Skip to content
Signal.

new homes

Why your development website is invisible to ChatGPT, and how to fix it

Most new-homes sites render in the browser, so search engines and AI assistants see an empty page. What they need instead, and how we build it.

Photo of Graham Wilsdon

Graham Wilsdon

· 5 MIN

TL;DR

  • Buyers increasingly ask an AI assistant “what’s available near X under Y” before they open a portal.
  • Most development websites are built in a design tool and rendered in the browser. Crawlers don’t run that code, so they see nothing.
  • The fix is a static content layer underneath the design: every plot, price and spec in plain HTML, a sitemap, and an llms.txt, regenerated whenever the plot list changes.
  • Done properly, the scheme becomes the answer when a buyer asks the question.

What a crawler actually sees

Open a typical development microsite, view the source, and look for the plot list. On most of them it isn’t there. The page is a shell that loads a script, and the script draws the site plan, the house types and the availability after the page opens.

A person sees a beautiful site. Googlebot sees a shell. ChatGPT, Claude and Perplexity, which fetch pages when a user asks a question, see the same shell. None of them can tell that plot 12 is a three-bed released last week at a price the buyer can afford.

Why it matters more now than last year

Search engines have been forgiving about this for a while. AI assistants aren’t. When a buyer asks “three-bed new builds in Lisburn under £300k with a garden”, the assistant goes looking for pages that answer that in plain text. A PDF brochure doesn’t. A portal listing answers it badly. A development site with a readable plot list answers it exactly.

The schemes that are readable get cited. The rest don’t exist.

What we build instead

We keep the design. Developers have paid for it and it works for the people who visit. Underneath it we add a layer a machine can read:

  1. A static content block in the HTML with every plot, house type, price, status, the site plan, the specification and the local area, generated from the same data the visitor sees. It’s removed once the app mounts, so the person never sees two versions.
  2. A sitemap and an llms.txt that tell crawlers and AI fetchers what’s on the site and what it’s for.
  3. Structured data describing the development, the house types and the selling agent, so search engines can treat the scheme as an entity rather than a page.
  4. A rebuild on every change. When the agent marks a plot sold, the content layer, sitemap and llms.txt regenerate. The plot list the buyer’s assistant reads is the one that’s true today.

Enquiries route straight into the selling agent’s WhatsApp, where the buyer is qualified against the plots that are actually released.

How to check your own site

Open the development page in a browser, right-click, and choose View Page Source. Search for a plot number or a price. If it isn’t in the source, the assistants can’t see it either.

If it isn’t, we can fix it without changing how the site looks.


Update, October 2026: we ran this check across the websites of all 260 Home Builders Federation members. 46% hide plots or prices from a plain fetch and 17% refuse the request outright. The full figures and method are in How 260 UK housebuilders answer a buyer online.