The brands AI actually recommends to UK consumers: we asked 40 questions and kept score (2026)

Last Updated on August 20, 2026

The question this answers: when a UK consumer asks ChatGPT or Google’s AI which broadband, bank, energy supplier, mobile network or airline to choose, which brands get named, and which get warned against?

Most UK marketing budgets are still spent as if the shop window is Google’s ten blue links. It is not. A growing share of buying decisions now starts with a question typed into an AI assistant, and the assistant does not show ten options. It names two or three brands, ranks them, and sometimes tells the customer who to avoid.

So we kept score.

On 7 August 2026 we asked ChatGPT and Google AI Mode the same 20 questions a UK consumer would ask, four each across five categories: flights, broadband, energy, current accounts and mobile networks. 40 answers in total. We logged every brand named, who got the top spot, and who got named as a warning.

The results should worry some very large marketing departments.

How we ran it

Plain consumer phrasing, no marketing language: “Which broadband provider should I switch to?”, “Which UK bank has the best app?”, “Who is the cheapest energy supplier?” Each question was asked in a fresh chat on a logged-in UK account, on the free tiers most people actually use. Share of voice below means the percentage of a category’s eight answers that named the brand at all. We also spot-checked Gemini on the flights questions and it named the same names, but its free tier would not complete a full panel, so Gemini is not in the numbers.

Flights: Ryanair wins by being exactly what it says it is

Both engines told the same story. Ask which airline is cheap, and Ryanair is first, easyJet second, Wizz Air third, with the same caveats about baggage fees almost word for word. Ask where to book, and it is Skyscanner then Google Flights.

Nobody at Ryanair paid for that. They earned it by being relentlessly, verifiably one thing. AI models reward brands whose reputation is consistent across thousands of sources, because that is what a language model can be confident about.

Broadband: the winner is a company most consumers have never shortlisted

Zen Internet has around two per cent of the UK broadband market. It took the top spot more often than BT, Sky and Virgin Media combined, in both engines, because Which? and Ofcom’s service data say Zen looks after customers, and that is the data AI trusts.

The other side of the coin is blunt. ChatGPT told us, unprompted, that “TalkTalk and Vodafone generated the highest broadband complaint levels in Ofcom’s latest figures.” That sentence is now part of the sales journey, and no media budget removes it.

Energy: Octopus takes almost everything

Octopus was named in every single answer, usually first, for everything from price to green tariffs to app quality. The interesting one is 100Green, a supplier most people have never heard of, named in seven of eight answers because Citizens Advice currently ranks it top for service. One good position in one trusted dataset put it in the AI conversation alongside a company fifty times its size.

British Gas appeared in three answers out of eight. One of those told the customer that legacy suppliers “tend to sit below the top performers for complaints and contact experience”. Then, under that exact answer, ChatGPT showed a paid British Gas advert. The ad and the answer disagreed, and the customer sees both.

Current accounts: the app banks have already won the recommendation

Ask which bank has the best app and both engines give you the same podium: Starling, Monzo, Chase. Ask where a small business should bank, and it is Starling first again. The big four appear mainly when there is cash on the table for switching.

And three established high-street brands were not named in a single answer across 40 attempts. If AI answers are the new shop window, TSB, Metro Bank and Virgin Money are not in the window.

Mobile: the big four own it, and EE owns the big four

EE won coverage, won business, and was ChatGPT’s “least hassle” default. The MVNOs exist in AI answers only as the answer to one question: what is cheapest. That is fine if cheap is your positioning. It is fatal if it is not.

What actually decides these answers

Look at what the engines cited and the pattern is impossible to miss. Which? appeared under answers in every category. MoneySavingExpert, Uswitch, Ofcom, Citizens Advice, OpenSignal and Reddit did the heavy lifting almost everywhere else.

The AI is not deciding who is best. It is reading the referees.

That is the same mechanism we found when we studied which sources AI actually cites: the answer layer is built on a small number of trusted datasets and communities. Brands that perform well in those places get recommended everywhere. Brands that do not are either invisible or, worse, named as the cautionary tale.

Two more things we did not expect to find:

Adverts have arrived inside AI answers. Twelve of our twenty ChatGPT answers carried a paid ad underneath, from Booking.com and Thomas Cook to Lloyds and giffgaff. In two cases the advertiser was a brand the answer above had just ranked low or warned about. The ad changed nothing about the answer. Reputation outranks spend, literally on the same screen.

The engines agree at the top. When we tested AI recommendations for local tradespeople earlier this year, the engines agreed just 8% of the time. For national consumer brands it is the opposite: the same names, in nearly the same order, on both engines. National reputations are settled. Local ones are still a lottery, which is exactly why a small business can still win its own town.

What a small business should take from this

The brands winning AI recommendations share one trait: they are easy to verify as good. Not the biggest, not the loudest. Zen, Octopus, Starling, 100Green. Their service record sits in datasets AI trusts, so the AI repeats it.

In Whito terms, this is Start and Build work, not Scale work. Get the verifiable basics right: reviews that are real and recent, a service record you would be happy to have read out loud, listings and data that agree with each other everywhere. Then make sure the referees for your industry, whether that is Which?, Checkatrade, Google reviews or your trade body, have something good to read. Paid reach comes after, if at all. The British Gas advert under the British Gas warning is what Scale-before-Start looks like in 2026.

The sharp takeaway

AI answers are already handing out buying decisions in every major UK consumer category, and they are doing it on service data, not ad spend. The most recommended brands are the most checkable ones. If your business is not verifiably good somewhere an AI trusts, you are not losing the AI shop window. You were never in it.

Structure before scale. The machines agree.

Common questions

Which brands does AI recommend most in the UK?
In our August 2026 test of 40 answers across ChatGPT and Google AI Mode, the most consistently recommended UK brands were Octopus Energy, Starling, Monzo, EE and Zen Internet, each named in at least five of eight answers in their category. Ryanair and easyJet led flights at 75% share of voice each.

Does AI ever warn people away from brands?
Yes, by name. TalkTalk and Vodafone were flagged for Ofcom complaint levels in broadband answers, and British Gas was placed below top performers for service. These warnings appeared unprompted in standard consumer questions.

Do ads change what ChatGPT recommends?
Not in our data. Twelve of twenty ChatGPT answers carried adverts, and in two cases the advertiser had just been ranked low or cautioned in the answer above the ad. The recommendation text was unaffected.

Why do small brands like Zen and 100Green beat household names?
Because AI engines lean on trusted assessors such as Which?, Ofcom, Citizens Advice and MoneySavingExpert. A strong position in one trusted dataset can put a small brand into AI answers ahead of far larger competitors.

Want to know how visible and verifiable your own business looks to AI search? The free Whito AI Visibility Scorecard checks it in about two minutes and tells you the one thing to fix first. Related Whito research: the AI recommendation lottery, which sources AI actually cites, and why AI engines disagree about the best tradesperson.

Whito

Whito exists to stop businesses scaling the wrong way. We focus on structure, leverage, and measurable growth, not noise, not vanity metrics.

See Full Bio