Last Updated on September 8, 2026
Reviewed by Jacob Whitmore, Whito · Fact-checked for accuracy
This research is free to read and the method is published in full. The complete dataset behind studies like this one, every firm named and the raw answer logs, is available as a report. Whito Intelligence
Finding
We asked ChatGPT and Google AI Mode to name the best mortgage adviser in 12 UK cities on 8 September 2026. They named 88 firms between them. Seventeen could not be matched to an active FCA permission in, or reasonably serving, the city they were recommended for, and 50 of the 88 were appointed representatives trading under a name that is not the authorised firm.
Source: Whito, whito.co.uk/research/ai-mortgage-advisers-fca-register-check, September 2026. Free to quote with a link to this page.
Last updated: 8 September 2026. Add Whito as a trusted source on Google
In August we asked three AI engines to name the best mortgage adviser in six cities and counted who they agreed on. This time we asked in twelve cities and then did the thing almost no reader does. We looked every name up on the FCA register.
A note before we start. Whito is not authorised or regulated by the Financial Conduct Authority and has no FCA firm reference number. This article is media research about what AI engines said on a stated date. It is not financial advice, not mortgage advice, and not a financial promotion for any firm. Firms and people are named only because an AI engine named them. Inclusion is not an endorsement. Anyone choosing a mortgage adviser should check the FCA Financial Services Register at register.fca.org.uk and take advice from an authorised person. Corrections: hello@whito.co.uk.
What we did
We reused the consumer question from our August 2026 mortgage adviser study, “Who is the best mortgage advisor in [city]?”, and asked it in 12 UK cities. The six from 17 August stayed in: Manchester, London, Birmingham, Leeds, Bristol and Glasgow. We added Preston, Norwich, Plymouth, Derby, Swansea and Aberdeen, which are smaller markets.
Two engines answered: ChatGPT on the free tier with web search on, and Google AI Mode. That gives 24 answers, all collected on 8 September 2026 and screenshotted. Google Gemini was in the original design and is not in these numbers, because its sends were silently discarded for most of the day and we finished with one city out of twelve. An incomplete engine would distort every comparison, so we left it out and said so here rather than in a footnote.
Then the new part. In the same week we looked up every named firm on the FCA Financial Services Register. For each one we recorded the legal entity behind the name, its firm reference number, its registered address, whether it is directly authorised or an appointed representative of a principal firm, and whether it holds mortgage permissions. Then we coded one outcome per firm.
The register check is the point of the exercise. An AI answer that names a firm reads like a recommendation, and the register is the only public record of who holds permission to give the advice being implied. We wanted to know how often the two line up, so we coded every name.
Shorter version of the regulatory note, because it matters here too: we are not the FCA, we hold no permissions, and every outcome below is our reading of a public record in a stated week.
The headline numbers
- 24 answers: 12 cities, two engines, one day.
- 88 unique firms named, filling 145 recommendation slots.
- 71 firms (80.7%) matched cleanly to a live register entry in or reasonably serving the city they were named for.
- 17 firms (19.3%) could not be treated as a safe local match. That covers wrong city (9), not found under the name given (4), lapsed or no longer registered (2), a name collision we could not resolve (1), and one firm whose register entry shows no mortgage permission.
- Those 17 firms filled 27 of the 145 slots, or 18.6%.
- 50 of the 88 firms (56.8%) are appointed representatives. The name in the answer belongs to one company; the FCA permission belongs to another.
- 26 firms (29.5%) hold their own authorisation. The remaining 12 are the unresolved and lapsed cases.
- By slot, appointed representatives took 90 of 145 placements (62.1%), directly authorised firms 33 (22.8%), national brands 9 (6.2%).
We are calling the middle finding the permission-name gap: the distance between the name an AI answer gives you and the name that holds the FCA permission. In this sample it is the normal case rather than the exception.
Outcomes on the register
| Outcome | Firms | Share |
|---|---|---|
| Appointed representative, principal confirmed, in that city | 43 | 48.9% |
| Directly authorised, mortgage permission, in that city | 25 | 28.4% |
| National or multi-city brand, authorised | 3 | 3.4% |
| Registered in a different city from the one recommended | 9 | 10.2% |
| Not found under the name the engine gave | 4 | 4.5% |
| Found but lapsed or no longer registered | 2 | 2.3% |
| Name collision, could not disambiguate | 1 | 1.1% |
| Named entity shows no mortgage permission | 1 | 1.1% |
| Not a safe local match, combined | 17 | 19.3% |
By recommendation slot, which weights firms by how often the engines surfaced them:
| Outcome by slot | Slots | Share |
|---|---|---|
| Appointed representative | 90 | 62.1% |
| Directly authorised local firm | 33 | 22.8% |
| National brand | 9 | 6.2% |
| Unresolved, lapsed or collision | 13 | 9.0% |
City by city
| City | Firms named | Matched locally | Not a safe local match | Named by both engines |
|---|---|---|---|---|
| Manchester | 7 | 7 | 0 | 4 |
| London | 11 | 11 | 0 | 1 |
| Birmingham | 10 | 6 | 4 | 4 |
| Leeds | 7 | 4 | 3 | 5 |
| Bristol | 8 | 8 | 0 | 4 |
| Glasgow | 10 | 7 | 3 | 4 |
| Preston | 9 | 7 | 2 | 4 |
| Norwich | 6 | 6 | 0 | 4 |
| Plymouth | 9 | 6 | 3 | 4 |
| Derby | 9 | 5 | 4 | 4 |
| Swansea | 10 | 8 | 2 | 4 |
| Aberdeen | 7 | 5 | 2 | 5 |
Four cities came back clean. Manchester, London, Bristol and Norwich produced no unmatched names at all. The added smaller cities did not perform worse as a group, which we had half expected. Derby and Birmingham produced the most unmatched names. In both, the unmatched entries came from directory suggestions appended to the end of a ChatGPT answer.
What the answers looked like
ChatGPT gives a shortlist in prose, usually five or six firms, and this time it quoted far fewer star ratings than in August. Where it did quote them the ceiling was still there. In Plymouth all six firms were a flat 5.0. In Derby all six were 5.0. In Norwich all four were 5.0.
Google AI Mode builds a panel of business cards with Google ratings, review counts, street addresses and phone numbers, then a comparison table. It is the more confident-looking of the two, and it is the one that names people. AI Mode named individual advisers in 11 of its 12 answers, more than 70 name-mentions in total. ChatGPT named one individual adviser across all 12.
Advertising has spread since August. Eleven of the 12 ChatGPT answers carried a sponsored placement beneath the answer. Six were Yorkshire Building Society promoting its own mortgage rates. Five were GetAgent, an estate agent comparison site, which is one of the two advertisers we recorded in August; the other was a double glazing firm. A lender advertising against a question about advisers is new in this series. Leeds carried no advertising.
On the regulatory side, both engines gestured at authorisation without ever telling the reader to check it. ChatGPT closed seven answers by relaying Unbiased’s “FCA-verified” framing for named firms. AI Mode cited an Unbiased page titled “Find the best FCA regulated mortgage brokers” in Plymouth, and appended “AI responses may include mistakes. For financial advice, consult a professional” to every answer. Neither engine, in any of the 24 answers, told the reader to look a firm up on the FCA register.
Three worked examples
A clean match. Independent Mortgage Store, named by both engines for Glasgow. The register shows Independent Mortgage Store Limited, firm reference number 301002, authorised since 31 October 2004, holding mortgages and home finance permissions, at a Glasgow G11 postcode on Byres Road. The name in the answer, the name on the register and the address all agree.
One firm, five cities. ChatGPT recommended Fairway Mortgage Advice as an “FCA-verified independent adviser” in Birmingham, Glasgow, Preston, Derby and Swansea, in each case relaying a line from Unbiased. Fairway Mortgage Advice Limited, firm reference number 1013529, is registered in Washington, Tyne and Wear, and has been an appointed representative of Stonebridge Mortgage Solutions since June 2024. It is a real, live entry. It is also a single firm in the north east offered as the local option in five cities, four of them hundreds of miles away.
Named, then nothing. ChatGPT named SJ Financial Solutions as a Birmingham option. Searching the register for that exact name returns SJR Financial Solutions Ltd in Northampton, no longer registered as an appointed representative, and SJ Mortgage Solutions Ltd. Widening to “SJ Financial” returns 145 firms, none of them in Birmingham. The firm’s own website gives a Birmingham address and carries no firm reference number and no authorisation statement on the page we fetched. We could not match the name to a register entry for that city in the verification week, which is what our “not found” code means and all it means.
What the register showed that the answers did not
Three patterns are worth setting out, because they explain most of the 17.
The first is the permission-name gap. Fifty of the 88 firms trade under a name that is not the authorised entity. Trinity Financial in London is Trinity FG Ltd, an appointed representative of Advance Mortgage Funding. aFinancial in Birmingham is a trading name of one person, Maryam Tala Maleki, an appointed representative of Openwork. First Mortgage, named in five of our twelve cities, is an appointed representative of Mortgage Advice Bureau. None of that is irregular, and appointed representatives are a normal and lawful part of the UK mortgage market. It does mean the name a consumer is given is often not the name they would need to search.
The second is distance. Nine firms are registered in a city other than the one they were offered for. Cloud Mortgages, recommended for Glasgow by both engines and described by AI Mode as having local advisers, resolves to Isuna Financial Solutions Limited in Nottingham, an introducer appointed representative of St James’s Place. JLM Mortgage Services Aberdeen resolves to JLM Mortgage Brokers Ltd, registered in Hitchin in Hertfordshire. The city is in the trading name; it is not always in the register entry.
The third is the dead namesake. Several brands have two entries, one live and one lapsed, and the lapsed one often sorts first. “Swan Mortgages” returns a firm no longer registered as an appointed representative before it returns the live Swansea company. So does “Mortgage Advice Brokerage”, where the dead entry and the live one are both in Glasgow and separable only by street address. A consumer doing the sensible thing and searching the register can land on the wrong record.
What this does not mean
It does not mean a matched firm is a good adviser. A register entry records permissions, not quality.
It does not mean an unmatched name is a scam. Most of the unmatched names are trading styles, network arrangements the register displays under a different company, or firms the engine placed in the wrong city.
It does not mean appointed representatives are second class. They are regulated business, and the principal firm carries responsibility for it.
It does not mean AI should be banned from answering the question. It means the answers should be checked, and mostly are not.
It does not mean Whito recommends anyone. We do not.
It does not mean the FCA register is wrong. Where an answer and the register disagreed, we treated the register as the record.
If you are choosing an adviser
Do the check we did. It is free and takes a few minutes. Put the firm name into the FCA Financial Services Register, confirm the entry is live, confirm it covers mortgages, and confirm the address matches the office you are dealing with. If the search returns two similar names, read both before you pick one. If the firm is an appointed representative, the register names its principal, and you can ask the firm to confirm in writing what it is permitted to advise on. Then take advice from an authorised person, not from an AI answer, and not from this page.
If you are a mortgage adviser
This is visibility research, not compliance advice. Two points in the data bear on how you show up.
The engines match names loosely. Where a firm’s trading name, register entry and Google Business Profile agreed, the answer got it right. Where they diverge, the answer drifted. Alignment across those three seems to matter more than any single listing.
Directories are carrying more weight than they were in August. Every one of the five multi-city appearances by a single north east firm came through a directory line relayed by ChatGPT. If you are listed, your listing is being read as local evidence in cities you may not serve. If you want to see what the engines currently say about your firm, the free Whito Visibility Check will tell you.
Where this sits next to the August study
The 17 August study measured agreement. Seventy-one firms across six cities, 38% named by two or more engines, five firms with full three-engine consensus, and almost every quoted rating a flat 5.0. It took the answers at face value and counted them.
This study asks the question underneath. When an engine names a firm, does the name stand up against the public record?
One thing did carry over. All five firms that had full three-engine consensus in August were named again in the same cities three weeks later: Manchester Money in Manchester, Trinity Financial in London, First Mortgage in Birmingham, and Independent Mortgage Store and West End Mortgages in Glasgow. The firms the engines agreed on in August are the ones that held their placements.
Method, limitations and disclosures
The AI answers are a one-day snapshot collected on 8 September 2026. Register verification ran in the same week. A firm’s register status can change after that week, and answers vary by account, session and day, so treat the direction of the numbers as the finding rather than any single placement.
Trading names were matched to register entries by hand. Local trading names of one brand were counted as one firm, which is why the two city-branded “moneyman” names in our sample count once. Where a register search was ambiguous we read the firm’s own website footer for its stated entity and firm reference number, then confirmed that entry on the register. We did not call any firm, did not assess anyone’s advice, and did not treat a directory listing as evidence of authorisation. Star ratings and review counts are what the engines displayed on the day and are not verified. Screenshots are cropped to the answer, and advertising is reported where it appeared. Name collisions are coded as collisions rather than as “not found”. Four firms could not be resolved to any register entry under the name given, and we have counted them as unresolved rather than assuming either way.
Google Gemini was excluded, as set out above, because only one of twelve sends completed. One firm, Reliant Mortgages and Insurances, could not be resolved because the register search did not return during the verification session; it is counted in the unresolved group.
Whito is not authorised or regulated by the FCA and holds no firm reference number. Nothing here is financial advice, mortgage advice or a financial promotion. No firm paid to appear, no firm was contacted before publication, and firms are named solely because an engine named them. Corrections and right of reply: hello@whito.co.uk. The full answer log and register-check sheet are available through Whito Intelligence.
Cite this research
Whito Research (2026). We Asked AI for a Mortgage Adviser. Then We Checked the FCA Register. Whito. https://whito.co.uk/research/ai-mortgage-advisers-fca-register-check/
Key finding: of 88 mortgage firms named by ChatGPT and Google AI Mode across 12 UK cities, 17 could not be matched to an active FCA permission in that city and 50 were appointed representatives trading under a name that is not the authorised firm.
Common questions
Did you check these firms with the FCA?
No. We read the public Financial Services Register, the same record any consumer can read. We did not contact the FCA and the FCA has no involvement in this research.
Does “not found” mean a firm is unauthorised?
No. It means we could not match the name the engine gave, for the city the engine gave it for, to a live register entry in the verification week. Trading styles and network arrangements can hide an authorised firm behind a name the register does not show, which is why “not found” and “name collision” are separate codes.
Is being an appointed representative a problem?
No. It is a standard structure in the UK mortgage market, and the principal firm is responsible for the regulated business its representatives carry out. The finding here is about searchability, not legitimacy: if you are given one name and the permission sits with another, checking is harder than it looks.
Why these twelve cities?
The six from August, so the results are comparable, plus six smaller cities to test whether thinner local markets produce weaker answers. On this evidence they did not.
Why only one day of answers?
Because that is how a consumer meets these engines. One question, one day, one answer. The August study three weeks earlier is the comparison point.
Can I quote these numbers?
Yes, free, with a link to this page and the collection date. The finding box at the top is written to be quoted whole.
The structural checks behind this research run on any website in about twenty seconds. Enter yours and read the full result on the page. Free, and we do not ask for your email.
Run the free Visibility CheckIf your business is named here and you think we have a fact wrong, email hello@whito.co.uk and we will check it and correct it. If you want to respond, send us a statement and we will publish it alongside. Full detail: Corrections, right of reply and removals.
