Why the Hogarfax report doesn't use artificial intelligence

When years of savings are on the line, "sounds convincing" isn't good enough. That's why the Hogarfax report doesn't use artificial intelligence to generate its findings: it uses explicit rules and calculations over official data, and when a data point is missing, it says so — it never fills the gap with a plausible-looking number.

The problem with a report that only "sounds" reliable

A language model generates the most probable answer based on text patterns, not a verified calculation over concrete data. That's fine for summarising a document or drafting an email. It isn't fine for telling you whether a parcel sits in a flood zone, what a plot's planning classification is, or what buying a property actually costs — there, the answer has to be what the official data says, not what a text pattern suggests is likely.

The risk isn't hypothetical: generative models can produce a figure or a claim with the same apparent confidence whether it's correct or not. In a property purchase decision, that unearned confidence is exactly what a report can't afford.

How the Hogarfax report actually works

• Every check is a query against a specific source, with its own date and scope (the property, the parcel, the census section or the municipality) — not an estimate from a model trained on generic text.
• The calculations are deterministic: the same property, with the same input data, always produces the same result. There's no variable interpretation from one run to the next.
• When a data point is missing, the report says so. It doesn't estimate a "reasonable" value to avoid leaving a blank — a clearly marked gap is more honest than an invented number that looks correct.
• Collective intelligence comes from people, not from a model: what other users have observed, visited or reported about a specific property is labelled as exactly that — a real user's contribution, not an automated inference.

The difference in practice