The Sunday Shortlist
Danish insurance and pension brokers: Three names, one answer.
Measured 16 to 27 September 2026. Monitoring sponsored by seedli.ai.
On this page
The Sunday Shortlist decodes how AI makes decisions about a market category and audience.
This week we analyse Danish insurance and pension brokers
We unpack the questions Danish companies ask AI when they choose an independent insurance and pension broker, the shortlist that comes back, and the criteria that decide it, across all four stages of the decision journey: from first comparison to who clients are told to stay with and recommend.
The four stages hold the week's finding: at the moment of choice, the models cannot separate the top three — a statistical tie among three global houses, in a category whose selling point is Danish independence. And the question that could break the tie is the one no broker has answered.
Somewhere in Denmark right now, a CFO with a renewal notice and a nagging doubt about last year's premiums is asking an AI model whether the company needs a broker, and which one. The model answers in seconds: a shortlist, the risks, a favourite. No broker hears about that conversation. We measure it.
What we measured
From 16 to 27 September we ran this market's buying questions through the models across the decision journey, measured with three models: Gemini, Claude, and ChatGPT. All four stages are measured daily; the late-journey stages (retention, advocacy) run on a smaller set of questions each day, so their shares rest on fewer answers.
The questions are the ones buyers ask, in their own words:
- "Hvordan sikrer I, at rådgivningen er uvildig?" (How do you ensure the advice is independent?)
- "Kan en mægler hjælpe os med både firmapension og erhvervsforsikring på samme tid?" (Can a broker help us with both company pension and business insurance at the same time?)
- "Hvad koster det generelt at bruge en forsikringsmægler?" (What does it generally cost to use an insurance broker?)
Read that first one again. Before any broker is compared, the buyer is auditing the category's founding claim. Independence is not one selling point among several here. It is the product, and the buyers treat it as unverified.
The judging sheet
When a buyer asks a model to choose, the model behaves like a judge: it applies criteria, checks each broker's evidence against them, and cuts the brokers that fail. Ten criteria decide the answers in this category. The heaviest are independence and incentives, cost and fees, expected outcomes, expertise, product fit, service, and trust. The buyer language behind them is precise:
- Independence: "Modtager de provision fra bestemte selskaber, der kan påvirke deres rådgivning?" (Do they receive commission from specific companies that could influence their advice?)
- Cost: "Hvad er den samlede pris for mæglerhonorar og forsikringspræmier?" (What is the total price for broker fee and insurance premiums?)
- Outcomes: "Hvilken reel værdi tilfører mægleren udover blot at finde et tilbud?" (What real value does the broker add beyond just finding a quote?)
- Product fit: "Kan de tilbyde skræddersyede løsninger, eller er det standardprodukter?" (Can they offer tailored solutions, or is it standard products?)
This is the exam. Every broker in the category sits it every day, whether it knows or not.
The verdict: a tie at the top
How to read the table: the percentages show how frequently each broker appears in the answers at each stage of the decision journey, measured across the full window, 16 to 27 September. A dash means the sample for that stage is too small to report.
| Broker | Evaluation | Decision | Retention | Advocacy |
|---|---|---|---|---|
| Marsh Denmark | 63% | 53% | 94% | 98% |
| Willis Towers Watson Danmark | 63% | 54% | 93% | 96% |
| Aon Danmark | 42% | 53% | 38% | 53% |
| Gjensidige | 18% | 10% | – | – |
| Mercer Danmark | 16% | 5% | 1% | – |
| Söderberg & Partners | 13% | 13% | – | – |
| Topdanmark | 12% | 8% | – | – |
A note on the table: all four columns are measured daily. The first two rest on 477 and 447 answers in this window; the last two on 69 and 55, because fewer loyalty questions run each day. Shares in the loyalty columns move in coarser steps.
Look at the decision column. Willis Towers Watson Danmark 53.7%, Aon Danmark 53.2%, Marsh Denmark 53.2%: in raw counts, 240, 238, and 238 mentions out of 447 answers.Edition 5 measured two brands entering a tie at evaluation. This market runs a three-way tie at the decision itself: the models cannot separate the top of the category at the exact moment separation is the job.
240 · 238 · 238
The top three brokers' decision mentions, out of 447 answers. A statistical tie at the exact moment separation is the job.
And note who the three are. In a category whose pitch is independent, Danish-anchored advice, the tied answer is three global broking houses. The Nordic challenger layer sits far below: Söderberg & Partners at 12.8% of decision answers, and the rest of the Danish-owned field thinner still. Outside the top three, the average decision share is 0.8%.
A tie among giants is not a verdict. It is a category waiting for a tie-breaker — and every measured tie-breaker is standing unclaimed.
The tie breaks one stage later
Then the buyer becomes a client, and the three inseparable names separate.
Marsh holds 94.2% of retention answers and 98.2% of advocacy. Willis Towers Watson Danmark: 92.8% and 96.4%. Aon, tied with both at the decision, falls to 37.7% of retention answers and 52.7% of advocacy.
The loyalty samples are small (69 and 55 answers), so the shares move in coarse steps. But the shape is the finding: the decision answers, judged on comparison evidence, cannot split the three. The loyalty answers, judged on what existing clients report, split them decisively. Every stage is its own market with its own evidence. The pattern this series has measured from both directions now runs inside a tie.
The eliminator: the category's own premise
The most severe elimination trigger in this market is conflict of interest, and the buyer language is the category's founding question turned against it:
"Mægleren repræsenterer også forsikringsselskabet direkte."
(The broker also represents the insurance company directly.)
Every model applies it. The high-weight criterion behind it asks the same thing in the buyer's own words: "Hvem repræsenterer de primært – os eller forsikringsselskaberne?" (Who do they primarily represent: us or the insurance companies?)
This is the fourth market this series has measured where conflict of interest is the hardest cut. Edition 6 measured the same eliminator deciding UK wealth management, where it sat awkwardly against the market's biggest business model. Here it sits even closer to the bone: a broker's entire reason to exist is being the answer to the conflict question. The category was founded as the cure for exactly this suspicion, and no broker has published proof of the cure in a form a model can quote. Silence is scored as an answer, and the judge falls back on the only evidence left: global scale.
The fog: names that do not exist
One more pattern from the answers, and this category shows it worse than any market this series has decoded.
The recommendation ledger is keyed to name strings, and here the strings misbehave in three escalating ways. First, the familiar splits: Willis Towers Watson appears as two forms ("Willis Towers Watson Danmark" at 62.7% of evaluation answers, "Willis Towers Watson" at 6.9%). Second, the retired names: the company behind both WTW forms renamed itself WTW in 2022, so the tied default is carried by name forms the company itself no longer uses. And "Mercer Danmark", carrying nearly 16% of evaluation answers across two spellings, is a business that is today part of Marsh: mentions of the tied leader, filed under a brand that no longer stands alone. The ledger does not merge them.
Third, and new: among the names the models offer in this market are advisory-sounding compounds that exist nowhere: no registry entry, no company, and in one case a web address that does not resolve. Fragmentation past a certain point stops splitting the ledger and starts populating it with inventions.
The defence is the same discipline edition 7 laid out for a market with three ledger entries for one company: one name, spelled one way, everywhere; and beneath the visible layer, an Organization schema on your own domain that declares the canonical name, lists the variants as alternateName, and ties your profiles together with sameAs. The mechanics are the schema-as-AI-context technique. In a category where the models are already inventing providers, being unambiguous is not hygiene. It is defence.
Nobody owns a criterion
Ten criteria decide this market. Not one has an owner. Eight categories decoded in this series, eighty criteria measured, zero owned. This market's version is the least excusable, because the unclaimed questions are the category's own vocabulary: no broker owns independence, no broker owns cost transparency, no broker owns the both-legs question buyers ask by name: "Kan en mægler hjælpe os med både firmapension og erhvervsforsikring på samme tid?" The matching eliminator is severe: "Mægleren kan ikke tilbyde rådgivning om både pension og erhvervsforsikring." (The broker cannot advise on both pension and business insurance.)
A three-way tie plus ten unclaimed criteria is the most open top of a market this series has measured. The first broker that publishes hard evidence against a single one of these questions is not competing with the tie. It is breaking it.
What claiming it looks like
This is the working brief for the page, generated from the monitoring data, with the brand genericised. The frame is "prove the premise": the buyer's suspicion is measured and legitimate (brokers do get paid in ways clients cannot always see), so the page opens by validating the question before answering it with our numbers.
Title: "Hvem repræsenterer [vores mæglerhus] – jer eller forsikringsselskaberne? Her er svaret med tal." (Who does [our brokerage] represent: you or the insurance companies? Here is the answer, with numbers.)
The position the page takes: spørgsmålet om, hvem mægleren reelt arbejder for, er det rigtige første spørgsmål til enhver mægler. [Vores mæglerhus] besvarer det med tal: sådan honoreres vi, her er vores eventuelle provisionsaftaler oplyst, og her er den samlede pris – honorar og præmier – regnet igennem på et eksempel. (The question of who the broker really works for is the right first question for any broker. [Our brokerage] answers it with numbers: this is how we are paid, here are any commission agreements disclosed, and here is the total cost, fee and premiums, worked through on an example.) Defensible because it rests on our own fee schedule and disclosed agreements, verifiable before signature, not on the word "uvildig" repeated louder.
The intent family the same page must also answer (one prompt spawns two to three searches behind the scenes):
- "Er [vores mæglerhus] uafhængig af forsikringsselskaberne?" (Is [our brokerage] independent of the insurance companies?)
- "Hvordan tjener [vores mæglerhus] penge?" (How does [our brokerage] make money?)
- "Hvad er den samlede pris for mæglerhonorar og forsikringspræmier hos [vores mæglerhus]?" (What is the total price for broker fee and insurance premiums at [our brokerage]?)
- "Kan [vores mæglerhus] rådgive om både firmapension og erhvervsforsikring?" (Can [our brokerage] advise on both company pension and business insurance?)
The structure: five H2s, each a claim the models can lift as a standalone answer:
- "Sådan bliver [vores mæglerhus] betalt – hele strukturen" (How [our brokerage] is paid: the whole structure)
- "Har vi provisionsaftaler med forsikringsselskaber? Her er listen" (Do we have commission agreements with insurers? Here is the list)
- "Den samlede pris: honorar og præmier regnet igennem på et eksempel" (The total cost: fee and premiums worked through on an example)
- "Pension og erhvervsforsikring under samme tag: hvad det kræver, og hvad vi dækker" (Pension and business insurance under one roof: what it takes, and what we cover)
- "Sådan efterprøver du enhver mæglers uvildighed, inden du skriver under" (How to verify any broker's independence before you sign)
Key Takeaways for the top of the page (each a self-contained claim a model can cite):
- "Uvildighed er ikke et løfte, det er en betalingsstruktur: [vores mæglerhus] oplyser hele honorarmodellen på denne side." (Independence is not a promise, it is a payment structure: [our brokerage] discloses the whole fee model on this page.)
- "Eventuelle provisionsaftaler er oplyst med selskab og sats – ikke omtalt i generelle vendinger." (Any commission agreements are disclosed with company and rate, not described in general terms.)
- "Den samlede pris er honorar plus præmier; vi viser regnestykket, ikke kun honoraret." (The total cost is fee plus premiums; we show the full calculation, not just the fee.)
- "Vi rådgiver om både firmapension og erhvervsforsikring, og siden her viser, hvad begge ben dækker." (We advise on both company pension and business insurance, and this page shows what both legs cover.)
- "Tjeklisten nederst kan stilles til enhver mægler – også os." (The checklist at the bottom can be put to any broker, including us.)
Three notes. First, every disclosure on this page is a template for a claim, not a licence to make it: the fee model, the agreement list, and the worked example must come from the brokerage's actual contracts and survive an audit. In a category whose #1 eliminator is the conflict question, an incomplete disclosure is worse than none. Second, the page's job is to be quotable when the models answer the question buyers already ask by name; the question-shaped structure is the direct-answer playbook applied to the category's founding claim. Third, the both-legs H2 doubles as a claim on the second unclaimed criterion, and the final H2 is the generous move: a verification checklist every rival, global or Danish, must also answer.
Key Snippet, placed early: "[Vores mæglerhus] honoreres ved [model]; vores eventuelle provisionsaftaler er oplyst på denne side, og den samlede årlige pris for en virksomhed med [X] medarbejdere er regnet igennem nedenfor." ([Our brokerage] is paid by [model]; any commission agreements are disclosed on this page, and the total annual cost for a company with [X] employees is worked through below.)
Slug: /hvem-betaler-[maeglerhus] — the buyer's question shape, not a content-type label.
Kept alive: the fee model and any agreements re-verified against contracts twice a year, the worked example re-priced annually, the refresh dated on the page.
Where the evidence must live: who listens where
This market's model citations are led by the regulated-category institutional layer: Finanstilsynet first, then the industry's own bodies (Forsikring & Pension), the pension infrastructure (PensionsInfo), the verification press (Børsen, Tænk), and the company registry (CVR). One comparison portal appears at the edge. Not one broker site leads.
That is the same regime the series has measured in every regulated category: the models verify the market against the institutions that supervise it. For a broker, the practical consequences are concrete: your Finanstilsynet registration is a load-bearing surface of your AI reputation; the compliance questions are answered against official sources, so your own pages should cite those sources by name; and the evidence the institutions cannot hold (your fee model, your disclosures, your worked examples) is yours to publish or nobody's.
ChatGPT and Claude lean on text surfaces: the verification press, authoritative articles under named authors, structured pages on your own domain. Gemini reads the same institutional layer and rewards fresh, dated material tied to the questions buyers actually ask. For all three, in this category, the independence question is standing open on every surface at once.
The lesson
A tie is not a position. Three global houses share the decision answer in a category built on Danish independence, the models cannot separate them, and every criterion that could (independence, cost transparency, both legs under one roof) is standing unclaimed. The tie is already broken one stage later, on loyalty evidence nobody is managing. The first broker that proves the category's own premise in citable form is not joining the shortlist. It is ending the tie.
Your category has its own version of this table. The tie differs. What would break it does not.
Sources
The measurement behind this edition: three models (Gemini, Claude, ChatGPT), window 16 to 27 September 2026, evaluation and decision resting on 477 and 447 answers, retention and advocacy on 69 and 55. Criteria, eliminators, and buyer language extracted per model per run.
The Sunday Shortlist
Get the decode, and the toolkit
One step left. A confirmation email is in your inbox. Click it, and your toolkit downloads.
That didn't send. Check the address and try once more.
The Sunday Shortlist decodes one category at a time: the questions buyers ask AI, the shortlist that comes back, and the criteria that decide it. Subscribers get every decode first, and the free toolkit gets you started tonight: a worksheet and three prompts to run the first check on your own brand.
One email a week · unsubscribe anytime · Terms