AI Visibility shows where you are named and where you disappear. The Action plan uses that evidence to diagnose likely causes, rank the work worth doing and watch the relevant signal after the work begins.
Captured AI answers, category and engine results, named competitors, citations, provider-exposed sources and the technical scan connected to the site.
Private inventory feeds, pricing decisions, paid placements, internal work and changes made in external systems unless you provide that context.
Models are used for diagnosis and judgment. Measured claims are resolved and calculated again in code before they reach the board.
01
Laup builds a 30-day evidence packet from captured answers, category and engine results, prompt-level visibility, competitors, cited and provider-exposed sources, and the latest technical scan. Seven, 30 and 90-day views provide wider context.
02
Three configured models from different providers analyze the same packet independently. A failed or unparseable analysis is excluded; the run continues with the usable proposals that remain.
03
A separate judging pass reconciles overlapping proposals into no more than 15 Actions. It records qualitative judgments such as confidence, evidence strength, expected horizon and cross-model support.
04
Code resolves the proposed category, prompts, competitors and source domains against the measured packet. It then calculates the prompt counts, competitor gaps and source records shown with the Action.
05
New findings enter the board without erasing work already underway. Planned, in-progress, measuring, done and dismissed states survive later analyses; open findings can refresh or become stale.
This illustrative dossier uses the same kinds of fields Laup stores and shows on the Action plan.
85
Your weakest tracked category repeatedly returns competitors and third-party explainers. A focused guide can address the questions and source gaps visible in the captured answers.
Support is the judge's record of proposal agreement. It is useful context, not an independently measured result.
The board score orders Actions consistently: 85% comes from the evidence signals, and up to 15 points from an AI review that ranks the whole plan by expected payoff against effort. It does not estimate revenue, and scores compare the Actions in one plan with each other, not with other companies.
Medium confidence
66% of 45
29.7
Three-model support
100% of 30
30.0
Measured evidence
100% of 15
15.0
Quick-win horizon
100% of 10
10.0
Evidence score, weighted at 85%
85% of 84.7
72.0
AI review: ranked 3rd of 15
16 minus the rank
13.0
Calculated total
72.0 + 13.0, rounded to the board score
85
Laup preserves three evidence levels instead of forcing every recommendation to look measured.
The Action resolves to measured categories, prompts, competitors or source records. Its supporting numbers are calculated from those records in code.
The diagnosis follows from the evidence packet, but the proposed mechanism is an interpretation. Treat it as a reasoned hypothesis to test.
A useful action can remain on the board without a resolvable data anchor. It is kept deliberately and presented without measured support it does not have.
The latest evidence refreshes the recommendation layer while the decisions your team has already made remain intact.
1Recommended
A newly ranked Action, ready for review.
2Planned
Accepted and queued, including Actions created through chat.
3In progress
The underlying work is being carried out.
4Measuring
The work is live. Moving a card here starts its effect measurement.
5Done
Completed work. A card that reaches Done is measured too, and its history stays on the board.
Dismissed Actions remain dismissed. Open findings that disappear from a later analysis can be marked stale without rewriting triaged work.
Measurement does not start on its own. When a change is live, you move its card to Measuring or Done on your action board in Laup, or ask your connected AI assistant to do it. The move freezes a before-picture: your presence on the prompts the Action targets over the 14 days before. Cards left in Recommended, Planned or In progress are never measured.
1
Your team or connected agents carry out the underlying work in the systems where it belongs.
2
You do thisWhen the change is live, you move the card to Measuring or Done on your board in Laup, or let your connected assistant do it. This is the step that starts the measurement.
3
Every AI Visibility run after the move compares the same prompts with that before-picture. The card shows it as day N of 14 and the change in percentage points.
4
For a category-anchored Action, Laup compares the latest trailing 14 days with the previous 14-day window.
What the verdict means
It is an early signal from two adjacent 14-day windows. It shows whether category visibility moved; it does not prove the Action caused the move.
When no category is attached
Laup does not invent an AI visibility verdict. The Action instead directs the team to the relevant prompt, source or implementation checks.
An Action describes and tracks work. Your team carries it out, or a connected AI agent can act through the tools and permissions you give it.
Connected AI Agents
Agents can analyze Laup data, discuss and manage the Action plan, then use connected external tools to perform the underlying work. Changes they make to the plan carry the agent's name, and your team can reverse them.
See AI Agent accessLaup needs repeated observations before it can distinguish a pattern from one answer. The default is approximately five days and at least three completed AI Visibility runs. Timing can vary with the tracking schedule and available evidence.
Different models notice different gaps in the same evidence packet. Laup runs three configured analyses independently, then a judging pass reconciles the usable proposals. The Action shows the judge's support record, such as 3/3, as context rather than a measured fact.
Models can propose categories, prompts, competitors and source domains. Code resolves those anchors against the measured evidence packet and calculates the displayed counts and gaps. Laup also permits inferred and open Actions, but labels them instead of presenting them as measured.
The Action plan diagnoses, ranks and tracks the work. It does not publish content or alter external systems by itself. Your team can carry out the work, or connected AI agents can act through tools and permissions you provide.
After a category-anchored Action moves to Measuring and a later run completes, Laup compares the latest trailing 14 days with the previous 14 days. The result is an early visibility signal, not proof that the Action caused the change.
The Action plan begins with captured answers. See how those answers, mentions and sources become measured evidence.
Start measuring the categories you care about. Laup builds the first plan after enough repeated evidence has accumulated.