Your brand can be the go-to answer in AI search right up until someone asks how much it costs. Here’s how we used prompt tracking to find that gap for an insurance client, and how you can look for the same thing on your own site.
Tom LightMost brands we speak to want to know one thing about AI search: are we showing up? It’s a fair question, but it hides a more useful one. Where are you showing up, and where are you missing?
When we looked at the AI visibility data for one of our insurance clients, the headline looked healthy. They were ahead of their competitors overall. But when we split the prompts by what the user was actually asking, a very different picture appeared. On advice questions, they were in almost every answer. On cost questions, they barely appeared at all, and competitors were picking up the conversation instead.
That matters because cost questions aren’t idle curiosity. They’re asked by people getting close to a decision.
None of this needed specialist tooling or a data science team. It needed two things: prompt tracking in place, so you can see which AI answers mention you, and a theory you want to test, so you know what to look for in the data.
In this article, we’ll walk through what we found, why we think it happens, and how we’re testing a fix. By the end, you should know how to find the questions your audience asks where your brand is missing, and what to do about them.
People rarely start their research with price, but almost all of them get there.
In insurance, the evidence is consistent. GlobalData’s UK consumer research found premium cost and brand reputation are the main drivers of provider choice, with around a third of people naming cost as the single most important factor. A Una Alliance survey of 1,200 UK consumers found more than half ranked price as the most important factor when buying insurance.
And it isn’t just insurance. TrustRadius research found 81% of B2B tech buyers want to find pricing on their own, and that buyers who can’t find it move on. In property, REA Group found 72% of house hunters skip listings that don’t show a price.
Search behaviour tells the same story.
The biggest single phrases are simple ones like “how much is life insurance” and “how much is public liability insurance”, each searched thousands of times a month. Around two thirds of these cost searches already show an AI Overview at the top of Google, so the answer is increasingly written by AI before anyone clicks.
If a buyer asks an AI tool about cost and your brand isn’t in the answer, you’ve left the conversation at the moment it matters most.
Prompt tracking works a bit like rank tracking for AI search. You choose a set of questions your customers are likely to ask, run them through AI tools such as ChatGPT, Google’s AI Overviews and Gemini on a regular schedule, and record which brands are mentioned and which sites are cited as sources.
On its own, an overall visibility score doesn’t tell you much. The useful step is grouping your prompts by what the person is trying to do. For our client, we split them broadly into two groups:
Advice prompts: early research questions, such as which cover suits a particular need or what to look for in a policy
Cost prompts: questions about price, such as how much cover typically costs, whether a particular risk makes it more expensive, or what affects the premium
You can use whatever groups make sense for your market. The aim is to see your visibility at each stage of the journey, not just as one number.
Once the prompts were grouped, the gap was hard to miss.
On advice prompts, our client was mentioned in almost every answer. Their guidance content was doing its job, and AI tools were clearly picking it up.
On cost prompts, it was a different story. Visibility dropped to a fraction of the advice level, and it stayed low across every product type we tracked. Same brand, same website, same audience. The only thing that changed was the question.
This is the kind of insight an overall visibility score hides. Averaged together, the brand looked healthy. Split by intent, there was a whole stage of the journey where they were barely present.
A gap only matters if someone else is filling it. So the next step was to look at which brands were showing up on cost prompts instead.
Across all tracked prompts, our client was ahead of the competition. On cost prompts alone, one competitor was seen far more often, and our client’s visibility fell well below its usual level.
The gap wasn’t even, either. Our client did best on direct “how much does it cost” questions, appeared less often on questions linking cost to a specific risk, and was rarely mentioned on comparative questions such as “is X expensive?”. On general questions about what affects the price, they didn’t appear at all.
We also read through the AI responses themselves. Mentions were inconsistent for every brand, and the same question could produce a different set of brands from one day to the next. But there were clear cases where competitors were named and our client was missing entirely.
When we looked at which sources AI tools were citing on cost prompts, a pattern stood out straight away. The sites being cited were the ones that actually published a number.
One independent comparison site was cited again and again. It had a page for each product type, and every page had a simple “key facts” panel near the top showing an average premium and a typical price range. When we asked an AI tool how much cover cost for one of those product types, it pulled that range straight into its answer, credited the site as the source, and kept doing so over time.
Here’s a simplified version of what that looks like:
AI answer: Basic cover typically costs around £X to £Y a month, with comprehensive cover nearer £Y to £Z depending on your circumstances and level of cover. Source: comparison site
Our client’s pages took a different approach. Like many insurers, the cost questions on their product pages explained what affects the price and then pointed to “get a quote”. That’s sensible from a pricing point of view, but it gives an AI tool nothing to work with. If there’s no number on the page, there’s nothing to cite.
Visibility also varied a lot between AI platforms, which is why we track several rather than one. A brand can be strong in one AI tool and almost absent in another.
All of this led to one simple theory:
If we publish an indicative price range on our product pages, AI tools will mention and cite us more often when people ask about cost.
To test it properly, you don’t change every page at once. If visibility went up, there’d be no way of knowing whether it was the change or something else, such as a model update or a seasonal shift in demand. So the test is set up with a small group of test pages that get the change, and a matched group of control pages that stay exactly as they are. Visibility and share of voice on cost prompts are then tracked every week for a fixed period, and the test pages are compared against the controls.
You don’t need anything more sophisticated than this. A handful of pages, a clear change, a like-for-like comparison and a fixed time period will tell you far more than changing everything and hoping for the best.
In a regulated industry like insurance, “just publish your prices” isn’t a realistic request. Premiums depend on the individual, and nobody wants a page that reads like a promise. So the change has to give AI tools something useful to cite without creating a compliance headache.
The answer is a structured pricing block near the top of each test page, built from information the business already has.
Four things make this an easier sign-off:
A range, not a quote. An indicative range with a clear disclaimer, the same approach already used across the market, rather than a single fixed figure.
Anchored to historical data, not pricing. It leads with what has actually happened, which looks backwards rather than promising what someone will pay.
Reuses approved copy. Wherever possible, it brings together content that’s already live and signed off, rather than writing new pricing language.
No change to how prices are calculated. It’s a content edit to the page template, not a change to the quote process.
If you work in a regulated sector, this framing is worth borrowing. Compliance teams are far more comfortable with “we’re reordering approved content and adding a disclaimed range” than “we want to publish prices”.
This isn’t specific to insurance, or to price. The same approach works for any gap between the questions your audience asks and the answers your site gives.
Set up prompt tracking. Build a set of prompts that reflect the real questions your customers ask, at every stage of the journey. Your search data, sales team and customer service inbox are good places to start.
Group prompts by intent. Tag each one by what the person is trying to do: research, compare, check cost, choose a provider. Then look at your visibility for each group, not just the total.
Find the weakest group. Look for a stage where your visibility drops sharply, especially one close to a buying decision.
Check who’s winning instead. See which competitors are mentioned and which sites are cited as sources. Then read the AI answers themselves to see what information they’re pulling through.
Compare their pages with yours. What do the cited pages say that yours don’t? Often it’s something concrete: a number, a comparison, a clear yes or no.
Turn it into a theory and test it. Write down what you think will close the gap, change a small set of pages, keep a set of controls, and track both for a fixed period.
The tooling helps, but the thinking matters more. A clear theory turns a dashboard of visibility scores into something you can act on.
We’re putting this theory to the test, and we’ll share what we learn, good or bad, in a follow-up piece.
Whatever the outcome, the analysis has already done its job. It turned a vague question (“are we visible in AI search?”) into a specific, testable one, and showed exactly where a valuable part of the audience was being missed.
If you’d like help setting up prompt tracking or finding the gaps in your own AI visibility, talk to us. To see where your brand sits on the wider journey, read AI search maturity: where brands are now, and how to move up.
We’ll set up prompt tracking, show you where competitors are winning the conversation, and help you test what closes the gap.
