Which? Is Becoming a Ranking Signal. What That Means If You Sell Insurance.

Which? has partnered with MoneySuperMarket to launch a co-branded insurance comparison platform, covering home, car, van and travel insurance. It was reported by the trade press on 14 July 2026. No launch date has been announced, and there is no press release from either company yet, so treat the detail as reported rather than shipped.

Here is the part worth your attention. According to Insurance Times, users will see Which? scores and endorsements displayed directly on the insurance results cards, with the ability to sort providers by their awarded score.

Read that last clause again. Sort providers by their score.

Why a sort button is not a small detail

A badge is decoration. You can win a Which? endorsement, put it in your footer, mention it in a renewal email, and it makes very little difference to who ends up on a shortlist.

A sort function is not decoration. It is a ranking mechanism. The moment a score becomes the axis a results page can be ordered by, it stops describing the market and starts shaping it. Whoever sits at the top of that sort gets the clicks. Everyone below the fold is, for practical purposes, not there.

That is the whole game, and it is a game most insurance businesses have never had to think about, because for twenty years the ranking inputs on a comparison site were price and, to a lesser extent, brand recognition. Both were things you controlled directly. A third-party editorial score is not.

Phil Amy, commercial director at Which?, put the rationale plainly: the aim is to embed the organisation's research "right at the point of purchase". That is an honest description of what is happening. A trust signal is being moved from the sidelines into the buying decision itself.

You have seen this before, without seeing it

Now hold that thought against the advice that has been hardening in AI search all year.

It has become a fairly blunt instruction: if you want to be recommended by ChatGPT, Perplexity or Google's AI Overviews, get yourself mentioned on the sources those engines read first. The listicles. The review sites. The comparison and accreditation bodies. Because when a buyer asks an AI assistant "who is the best insurer for a van", the assistant does not have an opinion. It goes and reads what the trusted sources say, and it repeats them.

Which? is one of those sources. It has been one for decades. We covered a related version of the same shift when Hiscox started paying to appear inside ChatGPT answers, buying its way onto a surface it could not earn its way onto.

Which means the same mechanism is now running in two places at once. A Which? score is about to become a visible ranking input on a comparison site. And it has been an invisible ranking input inside the AI engines for as long as those engines have been retrieving pages.

The comparison site is just the version you can point at.

The uncomfortable version

Put the two together and the picture for a UK insurance business, a broker or an accident management company looks like this.

You are being scored by systems you do not control. The scoring is happening in places you are probably not looking. And the score is becoming the thing people sort by.

That is not a hype line, it is a description of a mechanism. When a buyer asks ChatGPT for a recommendation, the shortlist it produces is assembled from third-party sources. If your business does not appear on those sources, or appears with thin or outdated information, you are not on the shortlist. Not ranked lower. Not there at all. The buyer never learns you exist, and you never learn you were considered, because there is no impression count, no lost-bid report, and no referral log for a conversation that happened without you.

The comparison site at least sends you a report. The AI layer does not.

The honest counter-argument

Now let me argue against myself for a moment, because this is where a lot of AI visibility commentary gets lazy.

None of this matters if the Which? score does not actually move purchase behaviour. It is entirely possible that consumers land on the results page, ignore the endorsement, sort by price the way they always have, and buy the cheapest policy that clears their excess threshold. Insurance is a famously price-led purchase. A trust badge has lost that argument many times before.

That is a fair objection, and anyone selling you a panic about this should have to answer it.

Here is why I still think the mechanic matters. The objection assumes the score is presented as a badge. It is not, or at least it is not only. It is being wired in as a sort option, sitting alongside price as an axis the page can be ordered by. That is a structural position, not a decorative one. And even if only a minority of buyers ever touch that sort control, the score is now a machine-readable, structured, publicly available signal attached to your brand, sitting on a high-authority domain, in exactly the format the answer engines like to consume.

Which is the second-order effect, and the one nobody is talking about. Even if not one consumer ever clicks "sort by Which? score", the existence of that score in a structured form on a major comparison site makes it more likely to be picked up, cited and repeated by an AI assistant answering a question about you.

The badge is optional. The data point is not.

Three things worth doing this month

None of these require a budget.

Check what the engines already say about you. Open ChatGPT, Perplexity, Google's AI Overviews and Claude. Ask the questions your customers actually ask. "Who is the best van insurance provider in the UK." "Who handles non-fault accident claims in Cheshire." "Best credit hire company." See whether you appear at all, and if you do, see whether the description is accurate and current. Most businesses that run this exercise for the first time are unpleasantly surprised, and the surprise is usually not that they rank badly, it is that they are absent.

Find out which sources the engines are citing. When the AI names a competitor, ask it where the information came from. The engines will generally tell you. Those cited sources are your target list. If a review site, a directory, a trade body or a comparison platform keeps appearing in the answers about your sector, that is where the authority sits, and that is where the work needs to go.

Audit your presence on those sources, not just your own website. This is the part most businesses get backwards. They pour effort into their own site, which the engines may read once, and ignore the third-party sources the engines lean on constantly. Your listing on an accreditation body, your entry on a comparison platform, your profile on a trade directory, the accuracy of your details on a review site. That is the plumbing. It is unglamorous and it is where the visibility actually comes from.

The bigger point

There is a pattern underneath all of this that goes wider than insurance.

The value in AI is drifting away from the model and towards the layer that surrounds it. The models themselves are commoditising fast, undercut on price, matched on capability, swapped in and out of products without users noticing. What is not commoditising is trust, authority, and the question of whose judgement gets repeated when a machine answers a question. Google's own recent guidance that AEO is really just SEO done well points at the same underlying truth: the fundamentals of being a trustworthy, well-cited source are what carry through into the AI layer.

Which? has spent decades building a trust signal. It has just turned that signal into something a computer can sort by. That is a very good business decision, and it should tell you exactly where the scarce asset is.

The question for everyone else is simpler. When the machines go looking for a trustworthy answer about your sector, do they find you, and do they find you described correctly?

Most UK service businesses have never checked. That is the whole opportunity. If you want a broader view of where your visibility sits, the Optimus AI Visibility audit scores what is there, flags what is missing, and writes the practical fix list.