The test.
Wave 9 put two German reinsurers, Munich Re and Hannover Rück, under one frozen task: the European Property Cat XL Treaty placement a primary insurer runs before its 1 January renewal. 18 datapoints per brand, across OpenAI, Perplexity, and Anthropic, in German [S1]. One task, picked and frozen by the public Task Selection rule [S4].
What we measured was discovery: does the agent name the brand, and does it surface what the buyer needs next. AI Visibility 44.2 for Munich Re, 28.2 for Hannover Rück. Half of agent runs name Munich Re. Less than one in three name Hannover Rück.
What we found on top of the number is a structural gap the score itself doesn't see, and it is the same gap for both reinsurers.
The brands surface. The route to the brands does not.
Munich Re: the capacity is visible, the route to it is not.
Munich Re is one of the two largest reinsurers in the world. Of course AI names it. There is no version of the question where it isn't on the shortlist. The score isn't the story.
The story is on munichre.com.
The capacity statement is there. NatCatSERVICE peril data, dated catastrophe loss curves, the line-of-business taxonomy. The page reads as a CFO expects it to read: serious, factual, deep. A reinsurance manager who clicks through gets the underwriting argument.
What's not on the page: a way to start a treaty submission.
Because there isn't one. Munich Re doesn't accept treaty submissions directly. Property Cat XL placements run through reinsurance brokers. The broker writes the submission, runs it past three or four reinsurers, builds the panel, signs the contract. Munich Re writes capacity. The broker writes the relationship [S3].
The agent never finds the broker step. It surfaces Munich Re, presents the capacity story, and stops. The shortlist looks like a list of names you can call. It is a list of names you can't call.
The capacity is visible. The path to the capacity is not.
Score: AI Visibility 44.2 · Usability not measured (the close runs through brokers).
Hannover Rück: the brand surfaces, its argument does not.
Hannover Rück shows up less than one in three runs. Below Munich Re, below Swiss Re, sometimes below SCOR and Lloyd's. AI Visibility 28.2.
That number is not the interesting one either.
The interesting one is the expense ratio. Hannover Rück runs one of the leanest cost bases in global reinsurance, a structural efficiency edge rating agencies tie to its lower-than-average administration expense. For a CFO building a follow-line panel, that is the buyer argument. That is why the treaty signs.
What surfaces in the AI answer: "Hannover Rück is a German global top-five reinsurer headquartered in Hanover." A Wikipedia line. A fact.
What doesn't surface: the expense-ratio differentiator. The reason a treaty signs.
The agent shows the brand. Not the brand's argument.
And the same broker wall sits underneath. Even if a CFO reads the right page, finds the right line about the cost discipline, decides Hannover Rück is the right follower on this treaty, the placement still runs through Aon, Guy Carpenter, or Gallagher Re [S3]. The brand doesn't transact. The brand persuades.
For a brand that persuades by being lean, surfacing "headquartered in Hanover" instead of "lowest expense ratio in global reinsurance" is the wrong fact at the wrong moment.
Score: AI Visibility 28.2 · Usability not measured (the close runs through brokers).
Allianz: the aggregator already replaced the broker.
Allianz is in the same DAX sector as Munich Re and Hannover Rück. It is also in a different world.
Ask an AI agent in German: "Welche Kfz-Versicherung soll ich 2026 abschließen?"
The list comes back. Check24. Verivox. HUK24 direct. Allianz once in five runs [S2].
Allianz has a Kfz quote-tool that works. It calculates a premium between €420 and €432 across runs, consistent and fast. The brand surface holds. The agent doesn't reach it. Check24 and Verivox absorb the unbranded query before the brand gets named [S2].
On the consumer side, the broker is already dead. Aggregators replaced him. The price-comparison portal owns the discovery layer that a human broker used to own ten years ago. The brand surface, even when it's good, like Allianz's, is a Layer 2 destination, not a Layer 1 hit.
Score: AI Visibility 19.8 · Composite 3.2 / 10. Allianz is the one brand here with a transactional close to score [S2].
Reinsurance doesn't have a Check24. There is no Cat-XL.com that compares treaty capacity across reinsurers. Broker placement is too high-context, too relationship-priced, too judgment-loaded for an aggregator to crunch. Yet.
Aggregators replaced the consumer broker. AI hasn't replaced the reinsurance broker. Yet.
That word, yet, is the whole point.
Named, not yet placed: what the three tell us.
The DAX insurance-finance sector now holds three intermediation modes in the index.
Allianz, consumer cover. Aggregator owns discovery; brand owns surface but reaches buyer second. Strategy: get aggregator-quotable in real time, treat the aggregator listing as the first surface, not the brand's homepage.
Munich Re and Hannover Rück, B2B reinsurance. Brand owns discovery, broker owns the close. Strategy: make the broker layer surface-able. Aon, Guy Carpenter, Gallagher Re as named, evidence-anchored cross-references on the brand's page. Right now they are not.
The fourth mode is the one not in the index yet. The one where AI starts handling broker workflows directly: submission analytics, panel optimization, treaty structuring. That isn't 2026. It is also not 2030. The reinsurance broker is currently the layer between AI and the close. If brand surfaces don't make that broker layer visible, AI will route around it the moment it can. The aggregator replacement on the consumer side took roughly fifteen years. The B2B replacement is a different animal, but the direction is the same.
What every brand in this article shares: the surface was built for humans navigating a known process. The AI is not navigating. It is deciding.
So the work is on the surface, and the reader changed. The reinsurance broker is still the layer between AI and the close, and no brand surface in this wave named it. The brands that put that layer on their own page, named and evidence-anchored, are the ones more likely to be handed back, not just cited.
