Follow
Follow The Observatory
Belles Demeures is the recurring organic first-row identity. Home Select opens the early AI tables, while YouTube is first throughout the citation record.
Quick answer
Do Paris property search and AI answers put the same sources first?
No. bellesdemeures.com is first in all 10 published organic tables, while the AI opening tables print the same portal as Belles Demeures: a label difference to preserve rather than merge. Across AI Overviews, homeselect.paris is first in the first five of nine opening tables, while youtube.com is first in all eight citation tables. These are distinct source roles and dated sequences, not one property-market ranking. The September 14–20 comparison adds the practical exception: Home Select is second in both AI lists while outside the organic top five.
At a glance — key facts
- The portal’s organic recurrence carries an alias caveat. bellesdemeures.com is first in 10/10 organic tables, while AI opening tables print the name Belles Demeures.
- Opening first rows follow a different sequence. Home Select is first in the first five of nine tables, YouTube in one, then Paris Ouest Sotheby’s in the final three.
- Citation first rows are consistent. youtube.com is first in 8/8 published citation top-five tables.
- The matched week still exposes a buying adviser. Home Select is second in both AI lists and outside the organic top five; DPE questions separately include commercial intermediaries beside public information.
A persistent organic portal, with an identity caveat
The retained organic series contains 10 tables from July 6–September 27, 2026, with publication gaps. The portal prints as bellesdemeures.com in those organic tables and as Belles Demeures in AI opening tables such as September 14–20, where it is third. Those labels remain separate: their values are not added or treated as independent businesses. The domain is first in all 10/10 tables, from July 6–12 to September 21–27.
This is portal visibility, not a finding about the similarly named Belles Demeures de France agency. The recurring portal identity makes the different first rows in the AI tables more informative: a buying adviser, local brokerage and content host enter different positions in the answer.
Citation recurrence differs from the changing opening order
The exact label youtube.com is first in 8/8 published Paris citation top-five tables from July 27 to September 20, 2026. That counts published tables, not queries, videos or independent creators. The host-level record does not identify every underlying channel or property adviser.
Opening results follow another sequence: homeselect.paris occupies the first row in the first five published weeks, youtube.com in the sixth, and parisouest-sothebysrealty.com in the final three. The span runs from July 27–August 2 through September 21–27; the changeovers are visible in August 31–September 6 and September 7–13.
The difference is a sequence of published first rows, not a growth claim. Historical aggregation wording changes and incomplete August evidence prevent treating the nine openings as a normalized share series. The citation sequence is linked in the evidence basis below.
Home Select’s visibility changes with the surface
The September 14–20, 2026 readings share seven comparable runs on Google desktop in France and a source context of 105 observed queries per run.
| Published selection | First source | Governing order | homeselect.paris |
|---|---|---|---|
| Organic top five | bellesdemeures.com | Frequency of holding the top organic position | Outside the published five |
| AI opening top ten | parisouest-sothebysrealty.com | Cumulative opening volume | Rank 2 |
| AI citation top five | youtube.com | Cumulative citation volume | Rank 2 |
Home Select’s declared service is property search for buyers. Its second-place positions bring that adviser role into both AI selections. The respective first rows identify a portal, a local brokerage domain and a content host.
Restricting the opening selection to its first five rows retains Home Select at rank 2. The longer list therefore does not explain this particular membership difference. Equal depth does not make the three measures or their denominators equivalent, and omission from the organic five is not absence from all organic search.
“A buying adviser can become visible in an AI answer without joining the organic shortlist beside it.”
The Observatory
For a Paris property business, the implication is a more precise visibility question: being prominent in organic results, opening an AI answer and supplying a cited source are separate observations. The surface-divergence Guide explains why they should not be collapsed into a single performance score.
A DPE question does not exclude property intermediaries
The exact query “comprendre le diagnostic DPE pour un appartement ancien à Paris” asks for an explanation of an older apartment’s diagnostic. Public-information sources appear in each of its four published top-five selections. Property intermediaries appear in two of them.
All four query readings cover seven comparable runs. Position below is the median of each source’s best organic position when present; presence records how many runs contained it.
| DPE window · 2026 | Public-information example: median position; presence | Intermediaries in the published top five |
|---|---|---|
| June 22–28 | economie.gouv.fr: 1; 7/7 runs | Neither Massimo’s nor Home Select’s exact domain |
| July 20–26 | service-public.gouv.fr: 1; 6/7 runs | Neither exact domain |
| August 17–23 | ecologie.gouv.fr: 1; 7/7 runs | massimoimmobilier.com: 4; 7/7 runs. homeselect.paris: 5; 5/7 runs |
| September 14–20 | service-public.gouv.fr: 1; 7/7 runs | massimoimmobilier.com: 3; 3/7 runs. homeselect.paris is outside the selection |
Across these four tables, massimoimmobilier.com appears in 2/4 and homeselect.paris in 1/4. The Massimo agency site and Home Select’s buyer-service description identify the commercial roles entering this technical-information question.
Massimo’s median position moves from 4 to 3 while its presence falls from 7/7 to 3/7 runs. Calling that a simple improvement would discard part of the evidence. The useful finding is repeated agency inclusion beside public information, not an exclusive division between official explanations and commercial property discovery. The official DPE guidance remains the place to understand the diagnostic itself.
A model’s account of discretion remains an account, not a census
In its July 13, 2026 native reading, Claude Sonnet 4.6 distinguishes trust based on public documentation from professional competence. In a separate September 21 native reading, the same model discusses under-represented mid-sized specialists, real-estate family offices and patrimonial advisers. Each publication records six prompts on one dated run.
These are two instances of a documentation-versus-discretion theme. They do not measure the number, quality or transaction share of less visible firms. Nor should the model’s organizational descriptions be used as a verified ownership map. The Official vs Narrative Layer reference explains the boundary between an attributed model account and established institutional facts.
Questions readers ask
Does the Paris organic order predict the AI Overview order?
The matched week shows different orders, including Home Select’s AI positions outside the organic top five. That is a dated contrast, not a predictive rule.
Is YouTube’s repeated position evidence for one adviser’s success?
No. The 8/8 count concerns the exact host label youtube.com in published citation tables. It does not resolve all creators or attribute their visibility to one property firm.
Do DPE results assess an agency’s technical advice?
No. They record inclusion, position and run-level presence for one query. Evaluating the accuracy of an intermediary’s advice requires evidence beyond its search appearance.
Has Paris representation improved across the period?
The record supports recurrence and dated contrasts. Gaps, aggregation changes and unresolved source identities prevent a single comparable June–September performance trend. The model narratives do not fill those measurement gaps.
Evidence basis and limits
The Paris archive reviewed through September 29, 2026 contains 82 publications: 10 organic and nine opening readings reaching September 21–27; eight citation and 14 query readings ending at September 14–20; and 28 Lens plus 13 Model Watch publications reaching September 21. The organic history begins July 6–12. Model Watch can reuse Lens responses, so publication totals are not independent experiments.
The eight citation tables are July 27–August 2, August 3–9, August 10–16, August 17–23, August 24–30, August 31–September 6, September 7–13 and September 14–20, 2026.
These are Google desktop observations in France within a French Paris/Île-de-France configuration, not a matched language experiment. Family gaps, changes from mean-based to cumulative aggregation, incomplete runs and distinct brand/domain labels limit longitudinal comparison. Host rows do not identify every creator. The observations do not establish service or property quality, transaction suitability, investment merit or endorsement.
The Observatory does not claim to measure real-world revenue, market share, or business leadership unless explicitly stated. It publishes observed AI and search positioning within a defined scope and time window.
Intelligence by AAI Kernel. Published by The Observatory.
Continue the reading
Home Select’s contrast depends on recognizing a buying adviser, just as portal visibility depends on keeping similar names distinct. The Paris source context separates those identities before a business compares itself with the results. The Luxury Real Estate context carries the DPE finding outward: information for a property check can come from both public sources and private services.
