Studies
Two baseline studies, publishing September 2026. Panels, screens, and questions are published in advance of collection. Findings are published in full.
How panels are built
Every AIWorthy panel is drawn from a public register by criteria published before any question is asked. Inclusion cannot be requested, declined, or purchased. No firm has paid to be included in a study, and no firm can pay to be left out of one.
Measurement is completed and frozen — with a cryptographic hash recorded — before any commercial relationship with a measured firm can exist. If a measured firm later purchases its own detailed results, that purchase is disclosed in the published study and does not affect what was measured.
We publish the panel so that anyone can re-derive it from the same register and the same screens. We do not publish which firm produced which result. Individual results go to the firm they concern.
We publish the roster. We publish the method. We do not publish which firm scored what.
What we measure, and on which surface
AIWorthy measures provider APIs, deliberately. API surfaces are controlled, repeatable and auditable — a frozen corpus can be re-read by anyone; a scraped session cannot. Consumer applications of the same platforms carry personalization, location and session state that API calls do not, and published guidance in this field indicates they may behave differently. That difference is a property of this method, stated here rather than left implicit.
A frozen corpus can be re-read. A scraped session cannot.
Results are reported as counts across repeated asks, never as rates, rankings or a composite score. Identical questions to the same platform return different answers — a documented property of how these systems are served, not a flaw in the measurement. Every question is asked five times on each platform so that variance is visible rather than hidden.
Automated matching produces candidates. Every count is read by hand before it becomes a number in a published study. Corrections are recorded, including corrections against our own earlier figures.
Triangle CPA Single Audit Baseline
Collected 11 August 2026 · Publishing September 2026
- Register
- Federal Audit Clearinghouse, AY2023–AY2025 general files
- Screen
- Named as auditor, at a Wake, Durham or Orange County address in the federal filing, for three or more distinct Triangle organizations
- Population
- Complete. Every firm meeting the screen is included — not a sample
- Excluded
- Five national firms, one state agency, and five out-of-region North Carolina firms, all named and reasoned in the panel sourcing document
- Scale
- 250 responses across five platforms · 1,443 citations
- Ground truth
- Federal Audit Clearinghouse filings; NC State Board of CPA Examiners register; 2 CFR 200.501(a)
- Status
- Collection complete, corpus frozen and hashed
The record publishes in September 2026 alongside the study — questions, fingerprints, and collection conditions.
Panel — 10 firms, listed alphabetically
| Firm | Auditor address in federal filing |
|---|---|
| Balance & Strategy Advisors, CPAs, LLP | Chapel Hill |
| Batchelor, Tillery & Roberts, LLP | Raleigh |
| Dean Dorton Allen Ford, PLLC | Raleigh |
| Langdon & Company LLP | Garner |
| Mauldin & Jenkins, LLC | Raleigh |
| McConnell & Jones LLP | Durham |
| Petway Mills & Pearson, PA | Zebulon |
| Romeo, Wiggins & Company, LLP | Raleigh |
| Sharpe Patel PLLC | Raleigh |
| Thomas, Judy & Tucker, P.A. | Raleigh |
Listed alphabetically. Order carries no meaning and no result is attributed to any firm on this page. Submission to the Federal Audit Clearinghouse is mandatory and names the auditor on every engagement — no membership, directory listing, or self-description was used. Three panel firms operate a Triangle audit office while headquartered elsewhere — Dean Dorton in Lexington, Mauldin & Jenkins in Atlanta, McConnell & Jones in Houston. Each is flagged in the panel file and each meets the screen on demonstrated Triangle work.
Triangle Registered Investment Adviser Baseline, Wave 2
Collected July 2026 · Publishing September 2026
- Register
- SEC Investment Adviser Public Disclosure (IAPD)
- Screen
- Retail-facing advisers with principal office in Wake, Durham, or Orange County, North Carolina
- Selection
- Random draw from a defined eligible pool of 138 firms, screens fixed before the draw and reproducible from the published seed and panel file. The draw was re-run six days later and produced a byte-identical result.
- Scale
- 2,625 responses across five platforms · 13,621 citations
- Ground truth
- Form ADV as filed with the SEC
- Status
- Collection complete, corpus frozen and hashed
- The record
- Publishing September 2026 alongside the study — questions, fingerprints, and collection conditions.
Panel — 15 firms, listed alphabetically
| Firm | CRD | City |
|---|---|---|
| Alohana Financial | 339433 | Durham |
| Beacon Financial Strategies | 132987 | Raleigh |
| Boundary Street Partners | 282300 | Chapel Hill |
| Broad Reach Capital | 144686 | Raleigh |
| Highwind Wealth Management | 335485 | Durham |
| Iseler Financial | 304647 | Chapel Hill |
| Onyx Wealth Management | 332972 | Raleigh |
| PieCapital | 328743 | Durham |
| Pragmatic Portfolios | 153764 | Durham |
| REH Wealth Advisors | 153826 | Raleigh |
| Reliance Financial Group | 118395 | Raleigh |
| Safe Harbor Family Capital | 317547 | Durham |
| Stone Legacy Capital | 334218 | Raleigh |
| The Wealth Group | 154106 | Durham |
| Wolfbridge Wealth | 148317 | Raleigh |
Listed alphabetically. Order carries no meaning and no result is attributed to any firm on this page. CRD numbers are included so any reader can verify each firm's registration and filings at adviserinfo.sec.gov.
What publishes in September
- Full findings for each study, ungated, with no email required
- The claims table — every published claim with its source, its sample size, and its verification status
- The question set and the screens used to build each panel
- A record of corrections made during verification, including figures that changed
Both studies are pilot waves. They established the instrument. The census criteria and benchmark instrument for the first full wave will be published before that wave runs.
Questions about the panels or the method: hharreld@aiworthy.ai