Not sure which package? Get your tailored plan — a 60-second fit-finder, no signup.
Bundles & prepay
Bundle the audit with a retainer, or prepay to save
Two simple ways to pay less for the same work — and a custom path if your situation doesn't fit a standard plan.
Launch Bundle
Full Audit + first 3 months of Growth — $5,500
Full Audit + first 3 months of Growth, at one fixed price.. The fastest way from "we don't know where we stand" to "it's measured and we're working it," at a fixed price.
Prepay a retainer & save
Prepay a term up front and pay less per month — here's exactly what each tier costs prepaid.
| Prepay term | $2,950/mo Growth | $6,500/mo Scale | $12,000/mo Authority |
|---|---|---|---|
| 3 months -10% | $7,965 save $885 | $17,550 save $1,950 | $32,400 save $3,600 |
| 6 months -15% | $15,045 save $2,655 | $33,150 save $5,850 | $61,200 save $10,800 |
| 12 months -20% | $28,320 save $7,080 | $62,400 save $15,600 | $115,200 save $28,800 |
Prices shown are the full prepaid total for the term (and what you save vs paying monthly). No lock-in beyond the term you choose. See retainers →
A bundle buys the same measured work: who AI recommends in a live category by share-of-model across all five engines, published with its method and date.
| # | Product | Share of model | Share | 95% CI |
|---|---|---|---|---|
| 1 | GitHub Actions | 21.8% | 20.1–23.7 | |
| 2 | GitLab CI/CD | 19.6% | 18.0–21.4 | |
| 3 | CircleCI | 19.1% | 17.5–20.9 | |
| 4 | Jenkins | 16.6% | 15.1–18.3 | |
| 5 | Argo CD | 6.4% | 5.4–7.5 | |
| 6 | Buildkite | 5.6% | 4.7–6.7 |
- Engines
- 5 of 5
- Prompts
- 10
- Answers
- 500
- Runs/engine
- 10
- Interval
- Wilson score interval, 95% (z=1.96)
Every bundle is measured this way for your category. Google's AI surfaces measured separately · 95% CI. Point-in-time; engines change. · Methodology →
Research evidence
We publish what we measured about whether each tier works, including when it didn't — see the full research programme.
| Study | Status | Finding |
|---|---|---|
| AI Answer Volatility Index | running | measured: 3 of 5 engines have a no-intervention band (anthropic +/-10.8pt, grok +/-7.2pt, openai +/-11.0pt); no valid null pair yet for gemini, perplexity |
| Cross-Engine Disagreement Index | proposed | not yet measured — proposed (pre-registration pending) |
| Newsletter/email presence vs AI citation | proposed | not yet measured — proposed (pre-registration pending) |
| Schema markup on NEVER-cited pages | proposed | not yet measured — proposed (pre-registration pending) |
| Content-freshness refresh test (causal, matched control) | proposed | not yet measured — proposed (pre-registration pending) |
| Developer-doc / technical-content citation study | proposed | not yet measured — proposed (pre-registration pending) |
| The negative-results ledger (what didn't work) | running | 2 pre-registered intervention(s): 0 moved, 0 no measurable change, 0 moved the wrong way, 2 unverified |
| AI citation -> funnel/revenue | proposed | not yet measured — proposed (pre-registration pending) |
| Own-brand Reddit test (disclosed participation only) | proposed | not yet measured — proposed (pre-registration pending) |
| Multi-baseline SCED across >=3 clients | proposed | not yet measured — proposed (pre-registration pending) |
| State of AI Visibility (the annual flagship) | proposed | not yet measured — proposed (pre-registration pending) |
Disconfirming evidence: at least one study above found this SKU's premise does not hold. See the study for the full method and limitations before you decide how we act on it.
Common questions
Is prepaying a lock-in?
What exactly is in the Launch Bundle?
My situation doesn't fit a standard plan — what then?
Need a custom package?
Multiple products, an unusual stack, an agency white-label, or a scope that mixes audits, retainers and add-ons in a way the standard plans don't cover? Tell us what you're trying to do and we'll put together a fixed quote — no call required.
Why Clear Cited
The five levers behind share of model
We get you recommended across the 5 AI engines — ChatGPT, Perplexity, Claude, Gemini, and Grok — plus Google's AI surfaces (AI Overviews & AI Mode), measured separately — and here is the honest system beneath it: what you are actually paying for, and the real limit on each.
Every engine and surface your buyers actually use
the AI engines your buyers use, plus Google's AI surfaces and Bing's Copilot answers — each measured separately, never summed into one score
The roster is five AI engines plus two answer surfaces — AI Overviews and Microsoft Copilot — reported separately and never summed into an engine count. The five engines are measured today. Both surfaces are positioned, not measured: an engine is one foundation-model family we query directly, a surface is a product that renders an answer, and a surface never becomes an engine by being added up. Outside the labelled separate-surface panel we name the Google surface plainly as AI Overviews, because naming it beside the engines would read as an engine claim.
A method you can reproduce
pre-registered prompt sets with published hashes, run counts and a confidence interval on every figure, and null results published on the same terms as positive ones - as standing practice, not a one-off study
Confidence intervals are reported honestly wide and named per product: presence uses a Wilson interval in both products; share-of-model uses a percentile bootstrap in the paid audit and a Wilson interval in the public Index, whose pooled denominator is clustered - so those Index intervals are narrower than a cluster-corrected estimate would give, and we say so. We never narrow a CI to look more certain than the data is.
All four tiers AI cites from
we work all four tiers AI cites from — not just the fraction of your presence you directly own
This is SCOPE — the tiers we work, not an outcome we promise. The per-tier percentages are illustrative (Yext-sourced) and the earned tier (news & forums) is best-effort, never guaranteed.
Almost none of your time
your time required is almost none — sign-off graduates to become yours by default once we've earned a track record, with continued sampling
A human approves every client deliverable until a clean track record is earned; graduation is opt-in, default-off, and reversible — client-facing and published content never reaches unattended auto (AI speed, human-approved).
A system that compounds
the system learns which of your pages get cited and compounds that signal every cycle
Honest-empty until measured — attribution shows the real citations a piece earns, with the engine, prompt and date, only once they are actually earned; no projected compounding curve.
The system compounds — see how it learns which of your pages get cited → · the four tiers of citations we work →
The transparency wedge
A method you can reproduce
pre-registered prompt sets with published hashes, run counts and a confidence interval on every figure, and null results published on the same terms as positive ones - as standing practice, not a one-off study
at least 10 runs per engine (12 by default), spent adaptively — more on high-variance engines, fewer on stable ones, never below a 5-run floor Every presence figure carries a Wilson 95% confidence interval. Share of model carries a percentile bootstrap 95% confidence interval in the paid audit, and a Wilson 95% confidence interval in the public AI Visibility Index — the method matched to the metric and named per product, both computed from the actual runs.
Confidence intervals are reported honestly wide and named per product: presence uses a Wilson interval in both products; share-of-model uses a percentile bootstrap in the paid audit and a Wilson interval in the public Index, whose pooled denominator is clustered - so those Index intervals are narrower than a cluster-corrected estimate would give, and we say so. We never narrow a CI to look more certain than the data is.
We publish our run counts, our uncertainty and our pre-registered prompt sets. Across the 8 providers we checked on 2026-08-02: run counts 4 of 8 verified; 4 not verified; uncertainty on published figures 2 of 8 verified; 6 not verified; the prompt set published with the results 0 of 8 verified; 8 not verified. The full comparison, with each vendor’s own wording →
That is the wedge: a number you can re-run and get back. See the full method →
Hold us to it, and hold the others to it too — the 12 questions to ask anyone selling AI visibility, each one answered here with a link.