How we build a dossier.

We resolve the address you submit to a coordinate pair, a postcode, and a country via Nominatim (OpenStreetMap). Nominatim is free, open, and rate-limited, but accurate enough for our supported countries. We filter results to UK, FR, ES, PT and TR at the API level — addresses outside that perimeter return a 422.

For French and UK addresses we pull real registered transactions from the official registries.

• France — DVF (Demandes de Valeurs Foncières): we query the CEREMA bbox endpoint around the property (initial radius 1.5 km, adaptive up to 20 km). We keep last-24-month sales only, exclude new-builds and zero-size records, and weight each comparable by distance, recency, and size similarity.
• United Kingdom — HM Land Registry Price Paid Data: we run a SPARQL query against the Land Registry endpoint, town-matched over the last 24 months. HMLR does not include square-metre size — we use median size by property type (flat 65 m², house 110 m²) as a placeholder until EPC integration ships.

From the candidate pool we compute a weighted median price-per-m². That figure × your property’s declared size = our fair-value estimate. Sample size and the median radius drive the “data confidence” field shown on the report (≥20 = high, ≥8 = medium, less = low).

Spain, Portugal and Türkiye do not publish free, machine-readable per-sale transaction registries. Instead we benchmark against published price indices, refreshed periodically and labelled with their review date on the report:

• Spain: INE house-price index and Colegio de Registradores regional €/m² medians.
• Portugal: INE and Confidencial Imobiliário €/m² medians, down to the city level where available.
• Türkiye: TÜİK / Endeksa ₺/m² medians, converted to EUR at the live FX rate.

We resolve the property to the most specific area benchmark we hold (city beats national) and multiply by the declared size. Every report from these markets is clearly labelled as an area benchmark, not a per-property appraisal, with the source and review date shown. Where no index granularity is available, a clearly-labelled AI estimate is used as a last resort.

We compute the all-in cost of acquiring the property for the buyer profile you select (currently UK buyer by default). Costs are coded by hand against published 2025-2026 rates from each country’s tax authority — we do not infer rates from AI.

Portugal

• IMT (sliding scale 0-8% by tranche)
• Imposto de Selo (0.8%)
• Legal fees (typical 1.5%)
• Notary fixed fee (~€1,200)

Spain

• ITP regional rate (4-11% by region — Madrid 6% default; switch in profile)
• AJD (1.5% on new-build)
• Legal fees (typical 1%)
• Notary & Registro (~€1,500)

France

• Notaire fees (7.5% all-inclusive — droits de mutation, registration, notary)
• Translation / sworn interpreter (~€500)

United Kingdom

• SDLT tiered 2025-26 rates (0% to £125k, 2%, 5%, 10%, 12%)
• Non-resident SDLT surcharge +2% (since April 2021)
• Solicitor fee (typical £1,500)
• Survey fee (typical £600)
• Land Registry (£135-£500 by band)
• GBP→EUR at the live ECB reference rate (Frankfurter API, cached 12h)

Türkiye

• Tapu harcı (transfer tax, 4%)
• KDV (VAT — 0% on resale)
• Avukat (lawyer, ~1.5%) + döner sermaye registry fee
• DASK earthquake insurance + iskan checks
• TRY→EUR at the live ECB reference rate

The Local Intelligence section is where we name region-specific laws and risks — okupas, AL containment, PEUAT, DPE G ban, leasehold tail risk, etc. The base set of risks per country is hand-curated by our editorial team from primary legal sources (gazette, ministerial decrees, local council resolutions) and refreshed quarterly.

On top of the base set, Gemini personalises the risk for the specific city / postcode — e.g. Eixample Esquerra is flagged as a higher okupa zone than Sant Andreu. When Gemini is unavailable (rate limits, network failure) the report falls back to the curated base set so you never get an empty report.

Rental ROI is computed in two scenarios:

• Long-term let: AI-estimated monthly rent range × 12 ÷ purchase price = gross yield. Net yield subtracts an effective tax rate (24% IRNR Spain, 28% non-resident IRS Portugal, prélèvement bands France, 40% higher-rate income tax UK).
• Short-term let: nightly rate × realistic occupancy × 365 = annual revenue estimate. Licence status (available / restricted / frozen / banned) is determined by the named local regime (PEUAT, AL containment, 90-day rule, etc.).

Lifestyle scoring (healthcare, education, transit, safety) is AI-generated on a 1-5 scale anchored to comparable cities. Cost-of-living index is normalised to London = 100. These are best-effort indicative scores, not census data.

The headline verdict (Good / Caution / Risk) blends three signals: price delta vs comparables, total cost burden as a percentage of purchase, and severity of named local risks. The threshold is currently:

• Good — price within ±5% of comparables AND no critical local risks
• Caution — price 5-15% above OR one high-severity local risk
• Risk — price >15% above OR any critical local risk OR multiple high-severity risks

• We take no referral fee or kickback from solicitors, surveyors, agents, lenders, or developers — that is what corrupts analysis. We may connect you to a vetted independent professional, but you pay them — or us, for coordination — directly; we never take a cent from them.
• We do not produce a chartered surveyor’s valuation suitable for mortgage, insurance, or court purposes.
• We do not inspect properties physically.
• We do not guarantee AI-generated narrative is free from hallucination — every fact-bearing field cross-checks with the underlying data layer.

For the regulatory and limit-of-liability boilerplate, see our Legal Disclaimer.

See the full list of public data we pull from on the Data Sources page — including the version, refresh cadence, and licence terms of each.