Vivienne Goizet (BILD Group) took us into the engine room of Germany’s largest publisher to answer the billion-dollar question: “Is the AI Overview (AIO) apocalypse happening?“

Her thesis is refreshing in its honesty: It’s complicated. While US studies scream “panic,” the German reality is nuanced. Goizet argues that AIOs aren’t a blanket “traffic killer” but a surgical strike on specific verticals. By triangulating data from Sistrix, Peak AI, and internal BILD analytics, she paints a picture of a fragmented landscape where “Health” is under siege, but “Politics” remains surprisingly untouched.

Approaching the topic:

The “Health” vs. “Politics” Divide

Goizet’s data reveals a massive vertical disparity.

• The “Siege” Sector: Health queries (vitamins, diseases) see AIO saturation of up to 80% in some tools. This confirms that “Your Money Your Life” (YMYL) topics are prime territory for Google’s direct answers.
• The “Safe” Harbor: Surprisingly, Politics sees significantly less AIO interference. Goizet suggests this might be Google treading carefully around misinformation, or perhaps just the nature of the queries in that vertical.
• The Takeaway: Panic is optional. If you run a political news desk, the AI threat is currently lower than for your lifestyle colleagues. So, as we say in German: “Keine Panik auf der Titanic – Wasser ist für alle da!” (essentially: “No panic on Titanic – there is water for everyone!” – and you thought Germans don’t have humor … 😉 )

The “Statista” Monopoly

A fascinating insight from her “Peak AI” tracking: When users ask AI for the “most important news sources,” it doesn’t list BILD, Spiegel, or Tagesschau.

• The Winner: It cites Statista.
• Implication (!): This highlights a terrifying disconnect between “Brand Authority” in the real world and “Source Authority” in the LLM. If the AI prefers a data aggregator over a legacy publisher for news queries, we have a fundamental branding problem in the dataset.

The “Question” Trigger

Goizet identified a clear pattern for when AIOs trigger.

• Pattern: Queries starting with “What is…”, “Definition…”, or “Simply explained…” are almost guaranteed to trigger an AIO.
• Strategy: If you are writing “Explainer” content, you are competing directly with the machine. Conversely, “Navigational” queries (e.g., specific restaurant names) remain AIO-free zones.

The “Old News” Gap

Addressing a key question on news timeliness, Goizet confirmed that AIOs and News Boxes rarely compete directly in real-time.

• The Observation: AIOs tend to appear only when the “Top Stories” box is 8-10 hours old.
• The Insight: This suggests a “freshness gap.” For breaking news, the publisher still owns the slot. The AI takes over once the news becomes “knowledge.”

My take on the talk

Vivienne’s talk was a brave look under the hood of a machine that is still being built. What I appreciate most is her transparency: She didn’t try to force a clean narrative where there isn’t one. The data is messy because Google is testing in real-time. Her observation about Statista winning the “News Source” citation is a wake-up call. It proves that being a household name doesn’t automatically translate to “LLM Salience”. We need to understand why the model prefers a sterile data provider over a journalistic brand.

Strategically, her distinction between the “Freshness Phase” (Publisher wins) and the “Knowledge Phase” (AI wins) gives us a tactical roadmap. We own the “Now”. The “Forever” is being contested.