The baseline is frozen at completion time
When you mark an action done, the current visibility score is written to that action as its before-value. It cannot drift afterwards, so the comparison is against the state that actually existed.
The operating loop
Set it once. Win recommendations every week.
Connect your site, then monitoring, scoring, change alerts and drafted fixes run on their own. Every number traces back to the exact AI answer it came from — so you always know what to do next.
- 60 checks a week
- Every Monday
- Runs itself
- Measured, not claimed
Short answer
What actually happens each week?
What happens in a normal week?
Nothing you have to do. Your ten buyer questions go out to six AI assistants, the answers are scored, any gap becomes a drafted fix, and anything you completed gets measured against where it started. You open your inbox and see what moved, what slipped, and what to ship next.
What is automatic, and what needs me?
Monitoring, scoring, change alerts, fix drafting and the digest all run without you. Your only job is the five-minute publishing step — or hand it to our engineers and do nothing at all. Nothing is ever posted or published under your name without you.
How soon can I trust the numbers?
Week one is a real reading and you can act on it. To talk about change you need two weeks, because a change needs something to compare against. Until then we say “waiting for next week” rather than pretending an improvement happened.
On autopilot
The routine, running while you don't
Read it as a line: each stage does one job automatically and hands the next exactly what it needs — ending in a fix you can ship and proof it worked.
01 · Setup
Onboarding · day one
Your market, mapped in one pass
Connect your site and competitors once. The system reads your positioning, pricing structure and feature differences — the same material buyers compare — so everything downstream is grounded in your real market, not generic keywords.
Output
Positioning claimsReal pricing structureFeature deltasMarket constraints02 · Questions
Onboarding · minutes later
The exact questions your buyers ask
Ten prompts across awareness, consideration and decision stages, written the way real buyers talk: team size, budget pressure, integration requirements, migration pain. No leading questions, no keyword strings.
Output
10 buyer promptsIntent-stage mappingCategory labels03 · Monitoring
Every Monday 00:05 UTC
Sixty fresh answers, every single week
Every tracked question goes to every AI surface — ten by six — automatically. Missed weeks don’t exist: the cycle opens on schedule whether you log in or not.
Output
60 checksWeekly scoreNever a missed week04 · Evidence
Weekly
Every answer kept as proof
Each response is stored verbatim before anything else happens, so you can always see the exact words behind a score. Refusals and empty answers are set aside rather than counted against you.
Output
Raw responseRefusal detectionEvidence trail05 · Grading
Seconds after collection
Graded on warmth, placement and proof
Every answer is graded on how strongly you were recommended, the sentiment of what was said, where you placed, and which sources decided it — including who won, why, your specific gap, and a verbatim quote. Anything misread can be corrected in two clicks.
Output
StanceSentimentPositionSourcesForensic block06 · Score
Same pass
One score that moves only when reality moves
Grades roll up into a 0–100 visibility score plus a simple citation share, stored per week for true like-with-like comparison. Corrections update the score immediately — the number always reflects the latest truth.
Output
Visibility scoreCitation shareStance strength07 · Alerts
On every answer
Warned before it costs you a deal
Five change signals ranked by severity — from an engine warning buyers away from you down to a five-point weekly slip. Each alert arrives with the evidence and a ready-made correction kit attached.
Output
Ranked incidentsCorrection kitDigest or webhook08 · Fixes
Your queue
The highest-payoff fix, ranked first
Five fix types ordered by time to value, each appearing only when your own results call for it. The first one takes five minutes. Nothing is invented to fill the queue — an empty list means you’re winning.
Output
Ranked actionsDraft assetsEffort badges09 · Proof
Next cycle
Proof it worked — or honesty that it didn’t
Completing a fix freezes your score as the baseline. Next cycle writes the measured after-value next to it. No second cycle yet? It reports as awaiting measurement rather than pretending — the discipline that makes every reported win trustworthy.
Output
Baseline lockMeasured liftBoard report
Proving it worked
Why a frozen starting score matters more than a confident claim
Anyone can show a chart going up. The harder discipline is refusing to take credit for a change you cannot actually attribute.
The after-value comes from a real cycle
A measured lift only appears once a later weekly snapshot exists. Until then the action reports as awaiting measurement — a deliberate refusal to round uncertainty up into a result.
Drift is compared between complete weeks only
The macro drop check uses the two most recent complete weeks. A partially collected current week can never trigger a false alarm, which is the failure mode that makes teams stop trusting alerts.
Corrections propagate instead of accumulating
If you override an extracted stance or flag a citation as incorrect, the affected weekly snapshot is recomputed in place. Historical comparisons stay consistent rather than carrying a footnote.
Example impact record
Before → after, with the cycle attached
- Objection-buster FAQ61.5 → 67.4+5.9
- llms.txt published63.2 → 66.8+3.6
- Entity disambiguation— → —Awaiting
Illustrative impact records. Your own deltas are computed from your weekly snapshots and appear only after a completed second cycle.
- W1
- First real reading
- W2
- Drift becomes visible
- W4
- First measured lift
- ∞
- Compounding cycles
What a bad week looks like
Progress is not a straight line
The chart shows a normal pattern: steady gains, then a five-point fall that raised a warning, then a recovery. A quarterly report would have averaged that fall into nothing, and you would never have known which week it happened.
- Every point is a complete weekly snapshot, not a rolling average.
- The alert threshold is five points, measured between complete weeks.
- Incidents carry the evidence, so a fall is investigated rather than guessed at.
- Recoveries are recorded too, which is what makes the series trustworthy.
Illustrative eight-week series.
Schedule
What happens without you lifting a finger
We are precise about this, because overclaiming automation is the fastest way to lose a technical buyer.
- Crawl your site and competitorsOnce, at onboarding
- Write ten grounded buyer promptsOnce, at onboarding
- Create the next cycle of checksWeekly, automatic
- Preserve every raw answerOn collection
- Extract stance, sentiment, position, sourcesOn every answer
- Recompute the weekly snapshotOn every write and correction
- Detect and rank drift incidentsOn every answer and weekly
- Draft ranked remediation assetsOn every detected gap
- Send the batched alert digestWeekly, with a cooldown
- Publish assets and seed citationsHuman step
Nothing in this list is presented as real-time. See how those boundaries are enforced.
Being straight with you
Where this stops
A technical buyer will find these limits anyway. Better they read them from us, on the page, than discover them later.
Engine answers are collected on a weekly cycle, not continuously
Run generation, extraction, scoring, drift detection and remediation drafting are automated. Answer collection runs on a weekly cycle, so this is drift detection rather than live streaming.
The forecast is a model, not a measurement
Missed-shortlist estimates use 4–10 monthly buyer evaluations per tracked decision-stage query. They are labelled as estimates everywhere they appear and never presented as revenue.
Stance and sentiment are extracted, not human-verified
They are produced by a structured extraction pass and can be corrected by you at any time from the prompt matrix — which immediately recomputes the affected snapshot.
Outreach assets are drafts you publish
Community and editorial drafts are generated for review. Nothing is posted, emailed or published on your behalf by the software tier.
Product walkthrough
Watch a full weekly cycle run
A recorded walkthrough that follows one week end to end: the rollover creating sixty checks, scoring landing in a snapshot, an incident firing, and the impact surface picking up the measured delta.
Recording in production. The player below activates the moment the Loom URL is set — no third-party script or iframe is loaded before that.
Walkthrough coming soon
Common questions
The practical details
Also published in a machine-readable format, so an AI answering the same question sees the same routine you do.
What happens during onboarding?
A bounded parallel crawl reads your homepage, pricing page, features or documentation and llms.txt, plus the same pages on each competitor, under a hard latency ceiling. That grounding produces ten realistic buyer prompts across awareness, consideration and decision stages, and opens sixty checks across six surfaces.
How long until the first useful reading?
The first snapshot exists as soon as the first cycle of answers has been extracted and scored, so you see real stance and citation data from week one. Trend, drift detection and measured lift need a second complete week, because a delta requires a prior point to compare against.
How is impact proven rather than asserted?
Marking an action done locks the current visibility score as its baseline. When the next weekly snapshot is computed, the same action receives a measured after-value, and the impact surface shows the before and after with the delta. Actions without a completed second cycle report as awaiting measurement rather than as a win.
What runs each week automatically?
An idempotent weekly rollover creates the next cycle of checks for every active brand, runs the macro drift check against the two most recent complete weeks, and dispatches the alert digest. Re-running it inserts nothing, so it is safe behind a scheduler or a retry.
What do the engines see when they read your own site?
The same standard we hold you to. This site publishes llms.txt, llms-full.txt and a capability manifest, allows retrieval crawlers explicitly in robots.txt, ships JSON-LD for Organization, WebSite, SoftwareApplication, FAQPage, HowTo and DefinedTermSet, and renders every page server-side with no client-side hydration.
Open your first cycle
Your competitors won't wait. Neither should you.
Connect your site and see what AI tells your buyers about you this week — every week you wait is a week of recommendations going to someone else. Fourteen days free, live in minutes.