The developer nobody could find once buyers stopped searching

The situation

Off-plan property is bought on research. A buyer commits significant money to a building that does not exist, on the strength of what they can establish about the developer, the escrow arrangements and the handover record. That research used to begin with a search engine.

Increasingly it begins with an assistant. The buyer asks a question in plain language and receives an answer assembled from sources the model considers reliable. If the developer is not among those sources, the developer is not in the conversation, regardless of where its pages rank.

The problem

This developer had an agency, a respectable content programme and rankings that looked healthy. Enquiries had been declining for three quarters and nobody could explain it.

We built a list of forty questions a real buyer would ask, in the words they would use, and put them to the major assistants. The developer was named in three.

The reasons were unglamorous. Its own pages answered questions about its communities but almost never answered questions about the category: what escrow protects, what happens when a handover slips, how payment plans actually work. Assistants answering those questions reached for sources that did.

Its entity information was inconsistent. The legal name appeared three ways across its own site, its listings and its filings. Faced with ambiguity a model hedges, and hedging means leaving you out.

And nothing it published was attributable. Pages carried no author, no date, and no sourced figures, which is precisely the profile a model discounts when it has to choose what to stand behind.

What we did

We told the developer to stop producing community pages for two quarters. That was an uncomfortable recommendation to make to a marketing team whose targets were set in campaigns, and it followed from the diagnosis: more pages of the kind that were not being cited would not change anything.

Instead we rebuilt the content programme around the category questions, one page per question, answering it fully in the opening paragraph before any context or caveat. A passage that resolves a question on its own is a passage an assistant can quote on its own.

Every claim was given a source and a date, linked at the claim rather than collected at the bottom. Where the developer had proprietary knowledge, its own handover record across four communities, we published it with the methodology, which turned an asset it had been sitting on into the most citable thing it owned.

We fixed the entity problem in parallel: one legal name, used identically everywhere, with structured data describing the organisation rather than only its articles, and named authors with real records attached to the technical pages.

Measurement changed too. We retired rank tracking as the headline and replaced it with a scheduled citation check against the same forty questions, recording not just whether the developer appeared but what was said about it.

The outcome

Six months later the developer was cited in twenty-seven of the forty questions, against three at the start.

Enquiries attributed to assistant referrals rose about eighteen per cent. Volume from that channel remains small next to paid media, and it converts markedly better, which is what you would expect from somebody arriving after a neutral third party has already endorsed your position.

One outcome was not anticipated. Two of the category pages began ranking conventionally as well, because the qualities that make a passage citable, a direct answer, a sourced figure, a named author, are not different from the qualities search has always claimed to reward.

Being ranked is being in the queue. Being cited is being the answer. The developer had spent three years optimising its place in a queue that fewer buyers were joining.