The test.
Wave 10 put three DAX industrial OEMs under their canonical big-ticket procurement task: the question an Asset Manager, a Klinikvorstand, or a Werkleiter asks before an RFQ goes out. One frozen task per brand, public, locked before measurement. 18 datapoints per brand, across OpenAI, Perplexity, and Anthropic, in German [S1]. The task slate is set by the public Task Selection rule [S4].
- Siemens Energy — shortlist an H-class gas turbine (SGT5-9000HL) for a 600 MW Combined-Cycle renewal. Close: an RFQ via the project EPC.
- Siemens Healthineers — shortlist a 3T MRI system (MAGNETOM class) for a tertiary-care hospital. Close: an RFQ via hospital procurement.
- GEA Group — shortlist a multi-stage spray-drying line (Niro, 8 tonnes per hour) for an industrial dairy. Close: an RFQ via the process-engineering firm.
What we measured is discovery: does the agent name the brand, and does it surface what the buyer needs next. AI Visibility 41.9 for Siemens Energy, 33.3 for Siemens Healthineers, 40.3 for GEA Group [S1].
One caveat, stated once for all three. Single task per brand, three-provider track, and AI Usability this wave is a structurally derived interim value, not a fleet-measured profile; the per-breed fleet pass lands in Wave 10b. The read is directional, Confidence C [S1].
The OEMs surface. The integrator that builds the project does not.
Siemens Energy comes up. The EPC that specifies the turbine doesn't.
Siemens Energy is the German national champion in heavy power-generation equipment, spun off from Siemens AG in 2020. Of course AI names it. There is no version of the H-class gas-turbine question where it isn't on the shortlist. The score isn't the story.
The story is on siemens-energy.com.
The SGT5-9000HL spec sheet is there. The Combined-Cycle efficiency benchmark above 64% gross at base load, the 50 Hz portfolio identity, the reference-plant register. The page reads as an engineer expects it to read: serious, factual, deep. An asset-development manager who clicks through gets the technical argument.
What's not on the page: the project the turbine sits inside.
A European 600 MW CCGT placement typically lands with an EPC consortium. Técnicas Reunidas with Ansaldo Energia on RWE's hydrogen-ready plants. Metlen alongside Siemens Energy on the Polish and Romanian builds. The EPC writes the technical specification, runs the panel of three or four OEMs, signs the contract that the turbine sits inside. Sometimes the OEM wins the whole plant turnkey instead, as Siemens Energy did at Leipheim. Either way, the project layer decides [S3].
The agent never surfaces that layer. It names Siemens Energy, presents the H-class efficiency story, and stops. The shortlist looks like the decision. The decision is one layer down.
The turbine is visible. The project that buys the turbine is not.
Score: AI Visibility 41.9 · AI Usability ~30 (structural interim) · Composite 3.6 / 10 [S1].
Siemens Healthineers sells the MAGNETOM. The procurement department signs the contract.
Siemens Healthineers is a top-three global medical imaging OEM, taken public by Siemens in 2018. Of course AI names it on a 3T MRI question. MAGNETOM Cima.X, MAGNETOM Vida, MAGNETOM Lumina. The MAGNETOM brand is on the shortlist by construction. The score 33.3 confirms it [S1].
That number is not the interesting one either.
The interesting one is the clinical-evidence library. Siemens Healthineers carries the BioMatrix sequence technology, the syngo.via clinical workflows, the published case series for high-field neuroimaging. For a Klinikvorstand making a multi-million-euro imaging-suite decision, that is the buyer argument. That is why the procurement signs.
What surfaces in the AI answer: "Siemens Healthineers is a German medical-imaging OEM headquartered in Erlangen, with a strong MAGNETOM MRI portfolio." A Wikipedia line. A fact.
What doesn't surface: the clinical-evidence argument. The actual reason a tertiary-care neuroimaging suite signs Siemens Healthineers over the GE HealthCare or Philips alternative.
And the procurement layer sits underneath. Even if a Klinikvorstand reads the right page, finds the right clinical story, decides Siemens Healthineers is the right MAGNETOM for the suite, the contract still runs through hospital procurement and, in larger clinic networks, through a group purchasing organization [S3]. The brand doesn't transact. The procurement department transacts.
The MAGNETOM is visible. The Beschaffungsabteilung that signs the contract is not.
Score: AI Visibility 33.3 · AI Usability ~30 (structural interim) · Composite 3.4 / 10 [S1].
GEA Group owns the Niro spray dryer. The dairy-plant builder owns the project.
GEA Group is the global market leader in spray drying, with more than 10,000 Niro plants delivered. The Niro franchise is the canonical multi-stage milk-powder line for industrial dairy plants. AI Visibility 40.3 on the unbranded spray-drying question [S1].
The Niro story holds.
The MSD spec is on gea.com. The dairy reference plants are on gea.com. The multi-stage-drying portfolio split between MSD, FSD, and SD is on gea.com. A senior project engineer at a European dairy greenfield who clicks through gets the engineering argument.
What's not on gea.com: the project that buys the dryer.
The 8-tonnes-per-hour decision is not a dryer order. It is a plant project, and the plant builder decides whose dryer sits inside it. In milk powder, the dryer makers themselves compete for that role: Tetra Pak builds complete powder lines around its own Wide Body dryers, SPX Flow's Anhydro plants cover the same scale, and specialist contract dairy-plant builders run the tenders [S3]. GEA makes the dryer, and GEA's rivals make the plants.
The agent never surfaces that plant-builder decision. It shortlists the dryer brand and stops, one layer short of where the contract is written. The pattern rhymes with Siemens Energy and the EPCs, and with Siemens Healthineers and hospital procurement. Different layer. Same architecture.
The Niro line is visible. The plant decision it sits inside is not.
Score: AI Visibility 40.3 · AI Usability ~30 (structural interim) · Composite 3.6 / 10 [S1].
Three procurements, one architecture: the integrator is the layer that doesn't get cited.
The pattern across the three Wave 10 brands is identical at the architectural level.
The OEM brand surfaces in the AI answer. The OEM product page carries the spec, the case studies, the differentiator. The OEM is findable by every breed of AI agent. The close, the project, the contract sits one layer down: with the EPC, the procurement department, the engineering firm, the specialist integrator who turns a brand on a shortlist into an asset on a project schedule.
On the consumer side, this layer already changed hands once. Aggregators like Check24 replaced the human broker; the price-comparison portal owns the consumer-discovery layer now. The brand surface, even when good, is a Layer 2 destination, not a Layer 1 hit.
On the industrial side, the integrator is still alive. There is no Check24 for utility CCGT projects. There is no Verivox for hospital MRI procurement. There is no aggregator that replaces the engineering firm in dairy-plant specification. Project specification is too high-context, too relationship-priced, too judgment-loaded for an aggregator to crunch. Yet.
Siemens AG, the parent brand, is the Wave 2 anchor at Composite 5.6 [S2]. It sits in an adjacent universe: factory automation, SCADA, rail. Siemens AG's industrial customers buy parts and systems, not projects. The intermediary layer is thinner, the OEM transacts more directly with the operator, and discovery and close converge closer. Siemens Energy, Siemens Healthineers, and GEA Group sit on the other side of that line: project-specification industries with a permanent intermediary layer that AI agents don't see.
Aggregators replaced the consumer broker. AI has not replaced the industrial integrator. Yet.
That word, yet, is the open question.
So the work is on the surface, and the reader changed. The agent surface keeps widening, and the integrator layer will get tested. The brands that make that layer visible, that name the EPC, the procurement path, the engineering firm on their own page, evidence-anchored, are the ones an agent can recommend as a buyable answer, not just as a citation [S4].
