How we analyze
The intelligence loop: measure, attribute, learn, refresh, prove.
Measuring share of model is the start, not the finish. Clear Cited runs a loop around it — attributing the citations each piece earns, learning what wins, refreshing on cadence, and proving it back to you. This is how the system works. The measurement half lives on the methodology page; here is what we do with those numbers.
- 01Measure
Share of model across the five engines — with confidence intervals, and Google's AI surfaces measured separately.
- 02Attribute
Every published piece is matched to the AI citations it earns — engine, prompt, date.
- 03Learn
What earned citations feeds the next brief — performance rewrites what we make next.
- 04Refresh
Published work is updated on researched decay cadences — substantive updates, not date bumps.
- 05Prove
A client scorecard ties the content back to the citations, mentions and refresh wins it produced.
01 · Measure
Measure share of model — rigorously
The loop opens with measurement: your real buyer prompts across the five AI engines, reported as a median with metric-matched 95% confidence intervals, and Google's AI surfaces measured separately — never summed in. The full procedure (adaptive sampling, the CIs, the engine set) lives on the methodology page.
Speed
We deliver a measured baseline snapshot within 72 hours of kickoff, and reply to client requests rapidly and asynchronously.
The activation flow (M0–M6) fires a baseline_snapshot within 72h of engagement (Wave 197); inbound comms are human-gated for same-day approved replies.
02 · Attribute
Attribute every piece to the citations it earns
When a new AI citation appears, we trace it back to the piece that earned it and record the receipt — the engine, the prompt, and the date. Mentions and citations are tracked as distinct things; Google's AI surfaces stay on their own track.
Citation receipts
Every piece we publish is attributed to the AI citations it earns — with receipts: the engine, the prompt, and the date it was cited.
The content ledger logs each published piece; the attribution engine matches new AI citations back to the piece that earned them and records engine, prompt and date (Wave 200).
03 · Learn
Let performance rewrite the next brief
The pieces that earn citations teach us what to make next. Our learning loop reads which of our published pages got cited — and which structural signals travelled with them, like leading with a direct, citable answer — and folds that back into the next content brief. What the engines actually reward in your category is documented on how AI cites sources.
04 · Refresh
Refresh on researched cadences — freshness as a lever
AI answers skew toward recent content, so we don't publish and forget. The refresh engine tracks each piece against decay cadences and queues substantive updates — a real re-measure and rewrite, never a cosmetic date bump.
The freshness system
We refresh published content on researched decay cadences — a systematic, substantive-update schedule, not a publish-and-forget.
The refresh engine tracks each piece against decay cadences and queues substantive updates under a substantive-update contract (Wave 200), because AI citations skew toward fresh content.
(SE Ranking, via Parse, Nov 2025)
pages updated within the last three months (~13 weeks) average 67% more AI citations than older equivalents (6.0 vs 3.6 per page) — why systematic refresh is a lever, not a nicety
(Ahrefs freshness study, via Digital Applied, Mar 2026)
AI-cited pages are on average 25.7% fresher than the pages ranking in Google's organic top-10 (1,064 vs 1,432 days old, across 16.975M cited URLs) — AI answers skew toward recent content
05 · Prove
Prove it — on a client scorecard
Every retainer report carries a "what your content earned" scorecard: the citation receipts, the mention-vs-citation split, refresh wins, and (where measured) the conversion premium on AI-referred visits. Wins get flagged the same cycle; engine drift comes with before/after data. And nothing publishes without your sign-off.
Proactive rigor
We flag citation wins the same cycle they happen and advise on engine drift with before/after data — proactively, not on request.
Win alerts fire when a piece newly earns a citation (Wave 198); drift advisories surface engine-behavior changes with the before/after measurement attached (Wave 199).
Human + AI quality
Every piece runs through calibrated quality gates and lands in your approval queue before anything publishes — AI speed, human-approved.
The content-dispatch choke point routes each item through risk-tiered judge panels and a trust matrix (Wave 196); nothing publishes without client approval.
See the loop start on your brand.
A free teardown runs step one — measured share of model in your category.
Get a free teardownKeep reading