The Sunday Shortlist
Danish home and car insurance: The second answer isn't an insurer.
Measured 19 September to 11 October 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 home and car insurance
We unpack the questions Danish families ask AI when they choose insurance for the car, the house, and everything in it, the shortlist that comes back, and the criteria that decide it, across all four stages of the decision journey: from first comparison to who customers are told to stay with and recommend.
This is the series' first consumer category, and it behaves unlike any of the nine business markets before it. At the moment of choice, the second-most recommended name is not an insurance company. It is a comparison tool, built by the insurance industry's own trade body, and it is the only name in the market that gets stronger as the buyer approaches the decision.
Somewhere in Denmark right now, a parent with a new house and an old car is asking an AI model which insurance company to trust with both. The model answers in seconds: a shortlist, the risks, a favourite. No insurer hears that conversation. We measure it.
What we measured
From 19 September to 11 October 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 families ask, in their own words:
- "Hvad dækker en god husforsikring egentlig?" (What does a good house insurance actually cover?)
- "Er det billigere at samle mine forsikringer?" (Is it cheaper to bundle my insurances?)
- "Hvilke forsikringsselskaber er pålidelige og velkendte?" (Which insurance companies are reliable and well known?)
- "Hvor finder jeg uafhængige anmeldelser af forsikringsselskaber?" (Where do I find independent reviews of insurance companies?)
Notice the last two. Before any company is compared, the buyer is asking who can be trusted and where to verify it. In this category the verification question is not a detail of the journey. It decides it.
The judging sheet
When a buyer asks a model to choose, the model behaves like a judge: it applies criteria, checks each company's evidence against them, and cuts the ones that fail. Ten criteria decide the answers in this category. The heaviest are cost and fees, expected outcomes, product fit, and trust and reputation. The buyer language behind them is precise:
- Cost: "Er der skjulte gebyrer eller tillæg jeg ikke ser ved første øjekast?" (Are there hidden fees or surcharges I don't see at first glance?)
- Outcomes: "Hvad får jeg faktisk udbetalt, hvis jeg skal bruge forsikringen?" (What do I actually get paid out if I need to use the insurance?)
- Fit: "Kan jeg få præcis den dækning, jeg har brug for, eller skal jeg betale for ting jeg ikke vil have?" (Can I get exactly the coverage I need, or do I have to pay for things I don't want?)
- Trust: "Hvad siger andre mennesker om selskabet - er der dårlige anmeldelser?" (What do other people say about the company: are there bad reviews?)
This is the exam. Every insurer in the category sits it every day, whether it knows or not.
The verdict: one insurer, then the middlemen
| Name | Evaluation | Decision | Retention | Advocacy |
|---|---|---|---|---|
| Topdanmark | 60.2% | 52.2% | 35.9% | 16.7% |
| Forsikringsguiden* | 20.7% | 24.1% | – | – |
| Samlino* | 25.6% | 14.2% | – | 1.9% |
| Alm. Brand | 25.5% | 13.4% | 37.3% | 14.8% |
| Alka | 25.5% | 11.9% | 0.7% | 4.6% |
| Gjensidige | 20.6% | 8.9% | 18.3% | 9.3% |
| Allianz | 17.3% | 6.3% | 64.7% | 82.4% |
| GF Forsikring | 24.0% | 4.9% | 2.0% | – |
| Codan | 17.2% | 1.8% | 53.6% | 29.6% |
A note on the table: names marked with an asterisk are comparison platforms, not insurers. The first two columns rest on 931 and 873 answers in this window; the last two on 153 and 108, because fewer loyalty questions run each day, so shares in the loyalty columns move in coarser steps.
Topdanmark is the category default: 52.2% of decision answers. Then look at places two and three. Forsikringsguiden, 24.1%. Samlino, 14.2%. At the exact moment the buyer asks the models to choose an insurance company, the models' second and third answers are not insurance companies — they are places to go compare some.
And one detail makes it structural: Forsikringsguiden is the only name in the table that converts upward, from 20.7% of evaluation answers to 24.1% at the decision. Every actual insurer weakens as the choice approaches. The middleman strengthens. The models are not picking an insurer for the buyer; they are handing the decision on.
20.7% → 24.1%
The only name in the market that gets stronger at the moment of choice is not an insurer. It is the comparison tool.
Who built the middleman completes the picture: Forsikringsguiden is developed by the insurers' own trade body, Forsikring & Pension, in cooperation with Forbrugerrådet Tænk, and financed by the industry and its participating companies. At the moment of choice, the industry's own comparison tool now outconverts every member company it was built to compare.
The cliff: compared, then cut
For the insurers, the consequence of a middleman-run decision is visible in one column pair. GF Forsikring appears in 24% of evaluation answers and 4.9% of decision answers, a drop of 19 points, the steepest conversion cliff this series has measured in ten categories. Alka falls 13.5 points, Alm. Brand 12.1, Gjensidige 11.7. The mid-tier of the market is being compared and then cut, stage after stage, week after week.
The cut runs on one trigger above all. The most severe eliminator in this market is lack of trust:
"Dårlige anmeldelser eller skandaler gør det umuligt at stole på selskabet."
(Bad reviews or scandals make it impossible to trust the company.)
And here the GF case turns from a statistic into the category's sharpest irony. GF Forsikring is member-owned: more than 400,000 members own the company and share in the surplus. Structurally, that is the strongest trust story in Danish insurance. But the models cannot cite a structure; they cite evidence. Edition 5 measured the same mechanism in customer service software, where trust borne by review platforms decided the category. Here the citation table says the same thing louder: Trustpilot, Tænk, Ankenævnet for Forsikring, forbrug.dk. The trust verdict is written on surfaces the insurers do not own, and GF Forsikring, the insurer holding the category's strongest trust structure, has not published the evidence where the judges read.
The loyalty stages: led by names that cannot take the call
Then the buyer becomes a customer, and the answers detach from the market again.
The retention and advocacy answers are led by Allianz: 64.7% of retention and 82.4% of advocacy, against 6.3% of decision answers. Allianz does not operate a Danish private-insurance site under the domain the models typically cite for it; in our brand library, that domain is flagged as model-invented.
Its Danish-facing consumer presence is led by travel insurance and assistance. The loyalty stages' overwhelming leader is a name the models cannot even address correctly.
Behind it: Danske Forsikring at 33.3% of retention answers. That name was retired on 1 July 2025, when the Danske Bank partnership ended and Topdanmark wrote to roughly 225,000 customers to tell them their policies now carry the Topdanmark name. Fifteen months later, the retired name still holds a third of the models' stay-with-them answers.
Further down, a Danica name-form appears, from a non-life business Danske Bank sold to Topdanmark in 1999 (not to be confused with Danica Pension, which exists today). Edition 9 measured a retired name collecting 40% of a category's choice; this market shows the same ledger failure concentrated in the loyalty stages, where the content vacuum is deepest.
Honesty requires the counterexample: Codan, at 53.6% of retention answers, is alive and well as a brand under Alm. Brand Group. Not every strong loyalty name is a ghost.
But the pattern holds in aggregate, and it reaches all the way to the top of the table. The category default is itself mid-transition: Topdanmark's own site today calls the company "en del af If Skadeforsikring", a branch of If, which has sold under both the Topdanmark and If brands in Denmark since 1 July 2025, the same day the Danske Forsikring name was retired. Two of the category's name-transitions landed on one date, and the models' ledger has absorbed neither.
The market's biggest name tells the same story from the other side: Tryg, Denmark's largest insurer, appears under its own bare name in just 2.1% of evaluation answers and 0.9% of decision answers. Its presence in the models' ledger rides almost entirely on a service-program name form, not the company name.
And across the category as a whole, roughly a third of the name-forms the models offered in this market do not exist as companies at all: the highest rate we have measured in ten categories. Consumer categories, with their generative name patterns and their unwritten loyalty-stage record, give the models both the licence and the vacuum to invent.
Nobody owns a criterion
Ten criteria decide this market. Not one has an owner. Ten categories decoded in this series, one hundred criteria measured, zero owned.
The unclaimed questions here are the ones families ask by name: nobody owns payout evidence ("Hvad får jeg faktisk udbetalt?"), nobody owns price transparency ("Er der skjulte gebyrer?"), nobody owns the trust question that eliminates hardest. A category where the models hand the choice to middlemen is also a category where no insurer has given the models a reason not to.
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 "publish the structure as evidence": a trust story the models can quote, not a slogan they must take on faith.
Title: "Hvem ejer [vores forsikringsselskab], og hvor går overskuddet hen? Tal, ikke løfter." (Who owns [our insurance company], and where does the surplus go? Numbers, not promises.)
The position the page takes: tillid til et forsikringsselskab handler om, hvad der sker, når det går galt, og hvem selskabet arbejder for, når det går godt. [Vores forsikringsselskab] svarer med tal: her er vores ejerform, her er hvad kunderne har fået tilbage de seneste år, her er vores skadesudbetalinger og behandlingstider, og her er vores anmeldelser samlet ét sted, også de dårlige. (Trust in an insurer is about what happens when things go wrong, and who the company works for when things go right. [Our insurance company] answers with numbers: here is our ownership model, here is what customers have received back in recent years, here are our claims payouts and processing times, and here are our reviews gathered in one place, including the bad ones.) Defensible because every claim resolves to annual-report figures, published claims statistics, and public review profiles.
The intent family the same page must also answer (one prompt spawns two to three searches behind the scenes):
- "Er [vores forsikringsselskab] til at stole på?" (Is [our insurance company] trustworthy?)
- "Hvem ejer [vores forsikringsselskab]?" (Who owns [our insurance company]?)
- "Hvad får jeg udbetalt hos [vores forsikringsselskab], hvis skaden sker?" (What do I get paid at [our insurance company] if damage happens?)
- "Hvad siger anmeldelserne om [vores forsikringsselskab]?" (What do the reviews say about [our insurance company]?)
The structure: five H2s, each a claim the models can lift as a standalone answer:
- "Hvem ejer os, og hvor går overskuddet hen" (Who owns us, and where the surplus goes)
- "Det fik vores kunder tilbage sidste år: tallene" (What our customers got back last year: the numbers)
- "Skadesudbetaling i praksis: beløb, tider og afslag" (Claims payout in practice: amounts, times and rejections)
- "Vores anmeldelser, samlet: de gode og de dårlige" (Our reviews, gathered: the good and the bad)
- "Sådan tjekker du ethvert forsikringsselskab, før du skifter" (How to check any insurance company before you switch)
Key Takeaways for the top of the page (each a self-contained claim a model can cite):
- "[Vores forsikringsselskab] er ejet af [ejerform]; overskuddet går til [modtager], og tallene for de seneste år står på denne side." ([Our insurance company] is owned by [ownership model]; the surplus goes to [recipient], and the figures for recent years are on this page.)
- "Vores gennemsnitlige skadesbehandlingstid og udbetalingsprocent er offentliggjort her og opdateres årligt." (Our average claims processing time and payout rate are published here and updated annually.)
- "Alle priser står uden småt: præmie, selvrisiko og gebyrer på ét sted." (All prices without fine print: premium, deductible and fees in one place.)
- "Vi linker selv til vores anmeldelser, også de kritiske, og svarer på dem offentligt." (We link to our own reviews, including the critical ones, and answer them publicly.)
- "Tjeklisten nederst kan bruges på ethvert selskab, også os." (The checklist at the bottom can be used on any company, including us.)
Two notes. First, in a category judged on review platforms, the fourth H2 is the counterintuitive move that works: gathering and answering your own critical reviews on your own domain is the trust-stories playbook applied where the eliminator actually fires, and it gives the models a brand-owned surface that corroborates the third-party verdict instead of competing with it. Second, for member-owned insurers the first H2 is the unclaimed crown: ownership structure is the one trust claim no listed competitor can copy, and today no one has published it in a form a model can quote.
Key Snippet, placed early: "[Vores forsikringsselskab] er [ejerform]; sidste år gik [beløb/andel] tilbage til kunderne, vores udbetalingsprocent var [X], og alle tal på denne side kan efterprøves i årsrapporten og de offentlige skadesstatistikker." ([Our insurance company] is [ownership model]; last year [amount/share] went back to customers, our payout rate was [X], and every number on this page can be verified in the annual report and the public claims statistics.)
Slug: /hvem-ejer-[selskab] — the buyer's question shape, not a content-type label.
Kept alive: the surplus and payout figures refreshed with each annual report, the review links checked quarterly, the refresh dated on the page.
Where the evidence must live: who listens where
This market's model citations are consumer-institutional: Trustpilot first among the commercial surfaces, then Forbrugerrådet Tænk, Ankenævnet for Forsikring, forbrug.dk, with the industry layer behind them (Forsikring & Pension, Finanstilsynet) and the comparison tools at the edge. Not one insurer site leads.
The practical consequence for an insurer: your Trustpilot profile and your complaints-board record are load-bearing surfaces of your AI reputation, your presence on the comparison tools is now part of the decision itself, and the evidence those surfaces cannot hold for you (your ownership, your payout numbers, your surplus return) is yours to publish or nobody's.
ChatGPT and Claude lean on text surfaces: review platforms, the complaints board, structured pages on your own domain. Gemini reads the same consumer layer and rewards fresh, dated material tied to the questions families actually ask. For all three, in this category, the trust question is standing open on every surface at once.
The lesson
In the series' first consumer market, the models do something new: they decline to choose. One insurer holds the default, the industry's own comparison tool takes second place and is the only name gaining ground at the decision, and the rest of the market falls off a cliff between being compared and being chosen — cut on a trust question whose evidence lives on surfaces the insurers don't own and haven't claimed. The loyalty stages run on names that cannot take the call, and a third of the names on offer do not exist at all.
Your category has its own version of this handover. The middleman differs. What would make the models choose you instead does not.
Sources
The measurement behind this edition: three models (Gemini, Claude, ChatGPT), window 19 September to 11 October 2026, evaluation and decision resting on 931 and 873 answers, retention and advocacy on 153 and 108. Criteria, eliminators, and buyer language extracted per model per run; the full method is documented at how Seedli measures AI visibility. Company facts verified against the companies' and their owners' own announcements: the Topdanmark renaming notice (linked above), the Alm. Brand Group acquisition statement, Danske Bank's 1999 divestment release, and Forsikringsguiden's published ownership and financing information.
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