The only platform that connects a deployed fix to a measurable lift in real AI-referred traffic. Every change is a logged event. The before and after run through statistical gates (an interrupted time series, a 95% confidence interval that excludes zero, a minimum volume floor). A causal line is drawn only when the evidence clears the bar. When it does not, you get an honest "not enough data yet" instead of an invented win.
Monitoring platforms show you that a number moved. They cannot tell you whether your work moved it. They have no record of what you deployed, and no baseline measured densely enough to compare against, so every movement is just as likely to be a model-load fluctuation as a result.
That is not a gap they can patch. A platform that only watches has no deploy event to define what was treated, and no dense daily series to build a comparison. Without both, there is no honest before and after.
TopSlot is an act-and-measure platform. It records the change, measures the response on a daily series, and applies the same statistical discipline a careful analyst would. That is the difference between a dashboard that reports movement and a platform that attributes it.
When you ship a fix or publish content, TopSlot records it as a dated event on your brand timeline. The event is the anchor. Without a recorded action, there is nothing to attribute a movement to. This is the part monitoring platforms structurally do not have.
TopSlot compares a 28-day baseline before the event against 7, 14, and 30-day windows after it, across AI mention rate, AI citation rate, and real AI-referred visits. The comparison runs on a dense daily series, not a single snapshot, so a movement has to persist to register.
The movement runs through statistical gates: an interrupted time series, a 95% confidence interval that excludes zero, and a minimum volume floor. If the evidence clears the bar, you get a verdict. If it does not, you get "not enough data yet" instead of an invented win.
A verdict (even a weak one) is only assigned when both the baseline side and the post-event side clear a minimum count of observed events. Below that floor, a rate moving from one-of-three to two-of-three queries reads as a large percentage but is one extra mention. The floor refuses to call that a result.
A strong or moderate verdict additionally requires the post-event level shift to be statistically significant: an interrupted-time-series fit with autocorrelation-corrected errors, and a 95% confidence interval that excludes zero. A large but noisy change is softened, never inflated. The gate only ever reduces a claim.
When evidence is below the floor, the result is marked insufficient. Insufficient rows carry no percentage and no causal phrasing, and they are excluded from any "wins" surface. This refusal to over-claim is the point, not a limitation. A confidently stated false claim on a proof surface is the worst outcome for a proof platform.
Every visibility-relevant event in your brand's life on one timeline: AI mentions, schema fixes, content publishes, rank changes, external industry events. You see what happened and when, so a movement always has context around it.
Each event is compared against a 28-day baseline across 7, 14, and 30-day windows. The result is a verdict (strong, moderate, weak, none, or insufficient) with a deterministic, plain-language explanation. No language model writes the conclusion, so the wording can never drift into a claim the math does not support.
For a deployed fix, one ordered receipt: this fix, on this page, moved these citations and earned these real AI-referred visitors. When a page URL is recorded, the receipt is page-scoped. When it is not, it honestly degrades to a brand-scoped receipt rather than imply a page that was not measured.
The sources AI cites in your category that you are not cited on yet, built from real audit citations and ranked by how often each source is cited and how easily you can get listed. When you get listed, the blueprint tracks the before-and-after and reports what was observed: visibility rose, or no change. It states what happened after the listing, never that the listing caused it.
Real human visits referred by AI engines (ChatGPT, Gemini, Claude, Perplexity), counted per page, captured through the Autopilot pixel and GA4. Actual traffic that arrived, not just where you rank. Crawlers do not run the pixel, so the count reflects people, not bots.
Attributes AI-referred visits and conversions above your baseline run-rate. It never invents a dollar figure. There is no fabricated monetary number; revenue is reported only when real value data is present, and the ledger shows "collecting data" until there are enough post-deploy days.
A daily categorized composite of your AI visibility direction over time, shown as a category (Strong, Moderate, Weak, Critical), never a vanity number you can game. Trajectory projects where the direction is heading, with bands that widen as the horizon extends.
Any figure shown in a sample receipt is illustrative.
AI search is a new channel, and leadership wants to know whether the work moved it. Attribution lets you report not just that a number changed but that your action is the credible reason, with a confidence interval and an honest verdict when the evidence is thin.
Show clients the closed loop: you shipped a fix, the AI citations on that page moved, and real AI-referred visitors arrived. A defensible process for arriving at an honest number beats a vanity dashboard that cannot tell whether the client's spend caused anything.
Rank tracking tells you position. Attribution tells you impact. Add a proof layer that connects a specific change to a measurable lift in real AI-referred traffic, and offer clients evidence their current reporting cannot produce.
If you optimize without proof, you cannot tell a real win from a model-load fluctuation. Attribution gives you a movement you can trust, gated by significance and a volume floor, so your next decision rests on evidence instead of a single lucky snapshot.
Attribution is the third pillar of a closed loop. The Search Tracker shows you where you stand across the AI models that matter. Autopilot ships the technical fixes. Visibility Trends proves whether that work moved your visibility and brought real AI traffic.
Each pillar makes the others honest. Without the deploy event from Autopilot, there is nothing to attribute. Without the daily series from the Tracker, there is no baseline to measure against. The loop is the moat.
Daily snapshots of your AI visibility across ChatGPT, Gemini, Claude, and Perplexity.
One pixel ships the technical fixes that make AI engines cite you, with auto-rollback.
Attribution and the Loop Receipt are included on Growth and Agency plans.
Visibility Trends is not included. Run a free Scorecard to see your current AI visibility category.
Event timeline and annotations with a 30-day lookback. Causal attribution, the Loop Receipt, and Trajectory are not included at this tier.
Full cause-effect attribution, the Loop Receipt, AI traffic analytics, Revenue Ledger, Trajectory, and Brand Health, with a 90-day window.
Everything in Growth, with a 12-month window, cross-brand comparison, real-time alerts, and the weekly executive-brief PDF.
A monitor shows you a number went up. It has no record of what you did, and no baseline to compare against, so it cannot tell you whether your work caused the movement. Attribution starts from a logged event (a fix or a content publish), measures a dense daily before-and-after, and only draws a causal line when the evidence clears a statistical bar. The honest verdict, including "not enough data yet," is the difference.
It draws a defensible causal line only when the evidence supports it. The engine runs an interrupted time series, requires a 95% confidence interval that excludes zero for a strong or moderate verdict, and enforces a minimum volume floor on both sides of the event. When those gates are not met, the result is marked insufficient and carries no percentage and no causal language. The system is designed to refuse rather than over-claim.
It is the closed loop on a single deployed fix: the change you shipped, the AI citations that moved on that page, and the real AI-referred visitors who landed there. When a page URL was recorded at deploy time, the receipt is page-scoped. When it was not, the receipt honestly reports at the brand level rather than imply a page that was never measured. Any sample receipt shown in marketing is illustrative.
It attributes visits and conversions above your baseline run-rate using absolute daily deltas, never a percentage that would over-state on low-volume days. There is no monetary column it can fabricate from. A dollar figure appears only when real value data is present, and the ledger reports "collecting data" until there are enough non-bot post-deploy days to stand behind a number.
Real human visits whose referrer is an AI surface (ChatGPT, Gemini, Claude, Perplexity), counted per page through the Autopilot pixel and GA4. Because crawlers do not execute the pixel, the count reflects people who arrived from an AI answer, not bots. This is the third leg of the proof loop: not just that your visibility moved, but that real AI traffic followed.
A verdict needs a full 28-day baseline plus the post-event window, so the earliest meaningful result lands roughly 36 days after a page starts being tracked. Until then, the Loop Receipt shows a "measuring" state. This is a structural property of honest before-and-after measurement, not a delay you can skip.
Run a free Scorecard to see your current AI visibility category, then close the loop with attribution on Growth and Agency. No credit card required.