Original research
The 150-Query Real Estate AI Search Study
By Abdul Subkhan Published Fieldwork 8 to 12 September 2026
83% of the websites AI engines cite on real estate questions are cited by only one engine: across 150 questions and 747 measurements, 1077 of 1297 cited domains reached a single engine and just 3 reached all five.
Every engine answered the same 150 questions, and every engine built its answer from a different set of sources. All five disagreed on their single most-cited source, and the largest overlap between any two engines' top-30 lists was 10 sources out of 30.
- Full data, CSV
- 747 measurements across 5 engines
150 questions, 5 engines, 747 measurements
83%
of cited domains reach only one engine
- 1 engine 1077 (83%)
- 2 engines 162 (12.5%)
- 3 engines 40 (3.1%)
- 4 engines 15 (1.2%)
- 5 engines 3 (0.2%)
-
83%
of cited domains reached only one engine (1077 of 1297)
-
3
domains were cited by all 5 engines
-
5
engines, 5 different most-cited sources
Do AI citations transfer between engines?
Barely at all. We counted how many of the five engines cited each domain at least once across the 150 questions. The distribution is extremely top-light: almost every domain that gets cited anywhere is cited in exactly one engine's answers, and the number reaching all five is small enough to list by name.
Cited by all 5 engines
- goflydragon.com , 82 citations
- citevantage.com , 54 citations
- clearleaddigital.com , 20 citations
CiteVantage is our own site. We ran the study, so the bias is declared rather than hidden, and the published CSV lets anyone recompute this list without taking our word for it.
Every engine had a different favourite source
Given the same 150 questions, no two engines leaned on the same source most often. Google AI Overviews behaves unlike the rest, drawing heavily on large platforms rather than on the specialist sites the chat engines prefer, which is why its most-cited domain looks nothing like the others.
| Engine | Most-cited source | Distinct domains | Avg domains / answer |
|---|---|---|---|
| ChatGPT | goflydragon.com (39) | 305 | 4.1 |
| Perplexity | teramok.us (26) | 618 | 7.7 |
| Gemini | rankfender.com (5) | 50 | 1.3 |
| Google AI Overviews | youtube.com (55) | 470 | 6.3 |
| Microsoft Copilot | firstpagesage.com (22) | 153 | 2.7 |
How much do any two engines agree?
We compared each pair of engines on their 30 most-cited sources. The closest pair shared 10 of 30. The most divergent shared 3.
Who gets cited most across all five engines
Totalled across every engine and question, this is the leaderboard of domains AI answers were built from. The engine-count column matters more than the citation count: a high total earned inside one engine is a narrow position, not a durable one.
How we ran this real estate AI search study
We wrote 150 questions a real estate professional would realistically ask when looking for help with search and AI visibility, put every question to 5 AI engines, and recorded the domains cited in each returned answer. That is 747 measurements. We recorded which sources an answer was built from, not how the answer was worded.
- Questions
- 150, real estate search and AI visibility
- Engines
- ChatGPT, Perplexity, Gemini, Google AI Overviews, Copilot
- Measurements
- 747
- Recorded
- Domains cited in each answer
- One citation
- One domain cited in one answer
- Fieldwork
- 8 to 12 September 2026
- Capture
- Each engine's own interface; AI Overviews by search API
- Excluded
- Grok, only 19 of 150 queries completed
- Licence
- CC BY 4.0, reuse with attribution
What counts as one citation
Every citation count on this page is one domain cited in one answer, so a site that appears three times in a single answer counts once. It is the only unit recorded the same way on all five engines, because only Google AI Overviews was captured link by link. That gives 3,038 citations in total, and every count here reproduces exactly from the row-level CSV. Google AI Overviews often lists several links from one site: it cited YouTube in 55 answers through 91 separate links. An earlier version of this page mixed the two units and was corrected on 23 September 2026; the Google AI Overviews links are published as their own CSV.
How each engine was queried
- ChatGPT, Perplexity, Gemini, Copilot: each query was put to the engine in its own interface and the cited sources in the answer were recorded.
- Google AI Overviews: captured through a search API.
Limitations
- Gemini captured fewer sources. It returned a citable source list on a minority of its answers, so its distinct-domain count is a floor rather than a measurement.
- Results can differ by place and by account. Engines can tailor an answer to where and how it is asked, and at least one engine volunteered locally relevant providers unprompted. We did not test every variation; questions that name a city are largely insulated from this, while nationally scoped ones are not.
- One run, one week. AI answers are not deterministic and change between runs. This is a snapshot of 8 to 12 September 2026, not a stable ranking.
- We are in the data. We ran the study and our own domain appears in it. The row-level CSV is published so the finding does not depend on trusting us.
Press kit: cite, download and reuse this study
Everything here is free to republish under CC BY 4.0, including commercially, with attribution and a link back. The row-level data is published alongside the findings so any figure on this page can be recomputed rather than taken on trust. No permission request is needed and there is no embargo.
- CSV
Full dataset, CSV
All 747 measurements: query, question shape, engine and every domain cited. Reproduces every count on this page.
- PNG
Charts, PNG
Three 1200px charts, each with the source line and the licence inside the image.
Download reach, per engine or overlap.
- STUDY
Companion study
The 181-Brand AI Visibility Study: how often brands are named at all, across four engines
- AUDIT
Apply it to one firm
A redacted AI visibility audit for real estate: the same method run on a single firm, scored line by line
How to cite it
Reference
CiteVantage (2026). The 150-Query Real Estate AI Search Study. https://citevantage.com/research/ai-search-real-estate-2026/
In a sentence
A 747-measurement study by CiteVantage found 83% of the domains AI engines cite on real estate questions are cited by only one of five engines (CiteVantage, 2026).
For journalists
Summary, free to quote: CiteVantage put 150 real estate search questions to ChatGPT, Perplexity, Gemini, Google AI Overviews and Microsoft Copilot and recorded every website each engine cited. Across 747 measurements the engines cited 1297 distinct domains, of which 1077 (83%) were cited by a single engine and only 3 were cited by all five. No two engines shared a most-cited source. The practical implication is that AI search visibility does not transfer between engines, so a business measuring itself on one engine is measuring one fifth of the picture.
Author: Abdul Subkhan, founder of CiteVantage. Available for comment, data questions and methodology detail at abdul@citevantage.com.
Questions about this real estate AI search study
The questions below cover what was measured, why Grok is excluded, how the engines were queried, and how we handled the fact that our own domain appears in the results. If something is not answered here, the full dataset is downloadable above and we answer methodology questions directly.
What exactly was measured?
Which websites each AI engine cited when answering 150 real questions about hiring help for real estate search visibility.
We recorded the domains behind every answer rather than the wording, across 5 engines, for 747 measurements taken between 8 and 12 September 2026. One citation is one domain cited in one answer, the only unit recorded the same way on all five engines, which gives 3,038 citations in total. Every count on this page reproduces exactly from the row-level CSV.
Why is Grok not in the results?
Because only 19 of the 150 queries completed on it.
A rate built on 19 completed queries out of 150 would misstate the denominator, so Grok is excluded from every figure on this page rather than quietly averaged in. The five engines that remain, ChatGPT, Perplexity, Gemini, Google AI Overviews and Copilot, account for all 747 measurements. The exclusion is listed in the method table beside the engines that were counted, so nobody has to find it in a footnote.
CiteVantage appears in the data. Is this study self-serving?
We ran the study and appear in it, so the bias is declared and the data is public for anyone to check.
CiteVantage is one of the 3 domains cited by all five engines, and the page names every one of them, ours included. The finding that matters is structural and has nothing to do with us: 83% of cited domains reach exactly one engine. The full row-level CSV is published under CC BY 4.0, so anyone can recompute every number, including ours.
How were the engines queried?
Directly, one engine at a time.
All 150 questions ran between 8 and 12 September 2026, and the row-level CSV records which engine produced each answer. Engine by engine:
- ChatGPT, Perplexity, Gemini and Copilot were queried in each engine's own interface, and Google AI Overviews were captured through a search API.
- Grok was attempted and excluded, since only 19 of 150 queries completed.
Does being cited by one engine help on the others?
Barely: in this data, a citation on one engine rarely carried over to another.
Of 1297 domains cited across the study, 1077 reached only a single engine and just 3 reached all five. The largest overlap between any two engines' top-30 source lists was 10 of 30, and all five engines disagreed on their single most-cited source. A domain visible in the engine you happen to check can be invisible in the other four, so measure each one separately.
