Nine commitments that define how Wiele uses AI, handles client data, and verifies the claims in our work. The trust layer is part of the platform, not a footnote.
Last reviewed · 2026-05-08
Every Wiele workflow that uses AI is named in the deliverable that ships. We don't hide or downplay AI's role in research, drafting, or analysis — and we don't let AI substitute for the strategic judgement clients pay for. If AI helped produce a slide, a report, or a citation, the methodology section says so.
No client deliverable ships without sign-off from a named Wiele principal. AI-drafted text is treated as a draft, not a deliverable. Engine output is reviewed against the source-level citation log before it leaves our hands. The principal listed on the SOW is the same principal whose name attaches to the work.
Lead-capture submissions live in Cloudflare KV in the EU edge. Transactional email is sent via Resend. Marketing analytics is Plausible — cookieless, no cross-site tracking, no fingerprinting. We do not sell, rent, share, or syndicate client data, ever. Client materials covered under NDA are segregated and never used in cross-client model training, ever.
Wiele runs on managed infrastructure with TLS in transit, encryption at rest, and access controls scoped to least privilege. Public-facing surfaces are gated by Cloudflare's security stack (WAF, bot protection, DDoS mitigation). We do not currently advertise SOC 2 or ISO 27001 certification — when we hold the audited posture, the badge will appear here with a verifiable link to the report.
Every claim, ranking, or competitive assertion in a Wiele deliverable traces back to a public, verifiable source. The Wiele OS engine logs source URLs, retrieval timestamps, and confidence scores per assertion. Methodology is open for inspection on every engagement; a citation log accompanies every report.
Outcome ranges, case-study metrics, and engine deltas in our materials are tied to engine runs we can show. We don't display aspirational ranges, unverifiable client metrics, or before/after numbers without the source data behind them. If a number appears on this site or in a deliverable, the underlying log is available on request to the client it applies to.
All client-facing content is editorially reviewed by a Wiele principal before it ships. Originality is non-negotiable — published work is not paraphrased from competitor pages, AI-rewritten from a single source, or syndicated from public-domain templates. Voice, accuracy, and authority belong to the client and the named author.
Mutual NDA is the default at the start of every engagement. Client materials, internal data rooms, and conversation context are walled off per client. Where a client agrees to be named in a case study, they review and approve the public framing in writing before publication. Where they don't, the engagement is anonymised and the methodology — not the brand — does the talking.
Automation is used to lift the volume of work a strategist can ship — never to replace the strategist. Where automation is in the loop, a named human is accountable for the output. Wiele will not deploy autonomous client-facing agents that act without human review, and will not stand up automation pipelines that bypass the editorial layer described above.
Methodology · AI Visibility Monitoring
Wiele's AI Visibility Monitoring retainer reports four core metrics: citation share, prompt coverage, source weight distribution, and named competitor citation share (Standard and Pro tiers). These metrics are measured directly from engine output. They are not modelled, projected, or interpolated. Every number in every monthly report is replayable from the citation log Wiele maintains for the client.
Every monthly run executes against ChatGPT, Gemini, Perplexity, and Claude — accessed via their public web applications, in fresh sessions, with no logged-in personalisation. The web app is what buyers see; the API can return different output. The web app run is what the retainer measures.
Each client has a versioned prompt panel — 25 prompts (Lite), 60 prompts (Standard), or 100 prompts (Pro). Prompts span branded queries, commercial-investigation queries, comparison queries, service-specific queries, local queries where applicable, and methodology queries. The panel is signed off by the client at kickoff and re-reviewed at each Quarterly Business Review. Mid-quarter changes are forbidden because they break month-over-month comparability.
Citation share is the percentage of panel prompts in which the client brand appears in the engine response, measured per engine, per month. It is reported per engine and as a four-engine average. It is not reported as a single combined “AI ranking” — there is no such number. The four engines have four citation functions, and collapsing them loses signal that drives action.
Not every citation carries equal weight. Wiele scores citations into four tiers — tier-1 (named press, peer-reviewed venues, recognised analyst firms, regulatory and government sources), tier-2 (credible secondary publications, named-author blogs, established forums), tier-3 (general web, anonymous content, content farms), and owned (the client's first-party content or the founder's owned surfaces). The taxonomy is maintained by founder and reviewed quarterly. Updates are logged with date and reason.
Citation share, prompt coverage, source weight distribution, branded search volume (pulled from Google Search Console), and featured snippet captures are measured lift — directly observable in engine output or in Search Console. Influenced pipeline (in Standard and Pro reporting where the client provides CRM access) is modelled attribution — opportunities the client's CRM tags as having an AI-surface or organic-search touchpoint in the buying journey. The distinction is named explicitly in every report. The two answer different questions.
Wiele does not guarantee specific citation share outcomes. Engine algorithms drift. Competitors ship work. New entrants enter the panel. The retainer commits to producing the cycle — engine runs, citation logs, action queues, monthly reports, and quarterly reviews — at the documented standard. Outcomes are measured against the prior baseline, reported honestly, and acted on through the next month's queue.
Wiele does not predict future citation share. The retainer reports observed lift. Forecasting is a Sovereign concierge service.
Wiele does not aggregate the four engines into a single “score.” Single-number rankings destroy the per-engine signal that drives action.
Every client has access to read the citation log that underpins their reports — on request. If a number in any report cannot be replayed from the log, the report is wrong and Wiele rebuilds it. This standard is not a value-add; it is the floor.
Wiele tracks observed shifts in citation behaviour across the four engines and publishes notable portfolio-wide shifts in client reports as engine algorithm field notes. Where Wiele believes a shift is engine-wide rather than client-specific, this distinction is named explicitly. Field notes inform panel construction at the next quarterly review.
The Lite, Standard, and Pro tiers differ in panel size, run cadence (Lite/Standard monthly; Pro bi-weekly), competitor tracking depth (Standard 3, Pro 5), featured snippet tracking (Pro only), modelled influenced pipeline (Standard light, Pro full), and action queue depth (3, 5, or 10 actions per month). Methodology rigour is identical across tiers. A Lite client receives the same standard of work; they receive less of it.
Have a specific question about how we handle data, AI use, or client confidentiality?
Book Strategy CallRun the audit
14 days. One engine run. Five signals. A 30-day implementation roadmap delivered with a Wiele principal on the call.
€2,500 one-off · No retainer required to start
