See the work
Illustrative example — not client dataStart with the evidence.
Before deciding whether an engagement makes sense, see what our analysis actually looks like. Every figure below is illustrative, and every figure in a real report is traceable to a captured response.
Private AI visibility report · Jersey City, NJ · residential
Sample analysis — illustrativeTeam A is recommended in 29 of 64 answers. You're in 1.
When buyers and sellers ask an AI assistant who to hire, it answers with specific names. We asked it 16 real questions for this market, four times each. Every answer is preserved, word-for-word. Everything here comes from public records and captured answers — no estimates, no proprietary scores.
- 756×
- the answer named someone specific to hire
- 8×
- it was you
- 752×
- that was an individual team
- 760×
- that was a brokerage brand
Asked: “Who is the best real estate agent here for sellers?”
“Team A — high review count and strong local team activity on the major listing portals.”
— the assistant, recommending a competitor
Who shows up when buyers in this market ask
- 01
We wrote 64 real buyer and seller questions in this market.
- 02
We asked each one 4 separate times against ChatGPT (OpenAI). The answers change a little on every ask, so we count the pattern across all 256, never one lucky reply.
- 03
Every answer is saved word-for-word, exactly as it came back.
- 04
We counted who was named and who was recommended. The table is those counts — nothing estimated.
Brought up = named at all. Recommended = the answer said to use them. One answer can name several teams, so the columns don't add up to 100%.
| Team | Market rank | Brought up | Recommended |
|---|---|---|---|
| The team ← you | #8 | 8/256 | 4/256 |
| Team A | — | 120/256 | 116/256 |
| Team B | #11 | 64/256 | 52/256 |
| Team C | #2 | 56/256 | 48/256 |
| Team D | — | 48/256 | 40/256 |
| Team E | — | 36/256 | 28/256 |
Market rank is the city ranking by closed sales volume. Visibility doesn't follow it: the #2-ranked team was recommended 12 times, the #11-ranked team 13 — and the #8-ranked team once.
The rest went to brand-level names, not teams
- Brokerage brand 1124/256
- Brokerage brand 2112/256
- Brokerage brand 372/256
- Brokerage brand 432/256
No individual team owns the answers in this market yet — that space is still open.
Why AI is overlooking this team
The answers cited their sources 2,024 times across the 256 responses — never the team's own website. Presence on the surfaces below is how the answer changes.
- zillow.com696×
- realtor.com680×
- compass.com304×
- building websites40×
- homes.com40×
- the team's own site0×
What we found, in one paragraph
AI visibility is far weaker than this team's documented market position. A record sale and a top-eight city ranking sit in public records, yet across 256 captured answers the team was recommended in 2%. The record isn't the problem — its visibility is, and that part is workable.
The track record AI is missing
- Record condo sale in the city's tallest residential tower, covered by the local property publication.
- RealTrends America's Best: #8 team in the city by closed volume ($25.4M).
- 20 transaction sides — roughly $1.27M of closed volume per side.
The questions we asked — and who was named
“Best listing agent in this neighborhood”
Recommended: Team A, Team E, Brokerage brand 2
“Who is the best agent for sellers here?”
Recommended: Team A, Team D, Brokerage brand 1
“Which team should I hire for a high-rise condo?”
Recommended: Team A, Team B
“Who sold the record-priced condo in this market?”
Recommended: The team (you), Team C
Try one yourself in ChatGPT right now. Any single answer varies — that is why the report gives rates over 256 captured answers, not one reply. The pattern is the finding.
How this was measured
- Window
- Single capture day
- AI assistant
- ChatGPT (OpenAI)
- Questions
- 64
- Answers captured
- 256
Questions represent realistic buyer and seller queries for the market and are run repeatedly against the listed assistant. Responses are captured verbatim and read for which businesses were mentioned and which were recommended. Rates are the share of captured responses in which a business appeared. AI responses are probabilistic: individual answers vary, which is why sample sizes are shown and no single response is treated as a result. Rates describe assistant behavior in the monitored window — not revenue, lead flow, or market share. Answers are content-hashed at capture and never edited.
Explore the same findings tab by tab
AI Visibility Audit
Manhattan Luxury · Seller Agent Recommendations
AI Recommendation Rate
18%
18 appearances across 100 eligible observations
Recommendation frequency
Market position
3 of 8 measured teams
Based only on this sample analysis
Biggest observed gap
Third-party market evidence
Competing teams appeared across more independent market-specific sources in this sample.
Illustrative sample — not actual client data.
What the report contains
An intelligence dossier, not a brochure.
- 01
Executive summary
What we found, what it appears to mean, and what we would do first.
- 02
AI visibility
Recommendation share, prompt coverage and the diagnostic index.
- 03
Prompt performance
Question-by-question results with the captured responses behind them.
- 04
Competitors
Who is being recommended instead, and which signals support them.
- 05
Sources
The publications and profiles that keep appearing behind answers in this market.
- 06
Entity analysis
How consistently the team is described across the information environment.
- 07
Opportunities
Where the evidence suggests movement is realistically possible.
- 08
Prioritized actions
Sequenced by impact, evidence, confidence, effort and dependencies.
- 09
Methodology
Prompt universe, platforms, window and version used for these numbers.
- 10
Evidence log
Every observation, timestamped, with the raw response attached.