The State of AI Visibility: UK Cigar Retail — July 2026

Ask an AI assistant where to buy a box of Montecristos in the UK and it will answer with confidence — naming specific shops, in a specific order, citing specific sources. In July 2026 we measured what those answers actually contain. We put the questions UK cigar buyers ask to ChatGPT, Claude, Gemini and Perplexity — 37 questions, sampled repeatedly across all four engines, 1,479 answers in all — and judged every answer against a defined field of 39 UK specialist premium-cigar merchants. The field is a published criterion, not a list of whoever we happened to add.

Five findings stand out.

1. Five names take six of every ten mentions

Across the full measurement, the four engines made roughly 3,700 recognisable references to specialist merchants in the field. Five retailers — C.Gars Ltd, Havana House, Smoke-King, James J. Fox and Simply Cigars — account for 59% of all of them.

The AI shelf has already consolidated. Thirty-nine merchants meet the specialist criterion; when an assistant answers a buyer's question, it draws overwhelmingly from the same five names. For the other thirty-four, the practical reality is that an AI-assisted buyer rarely hears they exist.

Two of those five — Smoke-King and James J. Fox — have not yet had their own full measurement; they earn their place in this report purely by how often the engines name them in others'. Their scorecards will follow.

2. The engines do not agree — and the disagreement is not small

The single most striking pattern in the data: which assistant a buyer uses decides which shop they hear about.

Take C.Gars Ltd, the most-visible brand we measured. Asked the same 37 questions:

• Gemini named it in 73% of answers
• ChatGPT in 70%
• Perplexity in 28%
• Claude in 11%

Same shop, same questions, a nearly seven-fold spread. And the engines don't merely disagree on magnitude — they disagree on order. ChatGPT names C.Gars in 70% of answers but Havana House in just 16%; Claude inverts that, naming Havana House more than twice as often as C.Gars. A brand's AI visibility is not one number in the wild — it is four different realities, one per engine. (Our score averages across all four, and publishes the per-engine split.)

3. Being named is not being recommended

Mentions are common; endorsements are rare. C.Gars is named in 45.6% of answers but recommended — actively steered toward — in only 9.1%. That 9.1% is the highest recommendation rate in the entire field. Most merchants measure between 1% and 6%.

The distinction matters commercially. A mention is shelf presence; a recommendation is the salesperson's nod. Right now, almost nobody in UK cigar retail has earned the nod — which means it is still winnable.

4. AI's sources are Reddit and the enthusiast web — not retailers' own sites

In every one of the five measurements, Reddit was the single most-cited source — cited between 65 and 91 times per measurement, more than any merchant's own website. Behind it sits a consistent cluster of enthusiast and community sites: londoncigarsmoker.com, cigarnights.co.uk, and a handful of retail sites outside the specialist field.

The conversation that decides a merchant's AI visibility is largely happening on forums and blogs the merchant doesn't control. Retailers' own sites do get cited — Havana House's site was cited in 33.8% of its answers, the highest in the field, and its visibility skews heavily toward live retrieval (75% of its mentions came with evidence of a live lookup). That is the fixable half of visibility working as designed: be citable, get cited, get named.

5. Heritage does not register

Gauntleys of Nottingham — trading since 1880, among the oldest family-run tobacconists in Britain — was not named once in 1,479 answers. Not by any engine, to any question.

And here is the sharper edge: across the entire measurement, not a single answer said anything negative about any merchant. The engines never criticise; they omit. In AI search the penalty is not a bad review — it is silence. A century of physical reputation, awards and expertise counts for nothing if the machines that increasingly mediate discovery have never read about it.

The five measured brands

Named / Recommended / Cited = share of that brand's 296 sampled answers (295 for Turmeaus). Share of voice = that brand's slice of all specialist-merchant mentions in its measurement, across the 39-merchant field. Full scorecards: C.Gars Ltd, Havana House, Turmeaus, EGM Cigars, City Cigars.

Even the leader scores 26 out of 100. Nobody in UK cigar retail is winning AI search yet — the field is early, open, and largely unaware it is being ranked.

How we measured

Each brand was measured on 21 July 2026 across ChatGPT (gpt-4o-mini), Claude (Haiku 4.5), Gemini (3.1 Flash Lite) and Perplexity (Sonar) — 37 buyer questions × 2 samples × 4 engines = 296 judged answers per brand. Mentions are counted only when attributable to the actual merchant (a name-match filter tuned against false positives), rates carry Wilson confidence intervals, and the field of 39 is a published criterion: UK specialist premium-cigar merchants a buyer could actually purchase from — not marques, distributors or department stores. Simulated data never feeds a published number.

Disclosure: The AI Visibility Index was built by the founder of City Cigars — which finishes last of the five measured brands, with a score of 6 and the only zero citation rate in the table. The numbers are published exactly as measured. An index you can trust about your competitors has to be one its own founder can lose on.

What happens next

This report will be re-run as the field is re-measured, and scores will move — that is the point of an index. If you run a UK cigar business (or any retail brand) and want to know what AI says when your customers ask: run a free scan, or see where your category stands on the Index.

Questions about the method? It's public: how the AVI Score works.