How Many UK Businesses AI Calls Trustworthy Are Not a Real, Active Firm in That Town

Last Updated on September 8, 2026

On 22 August 2026 Google AI Mode named 734 businesses as trustworthy local firms across 20 UK towns. We checked every preserved name against Companies House and the public record. 38 of 728, or 5.2%, could not be verified as an active company trading in that town.

That is the finding. We asked one question, “Recommend a trustworthy [sector] in [town]”, once for each of 8 sectors in each of 20 towns. 160 answers came back and named 734 businesses. The archived collection logs preserve 728 of those named recommendations, and each one was checked on 28 August 2026. 21 traced to a company that is dissolved, in liquidation, or filing dormant accounts. 15 could not be tied to any company or public footprint at all. 2 were too generic to match to one firm.

Cite this research: Whito AI Visibility Index, Q3 2026, unverifiable recommendations analysis. Collected 22 August 2026, Google AI Mode, 20 UK towns, 8 sectors, 160 answers, 734 businesses named, 728 preserved and checked, 38 unverifiable (5.2%). Verified against Companies House 28 August 2026. whito.co.uk/research/ai-recommended-businesses-unverifiable-uk/. Free to cite with a link.

What “unverifiable” means

We are not saying these businesses are fake. We are saying a public check could not confirm them as an active company trading in the town the question asked about. A named recommendation counted as unverifiable if any one of these was true:

1. No matching live company at Companies House, and no public footprint tying the name to a trading business in the town.
2. The company is dissolved, in liquidation, or filing dormant accounts.
3. The registered or trading address is not in, and does not serve, the town asked about.
4. The website is gone or the domain is for sale.
5. The name is too generic to match to one firm.

A firm can be legally active and still fail the town test. The full decision rules are in the codebook below. Every one of the 728 rows records which rule fired, and 31 rows are flagged as close calls.

The results by sector

Dentists fail most, at more than one recommendation in ten. That sector’s answers were nearly all prose rather than Google local cards, and practices often trade under a name that never appears on any company register. The most common reason across all sectors was a dead company: 21 of the 38 traced to a registered company that is dissolved, in liquidation, or dormant, including established multi-branch firms whose exact-name company has been struck off. The engine recommends the name. The register says the company behind it has gone.

The other side of the ledger matters as much. Only 305 of the 728 recommendations could be tied by name to a live registered company. The remaining 385 verified ones were confirmed through their public trading footprint or a national brand instead, because sole traders, franchises and practices trading under a different legal name leave no name-shaped trace at Companies House. AI engines do not check any of this. Neither do most customers.

What a small business should do with this

If AI engines are recommending firms that cannot be verified, a firm that can be verified has an edge. Make the check easy to pass. Your Companies House record, your website, your Google Business Profile and your trade register entry should show the same name, the same address and the same phone number. That consistency is what both a customer and an AI engine can confirm. Run the free check at whito.co.uk/visibility-check/ to see how your own site holds up.

Method in full

Source data. Whito AI Visibility Index, Q3 2026. Collected 22 August 2026. One engine, Google AI Mode. Prompt: “Recommend a trustworthy [sector] in [town]”. One run per prompt. No retries, no regeneration, no cherry-picking. Full collection method: whito.co.uk/research/uk-ai-visibility-index/. The Index recorded 734 named recommendations; the archived logs preserve 728 as named entries. The six not preserved, which include the Wolverhampton vets answer, are counted in the Index total but could not be checked here.

Towns (20). Reading, Sunderland, Swansea, Leeds, Nottingham, Coventry, Luton, Bristol, Doncaster, Ipswich, Preston, Stoke-on-Trent, Milton Keynes, Norwich, Plymouth, Glasgow, Belfast, Mansfield, Blackpool, Wolverhampton.

Sectors (8). Electricians, plumbers, roofers, cleaners, vets, dentists, accountants, estate agents.

Verification. All 728 preserved recommendations were checked against the public Companies House register (find-and-update.company-information.service.gov.uk) on 28 August 2026. Each name was searched, candidate companies were compared on name, status and address, and the chosen company’s status and registered office were read from its register page. A name-and-town match to an active company verified the row. Where the registered office sat outside the town, a postcode area covering or adjacent to the town, a branch named for the town, or a national brand or franchise counted as serving it. Sole traders have no company record, so a name with no company match was coded verifiable only where the collection log holds a public trading footprint for it in that town: a Google listing with reviews, a directory profile such as Checkatrade or TrustATrader, or the firm’s own site. Names failing both tests were coded unverifiable. Each row records the first rule that fired, and judgement calls are flagged: where the evidence was genuinely close, the row was coded verifiable, so the headline number is a floor, not a ceiling.

Counting. A business named in more than one town or sector was checked once per naming, because the claim being tested is “trustworthy in that town”. The unit throughout is the named recommendation, not the unique firm.

Limits

This is a single-run snapshot of one engine on one day. AI answers vary between runs, so the figures describe what was served on 22 August 2026, not a stable property of the engine. Companies House matching is imperfect: trading names differ from registered names, sole traders are absent from the register, and a dissolved namesake is not always the recommended firm, which is why 31 rows carry a close-call flag and why close calls were resolved towards verifiable. Websites were not fetched row by row; the website test fired only where the collection log already showed a dead or absent site, so rule 4 contributes nothing to this edition’s count. Dormancy was read from the register page and can lag reality. Six of the Index’s 734 named recommendations were not preserved in the archive and are excluded from the denominator.

Corrections

If your business is named in the underlying data and you think we have a fact wrong, email hello@whito.co.uk and we will check it and correct it. Corrections policy: whito.co.uk/corrections/.

Ethan Whitmore Co-founder and Analyst

Ethan Whitmore is a co-founder of Whito. He has spent over nine years in SEO and ecommerce inside global SaaS platforms, enterprise brands and payments companies. His work covers technical SEO, content strategy, conversion and video production.

See Full Bio