- Why do portals always get named?
- Portals hold listings, sale histories and suburb data in structured form, updated constantly, which makes them cheap for a model to use and easy to attribute. An agency website republishing the same feed adds nothing the model needs. The winnable questions are the ones portals hold no data on, which is what selling actually involves.
- Should we track every suburb we cover?
- Track the suburbs you genuinely compete in rather than every one in the footer. Vendors ask about a single area, and an agency that lists fifty areas while dominating four gets a truer read by tracking the four and their neighbours than by spreading allowance across a coverage map.
- Do agent names matter more than the agency name?
- Vendors choose an individual far more often than a brand, so named-agent questions carry the commercial weight. Most tools do not model people as entities, so those prompts have to be written by hand, and the underlying work is making each agent verifiable with a sales record, an area and a licence.
- Does community discussion really affect answers?
- Community and video discussion of commission, agent conduct and selling experience is abundant, opinionated and heavily referenced, which makes it useful corroboration for a model answering a question that involves money. An agency cannot write that material, and knowing what it says is still worth something.
- Is any of this bigger than the portals yet?
- Not yet, and pretending otherwise would be silly. Portals still carry most property discovery, which is a reason to sequence the work rather than ignore the surface. The vendor question is where assistants are already being used, and it is also the question that pays.