Google is building a quality bar for AI content. Marketing teams can meet it first

Google is hiring engineers to generate AI content for Discover, and others to check its quality. 

The standard inside those listings deserves the attention of every marketing team that publishes with AI. Google defines acceptable AI content for its own surfaces, and its public guidance for site owners uses the same vocabulary. Teams that adopt that standard now will publish with confidence as AI-led discovery grows.

What the listings describe

Three roles carry the story. A Staff ML Engineer role on the Discover Feed Recommendation team covers quality control, topic finding and content understanding for AI-generated content, including agentic flows that produce content for Discover and Notifications. A Senior Staff role focuses on the quality of AI-generated content in the feed. A Principal UX role, with a US base salary of $275,000 to $382,000, leads the design of a “proactive, hyper-personalized concierge experience”.

Discover is already heading that way. Google began adding AI previews to trending topics in October 2025, in the US, South Korea and India. On 18 September 2026 it started testing a Dive deeper button that opens a short topic overview, beginning with video. Marfeel measured AI summaries filling 51% of Discover positions in the US, Brazil and Mexico in December 2025.

A job listing describes intent. Treat these as a signal of direction, with launch timing still to be confirmed.

Why Google’s checks matter to marketers

Factuality, freshness, coherence and tone are the four checks Google lists for AI content in its own feed. They mirror the questions a marketing editor asks before publishing: is it correct, is it current, does it hold together, does it sound like us.

Brands publishing AI-assisted content face the same four tests, and readers apply them instinctively. A wrong statistic, an outdated reference or an off-brand tone costs trust, whether the page surfaces in Discover, an AI Overview or a client’s inbox.

The design listing also asks its hire to keep human voices and sources at the centre of the feed, and names experience building value exchanges with creators and publishers as a preferred qualification. An AI-led feed needs original expertise, named authors and verifiable sources to draw from. Those are assets a marketing team can build today.

What Google’s October guidance asks of you

On 1 October, Google updated its guidance on generative AI content. Generative models predict likely word sequences, so their output can contain hallucinations, and the page calls it “critical to manually factcheck and review all AI-generated content” before publishing.

The review extends to title elements, meta descriptions, structured data and image alt text, because these can appear in search results. Teams that automate metadata as a separate step now have a checking step to add.

The same page describes generative AI as useful for research and for adding structure to original content. It warns that generating many pages that add little value for users may breach the scaled content abuse policy. It also suggests sharing how a piece of content was created, which gives readers context.

A five-step review process for AI-assisted content

1. Source every claim. Attach a primary source to each statistic, date and named entity before the draft moves to review.
2. Check freshness. Confirm that every figure and product reference reflects the current month, and replace older material with a dated source.
3. Read for coherence. One editor reads the full piece for a single argument, consistent terms and a logical order.
4. Match the voice. Compare the draft with the brand voice guide and edit tone line by line.
5. Review metadata and context. Check the title, meta description, schema and alt text against the published page, and add a short note on how AI assisted where readers benefit from knowing.

What to track in Search Console

Since 3 June 2026, Search Console has offered generative AI performance reports, with a separate report for Discover. They count impressions of your URLs in AI features, grouped by page, country and date, and the Discover report combines all its AI features into one count.

Pair the page list from the report with clicks from the main Performance report over the same dates. Pages with strong AI impressions and few clicks show where content informs readers inside Google and where a stronger reason to visit the page would help.

The standard to adopt

Google’s listings describe a feed where AI creates content and engineers check it at scale. Marketing teams that apply the same four checks to their own AI-assisted work meet the standard Google is building for itself. Review discipline earns that content its credibility, and the next draft in your queue is the place to start.