Case study · Wealth management
First, win your own name.
More than a dozen firms share this independent wealth manager’s name inside the AI engines. Measured against the client’s own domain, the real story appeared: the first verified AI recommendations in the measurement record, wins up 300% from May to June, and a precise map of the identity problem to fix next.
+300%
verified wins, May to June
1 → 4, from a zero baseline in April
3 of 4
platforms citing the client's domain by June
Perplexity, Grok, ChatGPT
4 of 6
correct attributions converted to wins
when the engines resolve the right firm, it usually wins
Every number URL-verified against the client's canonical domain
The client’s brand is redacted due to compliance sensitivities in financial services. The measurements are reported exactly as taken. That caution is deliberate: Trinzik’s roots are in .gov and .edu web work, where compliance decides whether anything ships at all.
The identity problem
The engines answer the client’s name with thirteen different firms.
Other firms sharing the client’s name earned 99 URL-grounded first-place finishes across the 240 head-to-head query-answer opportunities, spread across 13 non-client domains. The client’s canonical domain earned 5.
A name-matched measurement would have credited all 99 of those finishes to the client and reported a market-leading result. URL-grounding is what keeps this report honest, and it is what exposed the real mandate.
The client must win its own name before it can win its market.
The issue is not awareness of the name; it is entity consolidation: making the client’s domain the firm the engines recognize and cite when they hear it.
What moved it
Content on the client’s own domain made that domain more authoritative in the engines’ reasoning.
The 20 benchmark questions never changed and neither did the competitor field. The variable we controlled was the content: 14 engineered articles on the firm’s core differentiators, each edited and approved by a human before it published, live between May 21 and June 23. The first verified win in the measurement record appeared 8 days after the first article went live, and the winning prompts match the article themes. At these counts that is a directional correlation, and we report it as such. It is also exactly the pattern the methodology predicts: raise the AI-perceived authority of the domain and the platforms’ own recommendations move behind it.
The trajectory
Zero verified recommendations at baseline. Four by June, across three platforms.
The strict standard: a win counts only when the platform ranks the client first and cites the client’s canonical domain. Nothing is inferred from names alone.
In head-to-head prompts every brand is named, so appearing is not the achievement. Being attributed correctly is. At baseline the platforms connected the name to the client’s domain in exactly 1 of 80 answers. The rest went to other same-named firms.
0% → 67%
attribution-to-win conversion
By June, when platforms correctly resolved the brand to the client’s domain, they usually ranked it first: 4 wins from 6 correct attributions. The immediate constraint is less preference than identity.
The second dimension
Discovery prompts measure whether a brand can be found. Head-to-head prompts measure whether a brand can be chosen.
This benchmark carries both: 20 head-to-head comparisons and 20 discovery prompts, every run. The discovery baseline is unambiguous. Across three runs and 240 discovery query-answers, the platforms never surfaced the client, by any name or domain. Organic recognition has not started yet, so every result here was earned in the head-to-head dimension, where the client is named in the prompt and the only question is whether the AI connects the name to the right firm and prefers it. As entity consolidation and the content library compound, the discovery series shows when that changes.
Platform detail
June was the first multi-platform pickup.
Perplexity
0 2
From never citing the client's domain to 3 attributions and 2 wins.
Grok
0 1
First to resolve the client, in May. It held the same family-ownership win in June.
ChatGPT
0 1
Resolves the client rarely, 1 of 20. Its June resolution converted to a win.
Gemini
0 0
Has never cited the client's domain. A strategically important platform, and the natural first beneficiary of entity-consolidation work.
Verified #1 recommendations, of 20 head-to-head prompts per platform
Ongoing authority building
14 engineered articles. The signal moved with them.
The client’s articles section was empty at the April baseline. The library was written for how AI reads, on the firm’s core differentiators: its fiduciary standard, its fee structure, its independence. The winning prompts match those themes. And every piece passed human curation and sign-off before it went live. Velocity comes from the system; judgment stays with people.
No content live
0 articles
measured · April 30
1 of 80 attributed · 0 wins
First articles live
4 articles
measured · May 29
5 of 80 attributed · 1 win
Full library live
14 articles
measured · June 29
6 of 80 attributed · 4 wins, 3 platforms
Each publishing phase preceded higher measured attribution and wins in the next benchmark. At these counts that is a directional correlation, and we report it as such.
Methodology & verification
Locked prompts
All trend claims use the April 30 prompt set, rerun without modification on May 29 and June 29. Byte-identical across runs, checksum-verified.
The strict verified-win standard
A win counts only when the platform ranks the client first AND cites the client's canonical domain. Answers that resolved the name to other same-named firms, 55 to 74 per run across 13 domains, were identified and excluded.
The AI picked the competitors, not us
The comparison set comes from the alternatives the AI platforms surface for this market, not from a list we or the client chose. The benchmark tests the field the AI already recommends.
Measured again next month
The benchmark reruns monthly on the same prompts, so a gain only counts if it holds. For this engagement the next runs also show when the entity-consolidation work starts winning back the client's name.
The benchmark demonstrates that sustained, structured content publication can shift AI recommendation preference in controlled head-to-head comparisons against the same LLM-surfaced competitors.
Does AI recommend the right you?
If your name is shared, conflated, or misattributed inside the AI engines, the benchmark will show it, verified against your own domain. A walkthrough runs it on your market.