The methodology disclosure for editorial transparency. Where each category of claim on syxoai.com comes from — observational data from Syxo's own voice builds, public regulatory documents, industry surveys, and hedged professional judgement informed by 17 years in marketing.
The Syxo blog and guide cluster makes many specific claims: 70-85 percent voice match on first draft, 30+ voice builds shipped, 60-200x cost ratios between AI-system and human ghostwriter paths, specific calibration timelines per ICP, percentage variance from posting time versus content quality on LinkedIn. Readers and search engines deserve to know where each category of claim comes from.
This page documents the sources. Four categories cover most claims on the site: observational data from Syxo's own builds, public regulatory and standards documents, third-party industry surveys, and professional judgement from 17 years in marketing. Each category has different reliability characteristics; the writing on this site tries to match the language to the source.
Source Category 1
The single most-cited source on this site is "30+ voice builds shipped" or variations. This refers to documented voice system builds completed for clients (and for Syxo's own use) since early 2026. The builds span solopreneurs, B2B founders, course creators, personal brand operators, and regulated practitioners (solo lawyers under SRA, accountants under ICAEW/ACCA, financial advisers under FCA, therapists under BACP/UKCP).
What gets observed: voice match score on first draft (using the 12-point audit), calibration time to stable voice match, failure modes during build, audience response patterns to voice-matched versus generic content for the same operator, edit time per piece pre-build versus post-build, ICP-specific calibration adjustments.
What doesn't get observed: long-term revenue outcomes for clients (Syxo does not have access to client financial data), retention or churn rates for clients' own customers, third-party platform-side metrics that require platform access.
How this source is cited on the site: "across 30+ voice builds shipped", "in our observation", "typical first-draft voice match", "calibration timeline for X ICP". The language hedges where the observation is from a smaller sub-sample (e.g. "regulated practitioner builds we have observed" when the sub-sample is closer to 5-7 builds).
Source Category 2
Specific regulatory claims are sourced from the published regulator's own documents. The pages on regulated practitioners (solicitors, accountants, financial advisers, therapists, real estate agents) cite specific frameworks. Where the citation appears in body content, the source is the regulator's published guidance.
UK sources:
US sources:
Where the site makes a claim about regulatory requirements, the source is one of these documents or the regulator's published guidance. Specific clauses are not always cited in the body content for readability; readers can verify by following the regulator link.
Source Category 3
Some claims rely on third-party industry data: LinkedIn algorithm behaviour, completion rates for online courses, ghostwriter pricing benchmarks, AI tool adoption rates.
Where these claims appear, the underlying sources include:
Where third-party data is cited, the writing tries to use ranged estimates ("5-15 percent typical", "varies by source, typically £X-Y") rather than point estimates that imply more precision than the source supports. If a specific point estimate appears, it usually comes from one of these sources and can be verified.
Source Category 4
Some claims on the site are professional judgement rather than observation or external data. Examples: "the 70/30 ratio (AI execution / human direction) covers most business content marketing", "buyer time required rises with budget rather than falls", "skipping voice infrastructure is the most common failure mode".
These claims are sourced from: seventeen years of marketing experience including senior in-house roles in MedTech, daily observation of how marketing teams and solopreneurs actually operate, conversations with hundreds of marketing operators across roles and industries, ongoing engagement with the marketing and AI tools communities.
How these claims are signalled in the writing: hedged language ("typically", "most operators we work with", "in our experience", "this pattern recurs"). The hedging matches the source — professional judgement is informed observation but is not the same standard as documented data. Where the language sounds more confident than that ("the right answer is X"), that's almost always because the same claim is also supported by observational data from Syxo's own builds.
A reference table for the most-repeated claims across the cluster:
| Claim category | Typical phrasing | Source |
|---|---|---|
| Voice match scores | "70-85 percent voice match on first draft" | Syxo build observation (Source 1). Measured using the 12-point audit across builds. |
| Build count | "30+ voice builds shipped" | Syxo build records since early 2026. Specific named clients: Sol Carey, Jacob Olenick, Anis Soomaroo, Rose Wang, Zac Murphy. |
| Pricing benchmarks | "£3,000-7,500 per month mid-tier ghostwriter" | Composite: Upwork/freelance marketplace data (Source 3) + Syxo conversations with practitioner ghostwriters (Source 1) + professional judgement on UK market (Source 4). |
| Sector regulatory rules | "SRA restricts outcome guarantees" | SRA Standards and Regulations directly (Source 2). Same pattern for FCA, ICAEW, BACP, etc. |
| LinkedIn behaviour | "posts that earn engagement in first 30-60 min get distributed" | LinkedIn published engineering blog posts (Source 3) plus observation across Syxo builds (Source 1). |
| Course completion rates | "5-15 percent self-paced, 30-50 percent cohort" | MOOC research and online course meta-analyses (Source 3). Specific sources include Class Central reports and university-published research. |
| Edit time per post | "15-25 minutes per post with voice infrastructure" | Syxo build observation across solopreneur clients (Source 1). |
| Timing variance on LinkedIn | "10-15 percent engagement variance from timing" | Syxo build observation (Source 1) compared against aggregate timing data from Sprout Social and Buffer (Source 3). |
| Cost-quality ratios | "60-200x cost ratio between AI-system and human ghostwriter" | Calculated from £713-1,453 year-1 AI path versus £24-90k+ year-1 ghostwriter retainers. Underlying prices from Source 1 (Syxo) and Source 3 (market surveys). |
| Voice prompt length | "500-800 words" | Syxo methodology specification (Source 1). The 500-800 word range emerged from build observation: below 400 produces voice prompt drift, above 900 produces context dilution. |
Honest disclosure of where the source data is thinnest:
1. Sub-sample sizes for niche ICPs. The "30+ voice builds" total includes builds across many ICPs. For specific regulated niches (solo lawyers, financial advisers, therapists) the sub-sample is closer to 3-7 builds per niche. Claims specific to those niches should be read with that sample size in mind. Where a claim is from a smaller sub-sample, the language hedges ("regulated practitioner builds we have observed" rather than "across 30+ builds").
2. Long-term outcome data. Syxo started shipping in early 2026. Long-term outcomes (year-2, year-3 client retention, audience compounding, pipeline contribution over multiple years) are not yet observable. Claims about year-2+ behaviour are extrapolation from year-1 patterns plus professional judgement, not direct observation.
3. US-market dynamics. Syxo is UK-based with majority of builds for UK clients. US-specific claims (state real estate rules, SEC nuances, US ghostwriter market) draw on Source 3 (industry surveys) and Source 4 (professional judgement) more heavily than Source 1 (direct observation). US readers should adjust accordingly.
4. Counterfactual claims. Many claims compare outcomes ("voice infrastructure produces X versus generic AI which produces Y"). The "Y" side is harder to measure directly because Syxo's clients did the voice infrastructure build — what they would have experienced without it is partly counterfactual. Where counterfactual claims appear, the comparison draws on Source 3 (industry surveys of operators using default tools) and Source 4 (professional judgement) rather than direct Syxo observation.
For full editorial transparency: the writing on syxoai.com is produced through the same voice infrastructure methodology Syxo sells. Specifically:
This is intended as the most honest possible demonstration of the methodology: the site that sells voice infrastructure runs on voice infrastructure. If the writing reads as the author's voice rather than generic AI, that is the demonstration the methodology works.
If a specific claim on the site reads as too confident without sourcing, or if a number does not hold up against published data, please email hello@syxoai.com. The site is iterated continuously; corrections appear in updated drafts within 1-2 weeks of identified errors. The dateModified field in each page's schema tracks when the page was last revised.
This methodology page itself is iterated as Syxo's build sample size grows, as new regulatory sources emerge, and as professional judgement claims either are confirmed by accumulating data or revised against contrary observation.