Google is hiring engineers and a Director-level designer to generate and recommend AI-generated content (AIGC) inside Google Discover, the personalized feed reaching more than 500 million daily active users. Job listings on Google Careers describe teams that generate, quality-check and recommend AI content, plus a design lead to turn Discover into a “proactive, hyper-personalized concierge experience,” according to Search Engine Journal’s October 11, 2026 report. One of those listings was first spotted by Damien Andell, who shared it on X.
What do the Google Discover job listings actually describe?
A Staff ML Engineer listing states the Discover Feed Recommendation team “leads by recommending AI-generated content (AIGC) to fulfill user interests.” The role covers topic finding, content understanding, and quality control for that content pipeline. Two additional listings expand on scope and seniority.
- Staff ML Engineer: builds agentic flows (automated multi-step AI pipelines) that generate content for what Google calls “proactive surfaces,” naming Discover and Notifications. Responsibilities include checks for factuality, freshness, coherence, and tone, plus feedback loops designed to surface engaging stories globally, per Search Engine Journal.
- Senior Staff ML Software Engineer: a more senior role covering AI-generated content quality across major surfaces in the Discover feed. The listing states the team uses AI-generated content across those surfaces.
- Principal UX / Director-level design lead: based in Mountain View or New York, with a U.S. base salary of $275,000–$382,000 plus bonus and equity. The listing describes building a concierge experience that helps users “make progress on their goals” and move “from inspiration to action,” working across YouTube, Shopping, Chrome, and Android teams.
The design listing also asks the hire to “balance AI capabilities while keeping human voices and sources at the center,” and lists experience building value exchanges with creators and publishers as a preferred qualification. None of the listings includes a timeline or specifies how much of the Discover feed AI content would fill, according to Search Engine Journal.
What has Google already shipped for Discover?
The hiring follows a run of recent updates to the feed. In October 2025, Google launched AI previews on trending topics in the U.S., South Korea, and India, with each preview linking to content on the web, per Search Engine Journal. In August 2026, Google announced users would be able to tune their Discover feed by typing specific interests directly into the app, as described in a blog.google post dated August 20, 2026. In September, Robby Stein, VP of Product for Google Search, announced a test of a “Dive deeper” button that opens a short topic overview with links to related stories, starting with videos. Stein said Google planned to test multiple designs for the feature in the weeks following his Sept. 18 post.
The August 20, 2026 blog.google post also announced a “Preferred Sources” button that publishers can embed on their own pages. When a reader clicks it, the site is added as a Preferred Source on Google, making it more likely to appear in Top Stories, AI Overviews (Google’s AI-generated answer panels in search results), and AI Mode. More than 600,000 unique sources had been selected by users as of that post.
What does the Search Console Discover report now track?
Search Console includes a Generative AI performance report for Discover. As of August 31, 2026, Google says it has rolled out these insights to all websites worldwide, per Google’s support documentation. The report tracks impressions, the number of times a link to your site appeared in Discover’s generative AI features, grouped by page, country, and date.
- The Pages dimension shows the canonical URL (the primary, indexed version of a page that Google treats as the definitive copy) that served as the source of information shown in a generative AI feature. This may differ from the page a user lands on after clicking.
- Impressions are the only metric described in the report’s help documentation. One impression is counted per result per session, even if a user scrolls away and back.
- Data from Search Labs (Google’s opt-in experimental features program) is not included.
- Sites with too few impressions may not see the report at all.
Whether a site appears in Discover’s generative AI features is controlled through the Search generative AI control in Search Console, the same setting that governs AI Overviews and AI Mode. Opting out keeps a site’s links and content out of all Search generative AI features, while content from other sites remains available, per Google’s help documentation.
Before adjusting your Discover strategy, check these first
- Confirm your site is not opted out of Search generative AI features in the Search Console settings panel.
- Open the Generative AI performance report for Discover and check whether impression data is available for your property.
- Note which pages generate impressions, since the Pages dimension shows source canonical URLs, not the click-destination URLs.
- Verify that key pages have correct canonical URLs, as the report attributes impressions to the canonical rather than the URL a visitor actually lands on.
- Consider adding the Preferred Sources button (code available in Google Search Central documentation) to high-traffic editorial pages.
- Review article schema markup (structured data that tells Google the type, author, and date of your content) on pages you want Discover to surface.
What this means for AI-search visibility
The clearest way to read these listings: Google is building a system that writes content for Discover, not just recommends content from publishers. That changes what showing up in Discover actually delivers.
Picture a reader who opens the Google app looking for weekend travel ideas. Today, they see cards from travel blogs and tap one. In the version these listings describe, they may see an AI-generated briefing that pulls from several travel sites without a click going to any of them. The travel blog’s URL may appear in the Discover generative AI report as a source. The visit may never come.
The listings support this reading. The Staff ML Engineer listing explicitly describes building agentic flows that “generate content” for Discover as a proactive surface. The quality-check pipeline covering factuality, freshness, and tone implies content being created, not just curated from existing articles.
The practical implication: impressions in Discover’s generative AI features are worth tracking now, before any click relationship is established. That impression count is currently one of the few places where you can measure how often AI-generated content draws from your pages. In the same way that ranking in AI Overviews without receiving a click still signals that your content is treated as a credible source, appearing in Discover’s AI features is a form of AI-search visibility even when it produces no session.
The Preferred Sources button is the most concrete publisher lever available today. Google says readers who mark a site as preferred find it more easily in Top Stories, AI Overviews and AI Mode; in Hingewise’s view, that preference may also carry into an AI-assisted Discover feed.
One dynamic the listings raise but leave open: whether the AIGC quality pipeline will treat publisher-provided structured data as a source selection signal. The pipeline’s emphasis on factuality and freshness checking suggests it needs external reference points to calibrate against. In Hingewise’s view, publishers whose content is machine-readable, clean schema markup, clear article metadata, stable canonical structure, are more likely to be drawn on as source material than those whose context lives in page layout rather than structured fields. The listings do not state this; it is an inference from what the quality-check criteria would require.
The metric to watch over the next several months is the Discover generative AI impression trend in Search Console, particularly for pages that perform well in standard Discover but have not yet appeared in AI features. A consistent gap between the two may point to structured data coverage as the variable worth addressing first.
Sources: Search Engine Journal, October 11, 2026 · Google Search Console Help: Generative AI performance report (Discover) · Google Blog, August 20, 2026
By Lam Nguyen · Hingewise
This article was drafted with AI assistance from the sources cited above and checked against them before publishing. The "What this means for AI-search visibility" section is Hingewise's editorial analysis, built on the frameworks and positions we use in our GEO work. How we work: Editorial policy.