Preferred Sources makes AI visibility personal
Crawlmind Engineering··5 min read
Preferred Sources is a Google Search setting that lets a signed-in user pick the websites they want to see more of, and since May 2026 it marks those sites with a "preferred" badge inside AI Overviews and AI Mode as well as Top Stories.
That last part changes what an AI visibility number means. Until now, the working assumption behind most tracking (ours included) was that Google's AI answers are roughly the same for everyone who types the same query in the same place. Preferred Sources is Google stating, in product form, that the answer is starting to depend on who is asking.
#What Google actually shipped
The feature started small. In August 2025 Google launched Preferred Sources for Top Stories, in English, in the United States and India, as a way to see chosen news outlets more often when they had fresh coverage for a query (Google).
On May 27, 2026 Google brought it into its AI surfaces. The announcement says users will be able to spot links in AI responses from the sources they already selected, reports that people are twice as likely to click through to a Preferred Source, and puts the number of unique sources selected at more than 345,000 (Google). It also widened eligibility well beyond news: "any website that publishes fresh content is eligible" (Google).
By August 21, 2026, when Google updated the publisher button so readers stay on the site after adding it, the count had passed 600,000 unique sources (Search Engine Roundtable).
Two details in the rollout matter more than the headline.
First, today the badge is a highlight, not a selection mechanism. It appears when a chosen site already shows up in the response. Google has said it is working on using Preferred Sources as a ranking signal across its AI features so that selected sites appear more often (9to5Google). That is the part to plan for.
Second, the AI surfaces sit behind another switch. Google's developer documentation says that to appear as a preferred source in AI Mode and AI Overviews, "you must make sure your site is included in Search generative AI features in Search Console" (Google Search Central). A site that turned off the Search generative AI control gives up this surface along with the citations.
#Why a logged-out audit misses it
Preferred Sources only exists for a signed-in user who has made a choice. Automated visibility checks, whether a SaaS tracker or an internal script, generally query from clean, logged-out sessions so that results are repeatable. That design is correct for measuring the shared baseline. It is blind by construction to anything layered on top of an account.
Right now the gap is mostly cosmetic: a badge and a higher click rate on links that were going to appear anyway. Once selections start influencing which sources get pulled into an answer, the gap becomes directional. A brand with a loyal audience that has added it will be cited more for those readers than any logged-out check shows. A brand nobody has chosen will look the same in the audit and do worse in the field, because a competitor someone did pick can take the slot.
Google is also nudging more AI Mode use toward signed-in sessions. In late August it was spotted testing a "Sign in to save" prompt inside AI Mode responses (Search Engine Roundtable). It is a test, not a launch, but the direction is consistent: more of the AI answer surface is going to be rendered for a known person.
We have written before about a version of this in assistant recommendations, where stating a buyer persona in the prompt shifts which brands get named. Preferred Sources moves the conditioning out of the prompt and into the account, where no prompt library can reproduce it.
#The eligibility trap in your URL structure
One rule in the documentation deserves an audit of its own. Only domain-level and subdomain-level sites qualify. Google's own example: https://www.example.com/ and https://code.example.com/ are eligible, but https://www.example.com/blog is not (Google Search Central).
For a publisher this is a non-issue. For a B2B company whose editorial content lives in a subdirectory, it means the reader cannot choose the blog on its own. They can only choose the whole domain, and the badge follows the domain. That is usually fine, since the domain is what you want remembered. It becomes a problem for brands that run a content property on a path under a parent company's domain, or a regional edition as a folder. The preference, and any future ranking weight attached to it, attaches to the parent, not to you.
#What to do about it
Preferred Sources is not an optimization you can apply to a page. It is an audience asset, and it rewards the same thing that earns repeat direct visits: people who already know your name and want more of it. That has a few practical consequences.
- Ask the readers you already have. Google provides an embeddable button and a deeplink for exactly this (Google Search Central). Put it where returning readers are: newsletter footers, the end of your most-read articles, documentation home pages. A first-time visitor from a search result is the least likely person to add you.
- Check your Search Console setting before you promote it. If the generative AI control is off, the button still works for Top Stories but not for AI Overviews or AI Mode. Decide on the control first.
- Track the adds yourself. The documentation covers eligibility and implementation but gives no way to see how many readers selected you (Google Search Central). Fire an analytics event on the button click. It will not tell you how many people finished the flow, but a trend line is better than nothing.
- Read logged-out visibility as a floor for loyal brands and a ceiling for unknown ones. Keep measuring the shared baseline, because that is still what most searchers see. Stop treating it as the whole picture. If your brand has strong direct traffic, your real AI visibility among existing readers is probably higher than the audit says. If you are new to a category, it is probably lower.
- Keep structure decisions on the table. If your content sits in a folder under someone else's domain and AI visibility matters to the business, a subdomain is now a slightly stronger option than it was a year ago.
#The broader shift
Search visibility used to be treated as a property of a page and a query. AI answers already made it a distribution across engines and reruns. Preferred Sources adds a third axis: the reader. Google has told publishers the signal is coming, told them how to collect it, and given them no report on it. The teams that start collecting selections now, while the badge is still only a highlight, will have the audience in place if it becomes a ranking input.
Related field notes
September 24, 2026 · 5 min
Google's goto links tax every SERP copy
Google now wraps result links in google.com/goto redirects. The target is scrapers, which includes your rank tracker and some AI answer pipelines.
September 23, 2026 · 5 min
A browser agent is not a crawler
Agentic browsing runs inside the user's own session, so robots.txt, bot allowlists and crawler analytics all miss it entirely.
September 23, 2026 · 5 min
Your AI Overview loss depends on the baseline
One Wikipedia study revised its AI Overviews traffic loss by a factor of three. What changed was the baseline.
Share or discuss
New posts, no spam. Roughly monthly. Unsubscribe with one click.