A general insurance shortlist is decided on the terms that reduce the cover, not on the premium. By the time a buyer sees a price, an assistant has already resolved their question into six comparisons: sum insured, room-rent and co-pay limits, waiting periods, network hospitals and garages, the claim settlement record, and what the exclusions and riders do. A policy whose answers to those six sit inside a PDF gets described second-hand, or left out of the shortlist entirely.
Six questions, asked in one breath
Health cover is now the largest thing a general insurer sells. IRDAI's 2024-25 annual report, analysed by the Observer Research Foundation in January 2026, puts health premiums underwritten by general and health insurers at INR 1,17,505 crore, a 9.19 percent rise on the year, and 41.42 percent of all gross direct premiums in non-life. Health has overtaken motor as the biggest segment in the category.
That scale matters here for one reason. A category this large is one an assistant has plenty of source material about, and almost none of that material has to come from you.
A prompt like best health insurance plans in India for family does not stay one question. It fans out into six: sum insured, room-rent and co-pay limits, waiting periods, network hospitals and garages, the claim settlement record, and what the exclusions and riders do. The assistant answers all six from whatever it can read, then writes one shortlist. The premium is the last thing it reaches, not the first.
Why do exclusions decide the shortlist?
Because the sum insured is a headline and the limits are the actual cover. A floater with a capped room rent, a co-pay on every claim and a sub-limit on the procedures a family is most likely to need is a smaller policy than its headline says. Buyers work this out. So do engines, because the buying guides they read say it explicitly.
The problem is where those terms live. When inclusions, sub-limits and waiting periods sit only inside a policy wording PDF, the assistant does not skip the question. It infers an answer. And it will sometimes infer in your favour, which is the worse outcome: the buyer arrives believing the cover is broader than it is, and discovers otherwise at the point of claim.
The same failure has a quieter version. A base plan and a rider disagree about a benefit, and the answer states the more generous one. Nobody wrote a false sentence. The engine assembled one out of two true ones.
A claim ratio without its basis is not a fact
The claim record is the last question a family asks and the one that settles it. It is also the number this category publishes worst.
Take the spread first. On IRDAI's 2024-25 figures, the aggregate incurred claims ratio for health insurance was 86.98 percent. That average hides public sector insurers at 100.59 percent, private general insurers at 87.59 percent and standalone health insurers at 68.73 percent. Those are not small differences in performance. They are different businesses, and a buyer reaching for one industry number is reading past all of it.
Now take the basis. The Press Information Bureau, announcing the same year's results in 2026, reported a claims paid ratio that had risen to 87.50 percent. Set beside 86.98 percent, that looks like one measurement confirmed twice. It is not the same measurement. A ratio means nothing without its period, its basis and its source attached, and an engine built to be careful will drop a figure it cannot qualify rather than repeat it.
The Observer Research Foundation calls this the transparency deficit, and notes that the most recent analytical report publicly available on the IIB portal dates back to 2019-20. That gap is the opening. Publish your settlement record with its year, its basis and a link to the filing it came from, and you are supplying the qualified number the engine has been looking for and mostly failing to find.
What does an assistant do with a motor policy?
It runs the same six-part comparison with different nouns. Insured declared value in place of sum insured. Own-damage and third-party cover in place of inclusions. Add-ons in place of riders. Cashless garages in place of network hospitals. And the same closing question underneath: who actually pays.
Motor carries one advantage and one trap. The advantage is that the buyer usually knows their own facts, the car, the city, the year, the claim history, so an answer can get specific quickly. The trap is that the specificity has to land somewhere.
A prompt like best car insurance in India ends with a buyer who has already decided. If your site cannot turn a vehicle and a city into a real premium, the answer hands them to a comparison portal that can, and the portal collects the sale you were shortlisted for.
The six places the answer finds its evidence
An answer about your policy is assembled from sources, and only one of them is yours. Product and policy wordings supply the cover, limits and exclusions, when they are readable. IRDAI filings and circulars carry the weight of the regulator. Comparison portals are where plans are filtered, ranked and quoted, and they are frequently the cited source. Claim and annual reports carry the settlement and grievance record. Financial publishers define the terms buyers search with. Reviews and grievance forums are where claim experience, good and bad, becomes public evidence.
You control one of those six directly. The rest you influence by being accurate and easy to verify.
So an aggregator listing that misstates your waiting period is not a listing problem. It is a sentence that will be repeated in answers given to people who never visit the aggregator.
How do you know what the answer says about your cover?
You read the answer. There is no proxy for it, and the traffic figures will not tell you. Pew Research Center's July 2025 study of Google users in the United States found that 18% of searches in its sample produced an AI summary, and that when one appeared, users clicked a traditional result on 8% of visits, against 15% of visits when no summary was present. Links inside the summary itself were clicked on 1% of visits. Whatever the answer says about your cover is, for most buyers, the whole of what they take away.
So measure it as text. Run the prompts buyers actually type, such as health insurance waiting period for pre existing disease or your claim settlement ratio, across ChatGPT, Gemini, Perplexity and AI Overviews, and capture the answer, your position in it, and the domains cited behind it. A wrong or over-generous restatement of your waiting period stops being a suspicion at that point and becomes a sentence you can act on.
The progression to aim for is a ladder. The answer compares insurers and you are absent. You are named among providers, with no detail attached. Your policy wording is cited as the source for a cover limit. Your plan is in the side-by-side the buyer is deciding from. Your plan is recommended as the fit for this family.
Each rung is earned by publishing what the rung below it was missing: the inclusions in readable text rather than a PDF, every limit and sub-limit that reduces the headline cover, each waiting period with its duration, the claim record with its year and basis beside it, and a quote path that takes age, city and cover and returns a real premium.