A complete page per agency
Address, opening hours, team, services, area covered, reviews: it is the source that most often comes with a first-place recommendation.
25% against 10% cited first when it is read AI Barometer · Real estate · Edition #1 · October 2026
Real estate: when you ask AI for an estate agent, who does it recommend?
When someone asks ChatGPT or Gemini for “an agency to sell my flat in Toulouse” or “someone to manage my studio in Montpellier”, does the AI name a network agency, an independent agency, a self-employed agent or a listing portal? We analysed nearly 13,500 answers to real estate questions that name no company: selling, valuation, buying, buy-to-let, renting and letting management.
Answers collected from April 1, 2026 to October 5, 2026 Published October 13, 2026 ChatGPT and Gemini, queried through their APIs with web search
53–62% of answers to a local question recommend a specific agency first, network or independent (manual coding – automatic measure).
78% when people look for a professional, the AI names an agency first. When they look for a property, portals come first (69%).
53% of the agencies named by the AI are independent agencies.
1 in 4 only: the same question, asked again a week later, gives the same first recommendation.
01 · Agency or portal
AI does not replace the estate agent: it replaces the shop window
On a local question, the first recommendation is a specific agency, network or independent, in 62% of answers according to the automatic measure (95% CI 59–65), and in 53% of a sample of 240 answers coded by hand. Portals and other online services come first in 26% of cases. It all depends on what the person is looking for.
What the AI recommends first, by what the person is looking for
Local questions, % of answers
- Looking for a professional
- Looking for a property
- Looking for information
See the data
| A specific network agency | An independent agency | A network, no agency named | A portal or online service | Another player or none | |
|---|---|---|---|---|---|
| Looking for a professional | 45.9% | 31.6% | 2.9% | 14.9% | 4.7% |
| Looking for a property | 9.8% | 11.5% | 2.8% | 69.4% | 6.5% |
| Looking for information | 9.1% | 15.4% | 4.8% | 45.4% | 25.3% |
What the AI recommends first, by need
Local questions, % of answers
- Selling
- Valuation
- Buying
- Buy-to-let
- Renting (tenant)
- Letting management
- Commercial real estate
See the data
| A specific network agency | An independent agency | A network, no agency named | A portal or online service | Another player or none | |
|---|---|---|---|---|---|
| Selling | 51.1% | 25.7% | 4.1% | 14.9% | 4.2% |
| Valuation | 38.0% | 23.9% | 4.3% | 30.4% | 3.4% |
| Buying | 28.7% | 22.8% | 2.8% | 36.8% | 9.0% |
| Buy-to-let | 28.4% | 33.8% | 2.9% | 27.3% | 7.6% |
| Renting (tenant) | 37.7% | 16.7% | 2.5% | 35.2% | 7.8% |
| Letting management | 41.3% | 31.4% | 3.0% | 17.1% | 7.3% |
| Commercial real estate | 15.9% | 36.9% | 1.4% | 13.9% | 31.9% |
What the AI recommends first, by town size
Local questions, % of answers
- Large city
- Mid-sized town
- Small town
See the data
| A specific network agency | An independent agency | A network, no agency named | A portal or online service | Another player or none | |
|---|---|---|---|---|---|
| Large city | 36.3% | 27.1% | 4.1% | 22.7% | 9.5% |
| Mid-sized town | 36.0% | 24.0% | 2.8% | 29.6% | 7.7% |
| Small town | 36.2% | 30.4% | 2.7% | 22.6% | 8.0% |
77%
of first recommendations go to a specific agency when people want to sell, against 52% when they want to buy: when buying, portals take over.
34%
of first recommendations go to an independent agency on buy-to-let questions, against 28% to a network agency.
67%
of answers in small towns recommend a specific agency first, against 60% in mid-sized towns: town size changes little.
What it changesWhen people look for a professional, the AI names an agency; when they look for a property, it sends them to portals. For an agency, the AI battle is fought on service requests — selling, valuation, letting management — far more than on listings.
02 · Networks and independents
Independent agencies compete on equal terms with the networks
At least one specific agency is named in 82% of local answers, 3.4 on average. And the AI does not reserve that place for big brands: 53% of the agencies it names are independent agencies.
Share of answers citing at least one player of each family
All retained questions. An answer can cite several families.
- Independent agencies 67%
- Agency networks 63%
- Listing portals 38%
- Online valuation 27%
- Online letting & management 15%
- Property managers 14%
- Self-employed agent networks 10%
- Banks & brokers 7%
- Insurers & services 3%
The most cited players, all questions combined
Share of answers citing the player. Hover a bar for the confidence interval.
- Orpi 39%
- Laforêt 30%
- SeLoger 25%
- Meilleurs Agents 22%
- Century 21 21%
- PAP 18%
- Nestenn 17%
- Guy Hoquet 10%
- Citya 9%
- Manda 8%
- Figaro Immobilier 8%
- Leboncoin 7%
- L'Adresse 7%
- Bien'ici 6%
- Arthurimmo 5%
How to read it: Orpi is cited in 39% of answers, and first in 13%. Independent agencies, counted together, are cited in 67% of answers. The panel's networks are favoured (see the network note and the method).
See the data
| Player | Family | Cited | 95% CI | In the top 3 | First |
|---|---|---|---|---|---|
| Orpi | Agency networks | 39.2% | 36.1–43.3 | 27.9% | 12.6% |
| Laforêt | Agency networks | 30.3% | 26.8–33.7 | 18.9% | 6.1% |
| SeLoger | Listing portals | 24.7% | 21.3–28.0 | 17.9% | 7.7% |
| Meilleurs Agents | Online valuation | 22.3% | 19.8–24.8 | 14.1% | 8.1% |
| Century 21 | Agency networks | 20.7% | 18.3–23.1 | 11.8% | 3.0% |
| PAP | Listing portals | 18.1% | 15.4–21.1 | 11.4% | 3.6% |
| Nestenn | Agency networks | 17.1% | 14.2–20.1 | 9.6% | 2.7% |
| Guy Hoquet | Agency networks | 10.0% | 7.4–12.7 | 4.5% | 1.0% |
| Citya | Property managers | 9.2% | 7.0–11.5 | 6.0% | 2.2% |
| Manda | Online letting & management | 8.0% | 6.3–10.0 | 4.1% | 1.3% |
| Figaro Immobilier | Listing portals | 8.0% | 6.4–9.5 | 4.7% | 1.4% |
| Leboncoin | Listing portals | 7.2% | 5.5–9.0 | 3.7% | 0.8% |
| L'Adresse | Agency networks | 6.6% | 4.5–8.7 | 3.7% | 1.2% |
| Bien'ici | Listing portals | 5.7% | 4.6–6.9 | 2.6% | 0.1% |
| Arthurimmo | Agency networks | 4.7% | 2.7–6.4 | 3.3% | 2.1% |
| Independent agencies (all) | Independent agencies | 66.6% | — | 48.7% | 26.0% |
34%
of local answers name at least five specific agencies: the AI offers a list, not a single name.
51%
of local answers put at least one network agency and at least one independent agency side by side.
68%
of local answers cite at least one independent agency, against 64% at least one network agency.
What it changesAI does not favour big brands: it cites the agencies it can read and locate. A well-described independent agency is as likely to be named as a network agency.
03 · Networks
Estate agency networks: cited branch by branch
Orpi is the most cited network, in 41% of local answers, and stays first outside the questions built from its own agencies (39.5%). When the AI cites a franchise, it nearly always names a specific branch; when it cites a self-employed agent network, it usually stops at the brand.
Share of local answers citing the network, and naming one of its agencies
Local questions. Hover a bar for details.
- Orpi 41% agency 37%
- Laforêt 32% agency 29%
- Century 21 22% agency 20%
- Nestenn 18% agency 17%
- Guy Hoquet 11% agency 10%
- Citya 9% agency 7%
- L'Adresse 7% agency 7%
- Arthurimmo 5% agency 5%
- Foncia 4% agency 3%
- ERA Immobilier 3% agency 3%
- Solvimo 3% agency 2%
- Imop 3% agency 0%
- Capifrance 3% agency 1%
- Square Habitat 3% agency 2%
- IAD 3% agency 1%
Conversion = share of answers citing the network that also name one of its agencies or agents. “Outside its own questions” = presence measured only on questions built from agencies of other networks or from independent agencies.
The questions are built from the towns and services of a panel of agencies, including agencies of the Orpi, Laforêt, Guy Hoquet, L'Adresse and Arthurimmo networks: these networks are cited more often there. The last column removes that effect. Century 21 and Nestenn, third and fourth, are almost absent from the panel.
See the data
| Network | Cited | Agency named | 95% CI | Cited first | Conversion | Cited outside its own questions |
|---|---|---|---|---|---|---|
| Orpi | 41.0% | 37.0% | 33.7–40.8 | 13.1% | 90% | 39.5% |
| Laforêt | 31.7% | 29.3% | 26.0–33.0 | 6.4% | 92% | 29.8% |
| Century 21 | 21.7% | 19.5% | 17.2–21.9 | 3.1% | 90% | 21.7% |
| Nestenn | 18.2% | 17.0% | 14.4–19.8 | 2.9% | 93% | 18.2% |
| Guy Hoquet | 10.6% | 9.9% | 7.3–12.8 | 1.1% | 94% | 9.5% |
| Citya | 9.3% | 7.2% | 5.0–9.5 | 2.2% | 78% | 9.3% |
| L'Adresse | 7.1% | 6.6% | 4.5–8.5 | 1.3% | 93% | 4.2% |
| Arthurimmo | 5.0% | 4.7% | 2.5–6.5 | 2.2% | 93% | 0.7% |
| Foncia | 4.2% | 3.0% | 2.3–3.8 | 0.9% | 72% | 4.2% |
| ERA Immobilier | 3.4% | 3.1% | 2.2–4.2 | 0.4% | 91% | 3.4% |
| Solvimo | 2.9% | 2.3% | 1.6–3.2 | 0.1% | 81% | 2.9% |
| Imop | 2.7% | 0.4% | 0.2–0.7 | 0.3% | 16% | 2.7% |
| Capifrance | 2.7% | 1.4% | 1.0–1.8 | 0.3% | 51% | 2.7% |
| Square Habitat | 2.7% | 1.9% | 1.0–3.1 | 0.6% | 72% | 2.7% |
| IAD | 2.5% | 0.6% | 0.3–1.1 | 0.4% | 24% | 2.2% |
| Human Immobilier | 1.9% | 1.8% | 0.9–3.1 | 0.4% | 92% | 1.9% |
| Nexity | 1.9% | 1.2% | 0.8–1.5 | 0.2% | 60% | 1.9% |
| Stéphane Plaza Immobilier | 1.8% | 1.5% | 0.9–2.2 | 0.2% | 86% | 1.8% |
| Engel & Völkers | 1.7% | 1.6% | 0.9–2.2 | 0.3% | 94% | 1.7% |
| Barnes | 1.5% | 1.2% | 0.5–1.9 | 0.3% | 77% | 1.5% |
90%
or more of the citations of each of the top four networks name a specific agency: the AI knows franchises branch by branch.
24%
of IAD's citations name a specific agent (51% for Capifrance): the AI knows self-employed agent networks by their brand, rarely by their agents.
39.5%
of local answers cite Orpi outside the questions built from its own agencies: the ranking is not just a panel effect.
What it changesBeing cited is not enough: on a local question, the network that wins is the one whose agency the AI knows — its name, its address, its page. For a self-employed agent network, the challenge is to make each agent identifiable.
04 · Town by town
In each town, the AI knows 7 networks and 13 independent agencies
Town by town, we counted the networks and independent agencies the AI names regularly (in at least 3 answers). In the median town, 7 networks and 13 independents compete for the recommendation.
Share of towns where the network has a regularly named agency
At least 3 local answers name one of its agencies · towns with at least 20 local answers
- Laforêt 92%
- Orpi 91%
- Nestenn 84%
- Century 21 83%
- Guy Hoquet 46%
- Citya 46%
- L'Adresse 36%
- Foncia 34%
- ERA Immobilier 31%
- Solvimo 31%
Share of local answers naming one of the network's agencies, by town size
Local questions, the eight most cited networks
| Large city | Mid-sized town | Small town | |
|---|---|---|---|
| Orpi | 43% | 35% | 36% |
| Laforêt | 30% | 35% | 20% |
| Century 21 | 17% | 23% | 15% |
| Nestenn | 10% | 21% | 17% |
| Guy Hoquet | 9% | 12% | 8% |
| Citya | 9% | 7% | 6% |
| L'Adresse | 4% | 10% | 4% |
| Arthurimmo | 4% | 4% | 8% |
92%
of the towns studied have a regularly named Laforêt agency, and 91% a Orpi agency: the two networks most rooted in AI answers.
73%
of answers in large cities cite at least one independent agency, against 66% in mid-sized towns and 69% in small towns.
The towns studied are those of the panel's agencies: the networks present there are favoured. No result is published town by town.
What it changesCompetition plays out town by town, and it is dense: a network does not only face the other brands, but a dozen independent agencies.
05 · Needs
Selling, renting, letting management: each need has its winners
Agency networks dominate selling. Portals are on a par with them on renting and buying, valuation tools stand out on valuation, and online management is taking hold in letting management.
Share of answers citing at least one player of the family, by need
All retained questions, in %
| Independent agencies | Agency networks | Listing portals | Online valuation | Online letting & management | Property managers | |
|---|---|---|---|---|---|---|
| Selling | 72 | 76 | 29 | 32 | 9 | 10 |
| Valuation | 65 | 68 | 46 | 55 | 9 | 6 |
| Buying | 56 | 53 | 49 | 35 | 7 | 6 |
| Buy-to-let | 72 | 58 | 38 | 31 | 20 | 14 |
| Renting (tenant) | 55 | 63 | 57 | 15 | 11 | 17 |
| Letting management | 80 | 73 | 15 | 10 | 37 | 27 |
| Commercial real estate | 74 | 51 | 38 | 6 | 4 | 7 |
The three most cited players for each need
| Need | The 3 most cited players | Independents | Portals |
|---|---|---|---|
| Selling | Orpi 52% · Laforêt 37% · Century 21 29% | 72% | 29% |
| Valuation | Meilleurs Agents 50% · Orpi 40% · SeLoger 32% | 65% | 46% |
| Buying | Orpi 32% · Meilleurs Agents 31% · SeLoger 30% | 56% | 49% |
| Buy-to-let | Orpi 37% · Laforêt 28% · Meilleurs Agents 24% | 72% | 38% |
| Renting (tenant) | SeLoger 45% · Orpi 40% · Laforêt 34% | 55% | 57% |
| Letting management | Orpi 47% · Laforêt 43% · Manda 23% | 80% | 15% |
| Commercial real estate | CBRE 27% · BNP Paribas Real Estate 27% · SeLoger 22% | 74% | 38% |
50%
of valuation answers cite Meilleurs Agents, first in 22%: the online tool often answers before the agency.
45%
of renting answers cite SeLoger, the most cited player, ahead of every network.
37%
of letting-management answers cite an online management player (Manda: 23%), against the networks and property managers.
Condominium management, life annuity sales and new-build each have fewer than 200 answers: they are not detailed here. As a signal, property managers and independent firms dominate them.
What it changesAn agency has several competitors, not one: a neighbouring network on selling, a portal on renting, a valuation tool, an online management platform. Each service has to be defended on its own.
06 · Prices and valuation
Price per m², valuation, notary: the AI looks for benchmarks
22% of answers give a price per square metre and 29% point to a notary. The figures come first from valuation tools, portals and public data.
Sources cited when the AI gives a price per m²
Share of ChatGPT answers with a price per m² that cite each source
- Meilleurs Agents 56%
- SeLoger 53%
- DVF or a public website 55%
- Notaries 34%
- An agency page 32%
22%
of answers give a price per m², and 38% of buying answers.
29%
of answers mention the notary, and 55% of buying answers: the notary is presented as the neutral judge of value.
55%
of valuation answers cite an online tool; yet a specific agency still comes first in 62% of local valuation answers.
What it changesAn agency that publishes price benchmarks by neighbourhood becomes a price source for the AI, just like portals and public data.
07 · Sources
What the AI reads before recommending an agency
ChatGPT cites 7.4 sources per answer on average. Network agency pages appear in 54% of answers, and public and notarial websites make up 18% of the sources cited.
Sources cited by ChatGPT
% of cited sources
- Public sites & notaries 18%
- Independent agency sites 17%
- Network branch pages 16%
- Listing portals 9%
- Network national sites 9%
- Valuation & price data 4%
- Documents (PDF, fee schedules) 4%
- Media & guides 3%
- Online management & agencies 3%
- Directories & reviews 2%
See the data
| Source type | Share of cited sources | Answers citing at least one |
|---|---|---|
| Public sites & notaries | 18.3% | 41.9% |
| Independent agency sites | 17.3% | 57.8% |
| Network branch pages | 16.2% | 54.2% |
| Listing portals | 9.1% | 35.6% |
| Network national sites | 9.0% | 40.5% |
| Valuation & price data | 4.3% | 24.8% |
| Documents (PDF, fee schedules) | 4.3% | 25.1% |
| Media & guides | 3.4% | 19.4% |
| Online management & agencies | 2.6% | 15.5% |
| Directories & reviews | 1.7% | 10.2% |
How often the cited network is recommended first
Depending on what the AI read from it before answering
- Its agency page 25%
- Another page of the network 16%
- No page of the network 10%
This is a correlation: the AI also reads the page of the agency it has chosen. The gap remains outside the networks whose agencies were used to build the questions (18.6% against 10.2%).
See the data
| What the AI read from the cited network | Cited first | 95% CI | Excluding the panel's networks | 95% CI |
|---|---|---|---|---|
| Its agency page | 24.8% | 23.7–25.9 | 18.6% | 16.4–20.4 |
| Another page of the network | 16.1% | 14.7–17.6 | 15.6% | 14.1–17.5 |
| No page of the network | 10.4% | 9.0–12.3 | 10.2% | 8.1–12.0 |
50%
of answers give an address when the AI read an agency page, against 19% without one.
What it changesThe agency page is the source a network fully controls. Complete, up to date and readable, it goes hand in hand with a first-place recommendation and an address.
08 · ChatGPT and Gemini
Gemini cites more independents, but without addresses
Gemini cites at least one independent agency in 94% of its answers, against 66% for ChatGPT. It gives an address in only 5% of cases, against 34%, and more often puts a network first without a specific agency (15.5% against 2.7%).
Share of answers citing the player, by AI
Gemini observed since July 31, 2026
- Orpi
- Laforêt
- SeLoger
- Meilleurs Agents
- Century 21
- PAP
- Nestenn
- Guy Hoquet
Gemini has been observed since July 31, 2026, on more than 500 answers: these gaps are signals, to be confirmed in the next edition.
See the data
| ChatGPT | Gemini | |
|---|---|---|
| Orpi | 39.6% | 31.3% |
| Laforêt | 30.4% | 29.6% |
| SeLoger | 24.9% | 20.0% |
| Meilleurs Agents | 22.7% | 13.2% |
| Century 21 | 21.1% | 11.5% |
| PAP | 18.3% | 12.2% |
| Nestenn | 17.4% | 10.5% |
| Guy Hoquet | 9.8% | 15.3% |
| Citya | 9.5% | 4.0% |
| Manda | 7.8% | 14.7% |
What it changesThe same network does not have the same visibility on every AI. Gemini relies more on directories and review sites: being present and well rated there matters as much as your own website.
09 · Volatility
An AI recommendation is never secured
Each question is asked again every week. From one week to the next, the first recommended player changes in 75% of cases.
25% of cases only give the same first player as a week earlier.
5% of cases only give the same top three.
26% of the players cited reappear from one week to the next, on average.
29% of first recommendations go to the three players most often cited first: recommendations are widely spread.
What it changesA screenshot proves nothing. To know whether an agency is recommended, you need to measure continuously, across a large number of questions.
10 · What to do
Five levers for an agency or a network
What the data of this edition tell agencies that want to be recommended by AI.
Clear, comparable information
Address, phone, published fees: the AI picks them up and compares agencies, especially in letting management.
An address in 50% of answers when the agency page is readMake each agent identifiable
For self-employed agent networks: one local page per agent, with their area, sales and reviews, to go from the brand to the person.
24% of IAD's citations name an agentPublish price benchmarks by neighbourhood
Price per m², time to sell, rents: the AI looks for figures and cites the ones it finds.
A price per m² in 22% of answersMeasure every week
The first recommended player changes from one week to the next: only a continuous measure, town by town, shows where to act.
75% of first recommendations change within a weekMethod and limits
How this barometer was built
- The corpus. Nearly 23,500 AI answers collected from April 1, 2026 to October 5, 2026. Nearly 13,500 are retained, on nearly 3,500 distinct questions and more than 250 towns; nearly 12,500 answer a local question.
- The questions. They are written like real requests from individuals and professionals: selling, valuation, buying, buy-to-let, renting, letting management, commercial real estate. Each question is asked again every week. Each one was classified one by one, with the help of an AI: main need, local or not, looking for a professional, a property or information, individual or professional. Mortgages, works, property surveys and holiday lets are excluded.
- No company in the questions. We excluded every question that names an agency, a network, a portal or a person (26% of search questions), as well as all brand comparison and brand perception questions.
- The AIs queried. ChatGPT through the OpenAI API (gpt-5-nano with web search, located in France). Gemini through the Google API (Gemini 3.1 Flash-Lite with Google Search grounding), since July 31, 2026. Calls are made without an account or history: answers are close to those of the consumer apps, without being identical.
- How a player is counted. A player is “cited” if it appears in the text of the answer or, for ChatGPT, in a cited source, and “first” if it is the first player named in the text. Networks, portals and platforms come from a dictionary of more than 160 brands in 11 families. Blind manual coding of 240 randomly drawn answers gives a precision of 98.5% and a recall of 99.4% on brands.
- Specific agency. A network agency is “named” when the answer points to a specific branch: agency name, town, address, phone or agency page. For a self-employed agent network, only an identifiable agent counts. Manual check: precision 93.8%, recall 95.1%. The share of answers that recommend a specific agency first is 60% with the automatic measure and 53% with manual coding of the same sample: we publish the range.
- Independent agencies. They are identified among the entities named in each answer (name and website), excluding institutions, media, directories and neighbouring trades (notaries, surveyors, mortgage brokers). This is the least precise measure: figures on independents are orders of magnitude.
- Towns and margins of error. Town size comes from official population figures (geo.api.gouv.fr). 95% confidence intervals are computed by resampling clustered by panel agency, since answers from the same territory are not independent. No result is published below 200 answers.
- Limits and panel bias. The questions are built from the towns and services of a panel of agencies, including agencies of the Orpi, Laforêt, Guy Hoquet, L'Adresse and Arthurimmo networks, and independent agencies. These networks are therefore cited more often: the ranking also publishes their presence outside the questions built from their own agencies. The results describe these territories, not the whole of France. No agency-level data is published. Wispra appears in 0.1% of ChatGPT answers and 4.4% of Gemini answers; it is excluded from the rankings.
- Next edition. January 2027.
Frequently asked questions
Does AI recommend estate agencies or portals?
Both, depending on the question. On a local question, the first recommendation is a specific agency, network or independent, in 53–62% of answers. When people look for a professional, that share rises to 78%; when they look for a property, portals come first in 69% of cases.
Does the network ranking depend on the choice of questions?
Partly. The questions are built from a panel of agencies, including agencies of the Orpi, Laforêt, Guy Hoquet, L'Adresse and Arthurimmo networks, which are cited more often there. That is why the ranking also publishes each network's presence outside the questions built from its own agencies: Orpi stays first (39.5%).
Can an independent estate agency be recommended by ChatGPT?
Yes: 53% of the agencies named by the AI are independent agencies, and 68% of local answers cite at least one. AI does not favour big brands; it cites the agencies it can read and locate.
Why does AI rarely cite the agents of self-employed agent networks?
When the AI cites IAD, it names a specific agent in only 24% of cases, against at least 90% for the top four agency networks. It knows self-employed agent networks by their brand, rarely by their people: one local page per agent is the most direct lever.
Do ChatGPT and Gemini give the same answers?
No. Gemini cites at least one independent agency in 94% of its answers, against 66% for ChatGPT, but almost never gives an address. And from one week to the next, the same question gives the same first player in only 25% of cases.
How can I find out whether AI recommends my agency?
Wispra measures every week where each agency stands in AI answers, town by town and against its local competitors, and helps improve it. Book a meeting to see the measure for your agency or network.
And your agency: does AI recommend it?
Wispra measures every week where each agency stands in ChatGPT and Gemini answers, town by town and against its competitors, and helps improve it.
Measure my agency