Tools · Phase 5 · in development
The Brand Visibility Monitor.
The audit is a photograph. The Monitor is the camera left running. Weekly automated probes across the same fourteen-model panel track your citation share, statement match, and 30-day persistence — and email you the week your visibility drifts, not the quarter after.
Not live yet — we're onboarding a small number of design partners while we harden the scheduled-probe backend and datastore. It's the recurring sibling of the one-shot LLM Visibility Audit. Request access below.
A snapshot can't catch a drift.
Twenty-seven years watching search taught one thing about visibility: it is never static, and it never warns you before it moves. Four reasons a quarterly audit leaves you exposed.
Visibility drifts between your audits
A one-shot audit is a photograph. Model weights get retrained, retrieval indices refresh, competitors publish — and your citation share can halve in the eleven weeks before your next quarterly run.
The drop is silent
Nothing pages you when a model stops citing you. There is no rank-tracker red line, no traffic dip in Analytics — the answer simply names someone else, and every query after routes elsewhere.
A quarter late is a quarter lost
Catch the slip in the next audit and you have already spent a full quarter feeding a competitor the queries you should own. The cost isn't the audit fee — it's the eleven weeks you didn't know.
Recovery needs the week, not the quarter
Know a claim slipped the week it slipped and you can trace it to the page or the rival that caused it. Ninety days later the trail is cold and so is the fix.
Four signals, tracked over time.
Every probe writes the same measurements the audit produces — but dated, versioned, and diffable, so the shape of the curve is the product, not a single number.
Citation share
How often your brand surfaces across the 14-model panel for your tracked query set, plotted week over week. This is the headline line you watch for drift.
Statement match
Which of your specific claims are reproduced verbatim, which are paraphrased without credit, and which get contradicted — per statement, per model.
30-day persistence
The decay curve for each claim you win. Commercial pages routinely fall below 12% citation retention by day 14; persistence tells you whether last month's wins actually survived.
Drift alert
The output you act on: an email the week a tracked metric crosses the threshold you set — with the model, the query, and the likely driver attached, not buried in a dashboard.
Probe, compare, alert.
The same controlled protocol as the one-shot audit, put on a schedule and wired to a threshold. No dashboard to check — the alert comes to you.
Probe
Your query set runs against all fourteen models, three repetitions each to dampen sampling noise — the same controlled protocol as the one-shot audit, on a schedule.
Compare
This week's citation share, statement match, and persistence are diffed against your rolling baseline and the prior run. Every historical run is retained.
Detect
A move past your threshold — a claim lost, a competitor surfacing where you used to, a persistence curve steepening — is flagged rather than left for you to find.
Alert
You get an email the week it happens, carrying the model, the query, and the probable driver — so you act while the trail is still warm.
The weekly digest, query by query.
Behind the one-line alert sits the detail: every tracked query, its citation share, and the week-over-week move. Held, gained, or lost — colour-coded so the losses read first.
| tracked_query | citation_share | Δ week | status |
|---|---|---|---|
| vector database pricing | 19% | −15pp | lost |
| best embedding model 2026 | 42% | +3pp | held |
| rag vs fine-tuning cost | 27% | −2pp | watch |
| self-hosted vector db | 51% | +6pp | gained |
| pgvector vs pinecone | 0% | −22pp | dropped |
Beta pricing, locked for partners.
These are design-partner rates, not a public checkout. You request access, we run a fit call, and nothing is charged until the Monitor is running on your domain. The rate you lock in now is the rate you keep.
Starter
- Weekly probe across the full 14-model panel
- Citation share + statement match, tracked over time
- 30-day persistence curve on every claim you win
- Email drift alert when a metric crosses your threshold
- Raw CSV export every run — no lock-in
Growth
- Everything in Starter, across three times the surface
- Competitor displacement — who fills the answer when you're absent
- Per-model breakdown, not just the panel average
- Weekly digest plus Slack or email alerts
- Query-set tuning as your category shifts
Scale
- Everything in Growth, at portfolio scale
- Twice-weekly probes — catch drift inside the week
- Custom query sets and segment-level tracking
- API access to every raw run
- Onboarding run personally by Gilad
Prices shown are the design-partner rates. The Monitor is in private beta; there is no self-serve purchase yet. Access is by request while we onboard the first cohort.
Become a design partner.
Design partners lock in founder pricing, get weekly access to the raw runs, and shape what the alerts flag. In exchange we ask for candid feedback — and a testimonial once the Monitor has earned it. Leave an email and, if you like, the domain you'd track.