series C at a $1B valuation in February 2026 (roughly $155M cumulative), and Sitecore acquired Scrunch for about $225M in June. Revenue share, contracts, tooling — all cost money. Google took a fourth road: outsource the choice to the user and pay nothing.

For small and medium businesses and agency teams this is one of the few structural windfalls in two years that needs no budget: two lines of HTML, no licensing fee, no domain-authority gate. The ceiling is equally hard — subdirectories are ineligible, Search Console has no dedicated report, and results depend on whether you already have returning readers. Below: the numbers, actions for three reader types, and why that "2x click rate" cannot be a KPI.

What Actually Shipped, With Numbers

Preferred Sources began as a star icon beside the Top Stories block: users flag outlets they like, and Google gives those sources more slots in later news-type queries. The 20 August update moved the trigger from Google's interface onto the publisher's own site. Per Google Search Central documentation, adding an element carrying the google-add-preferred-source-btn attribute is enough; the default script scans the DOM and renders an automatically translated badge.

The critical constraint sits in the same document: only domain and subdomain level sites are eligible. example.com and code.example.com qualify; the subdirectory example.com/blog does not. That single line disqualifies most corporate sites that park all their content under /blog.

On the data: Google says more than 600,000 unique sources have been selected cumulatively against 200,000 in May, and claims users click preferred-source links at roughly double the usual rate. Earlier Labs testing showed more than half of participants picked four or more sources — they treat it as a reading list. Winners: vertical sites with returning readers and clean domain-level architecture. Losers: aggregators living on one-off search traffic with content scattered across subdirectories.

Immediate Actions for Three Reader Types

Brand owners and SMB operators

☐ Confirm content sits on a domain or subdomain; if everything lives under /blog, add the migration to this quarter's technical debt
☐ Count how many people actually return: newsletter list, LINE friends, return-visit rate — that is the button's real ceiling
☐ Do not buy an AI visibility SaaS for this feature; measure on free tooling for three months first

Marketing and SEO practitioners

☐ Place the button at the end of articles beside the newsletter CTA, never in the header
☐ Build a GA4 custom event for clicks and segment new versus returning users for the honest conversion rate
☐ Watch Top Stories and Discover impressions via Search Console date comparison — there is no dedicated report
☐ Send your existing list one single-CTA email; it is the only lever that accelerates cold start

Developers and agencies

☐ Initialise via the SDK rather than letting the default script scan the DOM on every page
☐ Defer the Google-hosted JavaScript and measure LCP before and after launch
☐ Package "button install + GA4 event + Looker Studio dashboard" as a fixed-scope delivery

SaaS Tool Comparison

Tool	Price (2026)	Strength	Best for
Profound	From $99/mo; realistically $399–499+; enterprise $2,000–5,000+/mo	Cross-LLM citation tracking, crawler log analysis	Mid-to-large brands with a dedicated SEO team
Peec AI	From roughly €85/mo; Pro around €199; up to about $495	Prompt-level competitive visibility comparison	European and multi-language brands
Otterly.AI	Lite $29/mo (15 prompts, 4 engines); Standard $189	Entry price point, citation source lists	Single-product SMBs testing the water
GA4 + Looker Studio (DIY)	$0	Button clicks, AI Assistant channel traffic	Zero budget, own-conversion focus

None of those three SaaS products can measure Preferred Sources. They measure citations in ChatGPT, Perplexity and AI Overviews. They are listed because most executives will be sold a "one tool solves it all" story at exactly this moment.

What Nobody Will Tell You

600,000 is not a user count. It is the cumulative number of sites selected at least once by at least one person. Google has never disclosed how many people use the feature, nor any monthly active figure.
The "2x click rate" is selection bias. People who actively pick you are already returning readers and subscribers whose click rate was always higher. Installing the button will not double a stranger's click rate — you have only relocated where existing loyalty gets counted.
No report means no attribution. Search Console still has no Preferred Sources dimension, so you cannot separate a Top Stories impression earned by ranking from one earned by a user preference.
The subdirectory rule is a silent elimination round, and no Google interface will ever tell you that you are ineligible.

The No-Subscription SMB Alternative

Week 1 — Eligibility and install. Confirm content sits at domain or subdomain level, add the two HTML lines per the official documentation, and place the button between the article ending and the newsletter CTA.
Week 2 — Homegrown measurement. Create a GA4 preferred_source_click custom event carrying article slug and user_type; pull GA4's existing AI Assistant channel into a separate table so the two never blend.
Week 3 — Cold start and dashboard. Send one single-CTA email to your list, then assemble GA4 events and Search Console impressions into one page with free-tier Looker Studio.
Ongoing — owned channels. Preferred Sources is effectively Google's version of a subscription; the thicker your RSS, newsletter and LINE lists, the better it converts. The order cannot be reversed.

FAQ

Will installing the Preferred Source button improve my rankings?

No. It applies only to source-level surfaces such as Top Stories, Google News and Discover, and only for accounts that actively selected you. Ordinary query ranking logic is unaffected.

My blog lives at example.com/blog — is there a workaround?

Under current rules it is ineligible. The viable path is migrating to a subdomain such as blog.example.com, but that is a full URL migration requiring 301 mapping and internal link rework — not worth doing for this button alone.

Can this recover the traffic AI Overviews took?

Not most of it. Pew's data shows AI summaries cut traditional result clicks from 15% to 8%, while Preferred Sources only operates on news and content-discovery surfaces. It reads like compensation, not repair.

Should I buy an AI visibility tool at the same time?

Decouple the decisions. Measure Preferred Sources on free tooling for three months to get a baseline; for AI visibility tools, manually test 20 prompts in your category first and skip the $99–499 monthly spend if your hit rate is under 10%.

My Take

The mainstream reading is "Google gives power back to publishers." My judgement is the inverse: Google shifted the cost onto publishers. Perplexity pays $42.5M in cash revenue share; OpenAI pays licensing fees that buy a 48% citation premium. Both pay for content. Google offers a button, and every cost sits on the publisher side — you restructure the site, you email readers asking for a tap, you build your own dashboard. Google pays nothing and collects an extra behavioural signal.

The more contrarian call: Preferred Sources will grow a paid placement within 18 months. The pattern is well established — any interface where users declare a preference and demonstrably click above average eventually sprouts sponsored slots. Google is already testing conversational ad formats in AI Mode. Once 600,000 doubles again, the Top Stories preferred-source carousel becomes the next inventory line. Right now is the free window.

For ScriptWalker the implication is concrete: an immediately productisable engagement — "eligibility audit + button install + GA4 event tracking + Looker Studio dashboard", delivered in two to three weeks, clearing a very low client decision threshold (no annual contract, no subscription), and naturally surfacing the larger follow-on build: migrating site architecture from subdirectory to subdomain. Trading a cheap, high-completion entry project for a high-value restructuring build is the best bet an agency can place in the second half of 2026.

Sources

Primary

Third-party