We measured 502 AI answers across five home improvement trades: boiler replacement, EV charger installation, garden rooms, heat pumps and London loft conversions. Three engines, real buying questions, July 2026. The engines named 577 different firms across those answers, but the top of every list is short, and most local firms are not on it at all. This page shows who gets named, why the rest are absent, and what actually changes it. The full numbers live on the Aeova AEO Index and the individual readings.
Who the engines name, trade by trade
Boiler replacement: BOXT led on all three engines, with British Gas and Heatable the next most named. EV charging: Smart Home Charge led, ahead of Ohme and Octopus Energy. Garden rooms: Green Retreats led decisively, ahead of Noah Garden Rooms and Dunster House. Heat pumps: the three engines disagreed on the leader, with Octopus Energy, Heat Geek and Aira the most named. London loft conversions: LCCL Construction led, ahead of Simply Loft and The Loft Converter.
Across the same 502 answers, the verdict the engines gave flipped 40 times on repeat runs. The AI's opinion on who to recommend is measurably unstable, which is bad news for anyone taking a single answer as truth and good news for anyone able to be present consistently while a competitor is not.
Why good local firms are absent
Two reasons come up in every trade we measure. First, the engines corroborate machine-readable evidence: structured pages that say who you are, where you work and what verified customers say. Most trade websites state none of it in a form a machine can lift, however strong the work is. Second, some sites actively refuse the machines. In an August 2026 Aeova scan of roughly 490 home-improvement firm websites, nine served a full Cloudflare block against all major AI crawlers, and several of those also returned errors to AI fetchers at the firewall while serving Google normally. Those firms have, whether they know it or not, told every AI assistant to look elsewhere. A growing number of sites have also added llms.txt files automatically via their SEO plugin; that alone does not fix the problem, and Google has said it does not use the format.
What changes it
The sequence is the same for a sole-trader plumber and a national installer. Measure where you actually stand, because a single answer is directional and self-assessment is unreliable. Remove anything blocking the crawlers that feed the answer engines. Put your evidence where machines can read it: services, locations, guarantees, accreditation, reviews, consistent company details everywhere. Then re-measure against the same questions, because 19% of same-question repeats change the recommendation and only repeated measurement separates real movement from noise. The practical guide covers the steps, our method is public, and pricing starts at a £500 pilot.
Where to start: the free ten-question audit, run on the questions your customers actually ask.
Common questions
How does a trade firm get recommended by ChatGPT or Google's AI?
Three things, in order: let the AI crawlers read your site (many trade sites unknowingly block them in robots.txt or a firewall), give the engines machine-readable evidence about who you are, where you work and what customers say about you, and be named in the public sources the engines corroborate against. None of it is a toggle in your SEO plugin. Source: Aeova's published measurement method, aeovapartners.com/how-we-measure/.
Which home improvement firms does AI recommend today?
In our July 2026 measurement of 502 answers across five trades (boiler replacement, EV charging, garden rooms, heat pumps, London loft conversions), each trade's answers concentrated on a short list: BOXT led boiler replacement on all three engines, Green Retreats led garden rooms, Smart Home Charge led EV charging, and in heat pumps the three engines disagreed on the leader. One run is directional, not a verdict. Full per-trade data: aeovapartners.com/aeo-index/.
We rank well on Google. Why are we absent from AI answers?
Ranking and recommendation are different systems. AI engines build shortlists by corroborating machine-readable evidence across sources, and some trade websites actively block the AI crawlers that gather it: an August 2026 Aeova scan of roughly 490 home-improvement firm websites found nine serving a full Cloudflare block against all major AI bots, some returning errors to AI fetchers at the firewall while serving Google normally.
Does adding llms.txt fix AI visibility?
No, and we would be selling you snake oil if we said so. llms.txt is unproven and Google has said it does not use it. It is appearing on trade websites anyway because SEO plugins now generate it automatically. Access, evidence and corroboration are the levers; a text file is, at best, a small part of the last one.
Updated July 2026