The evidence / Method
The questions, collection conditions and definitions used to interpret the results.
44 locked English-language questions cover 11 employer hiring situations. Each question was observed twice in ChatGPT, Gemini, Perplexity and Claude, giving 352 selected full answers: 320 UK and 32 US. Earlier extra pilots and access blocks are outside those totals.
UK observations were collected from London; the US domestic hire was collected from Ashburn. These are collection settings, not proof that every product used location identically.
| Tool | Recorded condition |
|---|---|
| ChatGPT | 88 logged-out Free guest observations. The underlying model is not identified when the UI did not disclose it. |
| Gemini | Two logged-out Flash-Lite observations for S01-A; 86 dedicated signed-in Free Auto / Temporary observations. |
| Perplexity | 88 dedicated signed-in Free Search / Incognito observations; automatic preview routing is retained where recorded. |
| Claude | 88 dedicated signed-in Free / Incognito observations. Visible Sonnet 5.5 Medium; repeat 2 records Manual. |
ChatGPT used logged-out guest access. The other tools also used dedicated signed-in Free accounts, as recorded above. This differs from the all-logged-out collection originally proposed; private or temporary chat does not itself mean logged out.
All 352 selected answers name at least one recruiter. The completed-answer and recommendation-bearing denominators therefore match in this collected set. Questions are related and purposively chosen; the count is not an independent random market sample.
JGA, JGA Recruitment, JGA Recruitment Group and James Gray Associates count where they identify the recruitment business. Brand and legal-entity references are distinct; collection from the US does not itself identify the US entity. Numerical agency rank is retained only where explicit; a numbered category is not automatically a ranking.
Across 176 exact question/tool pairs, JGA appears in both for 80, in only one for 48 and in neither for 48. Both answers remain visible. The repeats are not before-and-after measurements of a website change.
170 pairs have both submission timestamps: spacing is 28.79–59.63 hours, median 37.27. Six early pairs lack one or both exact submission times but retain dated capture. Eight known pairs exceed the proposed 48-hour spacing.
A domain counts once per answer within a source role. Hosts are grouped using the preserved Public Suffix List snapshot, with original URLs retained. Supporting citations, provider links and retrieved results have distinct views. Ties share a rank; alphabetical order resolves display order only.
Incomplete captures stay in the denominator, so observed counts can be lower bounds. The sensitivity view uses only complete/no-source captures for both numerator and denominator. Source frequency is not authority, ranking placement, market share or sales attribution.
The study includes 352 complete answer texts and 5,596 recorded source events. Source capture is complete for 345 answers and partial for seven, so source frequencies are observed minimums where capture is partial. Q44 Perplexity retains all 21 supporting events; its separate retrieved list is not captured. Dated text and HTML exports cover every answer. Two original ChatGPT screenshots, Q31 and Q41, are unavailable; their complete answer texts are included.
The company-page reviews are separate from model-cited sources. Public service claims, job adverts and publisher case studies are attributed; they do not independently verify all completed placements. Observed citation use does not reveal hidden model reasoning.
Payroll recruitment page describes UK/global talent acquisition and publishes combined-experience, reviews and retention claims. These are self-reported; the retention cohort, period and calculation are not defined in the inspected text.
JGA · checked 2026-10-22 · separate public research
Homepage presents payroll and HR recruitment, UK/EMEA coverage, podcasts and an employer vacancy route. Public brand and service description, not independent evidence of successful placements.
JGA · checked 2026-10-22 · separate public research
London role page lists operational payroll and implementation roles and links employers to the payroll recruitment page. It carries 2019/2020 award references and a practice-lead experience statement; date-specific claims should remain dated.
JGA · checked 2026-10-22 · separate public research
A public payroll recruitment page uses a test-labelled URL/title. An unauthenticated technical fetch returned 200, a self-referencing canonical and index/follow metadata. Whether it should be retained or consolidated requires content-owner review, not an automatic redirect.
JGA · checked 2026-10-22 · separate public research
US website names JGA Recruitment, Inc., a Delaware address and US contact route, and describes global payroll/HR technology recruitment. This does not establish coverage or placements for every domestic US Payroll Director mandate.
JGA US · checked 2026-10-22 · separate public research
US payroll service foregrounds global payroll talent acquisition. It repeats broad experience, review and retention assertions. Public copy does not itself document a domestic multi-state director search or the methodology behind the retention claim.
JGA US · checked 2026-10-22 · separate public research
Employer-facing page separates permanent, temporary/contract, executive search and client solutions. It describes interviewing, role briefing, vetting and knowledge evaluation for temporary candidates. Service statements and largest-specialist positioning are publisher claims, not independent comparative performance.
Portfolio Payroll · checked 2026-10-22 · separate public research
Payroll practice page specifies operational-to-global roles, international coverage and different hiring models. It connects expertise, team members, resources and employer enquiry. Network and client-retention assertions remain self-reported.
Frazer Jones · checked 2026-10-22 · separate public research
Payroll-specific enquiry page distinguishes executive, permanent and interim requirements and identifies UK payroll practice contacts. The page makes the buyer action explicit; it does not verify candidate availability or a promised time to hire.
Frazer Jones · checked 2026-10-22 · separate public research
Public profile identifies John O’Brien as Head of Payroll and describes UK/EMEA/global executive payroll and benefits recruitment. An attributed candidate testimonial offers context but does not independently audit the whole practice.
Frazer Jones · checked 2026-10-22 · separate public research
Capabilities page spans operational payroll through global leadership and permanent/interim assignments. It supplies sector-and-role placement examples and practice statistics. These are self-published examples without an independently verified outcome dataset or dates for every placement.
Oakleaf Partnership · checked 2026-10-22 · separate public research
Payroll page explicitly addresses transformation, systems, governance and vendor change across UK/EMEA/US mandates, spanning analysts to senior leaders and interim/permanent hiring. It also describes post-search reporting. This contradicts a blanket interpretation that its public scope is senior-only.
Middlemore · checked 2026-10-22 · separate public research
Employer finance page includes payroll, permanent/interim/contract hiring and systems/transformation roles within a broad finance proposition. It supports public scope, not evidence of success on a specific Global Payroll Director search.
Hays · checked 2026-10-22 · separate public research
Payroll page lists operational-to-head roles and UK regional recruitment coverage. It is primarily candidate-facing. Job categories and current advertisements show market activity; they do not independently prove completed employer placements.
Hays · checked 2026-10-22 · separate public research
Employer finance page includes Payroll Manager, different hiring models, technical assessment options and a described recruitment process. Broad finance scale and client figures are self-reported and are not a payroll-director-specific success rate.
Robert Half · checked 2026-10-22 · separate public research
A public US results page includes permanent director-level adverts describing multi-state payroll, team leadership, systems and controls. This is role-specific recruitment activity, not confirmation that advertised jobs were successfully filled or remain open.
Robert Half · checked 2026-10-22 · separate public research
A native browser check saved with Q43 matched the publisher’s statements about competency assessments, 12 hires and 85% retention after 12 months. These are publisher-reported case outcomes, not independently audited results or evidence of a US director placement.
Robert Half · checked 2026-10-22 · separate public research
Google states that its Search AI features use established search eligibility and SEO practices; no special AI file or schema is required. This guidance concerns Google Search AI features, not a guarantee or a complete specification for the four consumer products tested.
Google Search Central · checked 2026-10-22 · separate public research
Structured data should describe the visible page accurately and be validated. Eligibility does not guarantee a rich result or AI recommendation. Recommendations in this report do not depend on adding unsupported ratings, affiliations or special AI markup.
Google Search Central · checked 2026-10-22 · separate public research