We asked five AI engines who to sell with in Teignmouth

First, the disclosures, because they are the point. Dart & Partners is our client: we built their website, and this measurement is the baseline for the AI Visibility service they were the pilot for. Nothing here is paid placement, because there is no such thing: none of these engines sells a seat in its answers. Every quote below is verbatim from a captured answer, dated 30 July 2026, and we have kept the misses in alongside the wins, because a measurement you can only pass is not a measurement.

What we asked, and who we asked

The panel was eight questions a real seller, landlord or buyer in Dart's patch would ask: who to sell with in Teignmouth, Dawlish and Shaldon, the best letting agent for landlords in two towns, who should value a house near Shaldon, who sells fine and coastal homes in South Devon, and the best agent in Bishopsteignton. We put them to Google's AI Mode, Perplexity, Microsoft Copilot, ChatGPT and Claude, and recorded who each one named, in what order, and in what words.

The headline

Across the full panel, Dart & Partners was named in 29 of the 31 answers captured, and was first or the outright recommendation in 14 of them. Not because anyone paid, and not because the questions were soft: the same panel produced two answers where they were not named at all, and several where a competitor beat them.

What the engines actually said

Google's AI Mode, asked who to sell with in Teignmouth: "Dart & Partners represents the gold standard for independent, long-standing local expertise." Perplexity opened its answer with a decision: "I'd start with Dart & Partners, Nexmove, and Bradleys", citing their volume of reviews. ChatGPT called them "probably my first call" for selling, "the safest all-round choice" for landlords, and for a valuation near Shaldon it went further than words: it offered the questioner a button to book a Dart valuation on the spot. Copilot reached back into their history: "Established in 1971, award-winning independent agent, known for excellent customer feedback." Claude, measured with live web search, put them first in Shaldon and for Teignmouth landlords, and honestly second in Dawlish behind a lettings-only specialist.

The misses matter more than the wins

Two findings were worth more than any first place. The engines showed the wrong Google profile for one question, because one branch's listing was labelled with the postal town rather than the village it is actually in: a 40-review card was shown while the 308-review branch went unused. And for fine and coastal homes at the county level, two engines did not name the agency at all, because its premium sub-brand had no page of its own for a machine to find. Both became the top recommendations in the report, and both are fixable. A machine cannot recommend a brand it cannot find.

Why the machines choose the way they do

Reading 31 answers side by side, the pattern is not mysterious. The engines lean on the Google Business Profile rating and its volume, on review platforms like allAgents and GetAgent, and on what an agency's own website says about itself: several quotes above are the client's own copy read back. Structure does the quiet work underneath, which is why the same agency whose website describes all three branches to machines on every page keeps winning village questions its rivals never even appear for.

One number to keep

AI answers vary by run, location and account, which is why this is a dated snapshot and why the panel reruns monthly. But the direction is not subtle. When five different AI systems, asked the questions your next instruction is already asking, name the same agency 29 times out of 31, that is not luck. It is what a readable, review-backed, properly structured presence looks like from the machine's side of the glass.