Win the AI answer in your market

Why a global discipline has city pages at all

Nothing about how a language model assembles an answer is local. The retrieval mechanism is identical whether the question is asked in Dubai or Dublin. What is local is the question itself, the corpus it reaches into, and the set of names that comes out — and those three are different enough, market to market, to be worth measuring separately.

The evidence for that is sitting in the data behind these pages. We asked ChatGPT, Perplexity and Google the same shortlist question localised to each of our 31 markets, and kept only the answers that survived two checks on the named firm's own website: that it genuinely offers this work, and that it names the city or country rather than ranking globally for a local query. That left 142 engine-named listings, which together with our curated entries cover 169 distinct companies. ChatGPT named 95, Perplexity named 58, Google named 29 — a listing can be named by more than one engine, and most are not. The engines agree with each other far less than you would expect for a question this specific.

What genuinely changes between markets

• The language the question is asked in. English is a working business language in all 31 of these markets, but 11 other languages appear across them, and a prompt corpus in English only will under-measure a market where the buyer researches in German, Arabic, French or Japanese.
• Which sources carry weight. A local business publication, a national directory or a regional subreddit can dominate a city's citation set while being invisible at global scale. The citation map is the part of a measurement run that changes most when you move the query across a border.
• Who is actually competing for the answer. Almost none of the firms named for one city are named for another. The shortlist an engine produces is assembled per question, so a national incumbent and a two-person local studio can appear side by side.

What we are not claiming

We do not have an office in every city on this page, and we are not going to imply otherwise by putting a local address on a generated page. We are based in Edinburgh and work remotely with clients in these markets. Each page exists because the buying question is market-specific, not because the delivery is.

We also do not rank the firms listed on each city page. Recording that an engine named a company for a query is a measurement; deciding that company is good is an opinion, and we have not audited their work. Where a blurb appears it is a quotation from that firm's own site rather than our description of them. If you are choosing between agencies, the buyer's guide sets out the questions that actually separate them, and our own measured position sits on our numbers alongside everyone else's.

Where to start

Open the city closest to your buyers and read the local prompt examples first — if they do not sound like the questions your customers ask, the rest of the page will not fit you either. Then either run the method yourself, which is written up in full and costs nothing, or send us a domain for a free demand snapshot. If your market is not listed, the approach does not change; the methodology is the same everywhere and we can point the corpus at any market.