How it works

Detection to verified fix.
One closed loop.

AI systems decide who gets discovered. Your site decides who converts. PlayHero monitors the second half — and this chart is the whole system.

Journey Score · acme.com/pricing · last 12 scans76 · +2 vs 90d avg
6065707580Drop detected−7 ptsRallycard approvedhuman decisionFix verified by scan+8 pts, card resolved
This site was up 99.97% of that window. Its demo-request conversion dropped 18%. Only one of those had a monitor — until scan five.
01The Score

One URL in. One number out.

A scan renders your page the way visitors and AI agents both see it, then scores seven pillars. The headline blends them — human experience weighted 0.88, agent visibility 0.12 — because people still sign the contracts.

Seven pillarsheadline = human×0.88 + agent×0.12
Speed
81
Accessibility
76
Hierarchy
58
Navigation
78
Readability
88
CTA
74
Agent Visibility
69
Journey Score74
Four passes, in parallel
DOM fetchDeterministic · ~3s

Full rendered DOM at 375 / 768 / 1440px — what visitors actually see.

Heuristic analysisDeterministic · ~14s

CTA visibility, contrast ratios, heading nesting, form friction, per viewport.

LighthouseObjective · ~15s

LCP, INP, CLS against Google's published thresholds. Scripts weighed by origin.

LLM interpretationProbabilistic · ~18s

Value-prop legibility and messaging hierarchy — catches what rules can't, labeled as interpretive.

Total scan time~50s
02The Agents

Four specialists. One domain each.

Every agent owns one narrow detection domain and rates its own severity. Narrow on purpose — a UX finding and a Web Vitals finding argued by the same generalist would weaken both.

UX Sentinel
Conversion friction · human UX signals

High and critical severity, plus lower-severity findings with strong supporting evidence.

CTA visibility across viewports
Heading hierarchy breaks
Mobile layout regressions
Drift Detector
Metric drift · engagement anomalies

Requires statistical significance before surfacing — a bad Tuesday is not a trend.

CVR trend deviation
Click-through anomalies
Deploy-correlated degradation
Page Vitals Enforcer
Load intelligence · AI crawl performance

Enforces Google's published Core Web Vitals thresholds — slow pages lose crawlers too.

LCP regression detection
CLS layout-shift scoring
Third-party script weight
Lumen
AI discoverability · citation monitoring

Scores AI readiness across the models your buyers actually ask, and flags misframing.

10-signal readiness scan
Citation Scout · 5 platforms
Framing accuracy check
03The Decision

Every finding arrives as a decision.

A Rallycard is a governed decision artifact, not a notification. Every field exists so a reviewer can decide without opening another tab.

UX SentinelHIGH
Hero CTA invisible on mobile viewports

Primary CTA renders below the fold on 74% of mobile screens. Visitors who never see the CTA cannot convert.

Decision impact (est.)
$88k – $192k / yr
Critical
Code fix
.hero-cta {
-  position: relative;
-  margin-top: 480px;
+  position: sticky;
+  top: calc(100dvh - 96px);
}
Agent + severity
Who found it, how bad, and by which detection method.
Decision impact
A benchmark-based dollar range, labeled as the estimate it is.
How it surfaces
High or critical severity, judged by the detecting agent — not a published percentage. Lumen's framing-accuracy check is the exception and keeps its own measured threshold.
Code fix
The exact diff, copy-ready. The reviewer sees the fix before they weigh the finding.
Approve · Defer · Dismiss
Nothing executes without this. Every choice is logged with actor and timestamp — dismissals included.
04The Fix

Approval starts a loop, not a ticket.

Approving a card doesn't file a ticket into the void. It starts a loop the system finishes on its own — and no one has to ask "did that work?" The record answers.

1
Approve
Card locks. The original finding is frozen — no quiet edits.
2
Brief drafted
A Resolution Brief with verification criteria, frozen at draft time.
3
Team ships
The fix lands in your codebase. PlayHero never touches your site.
4
Scan evaluates
The next scan reads the frozen criteria — not vibes, not a checkbox.
5
Resolved
The card closes itself, score delta attached.
RESOLVED
Hero CTA invisible on mobile viewports · verified by scan #48 · +8 pts
05The Record

The Heroboard remembers by page, not by ping.

The Heroboard is not a task manager. Every decision — approved, dismissed, deferred, resolved — lives under the page it belongs to, with actor and timestamp. Over time it becomes the only complete history of why your site converts the way it does.

acme.com/pricingJourney Score 76 · +8 last 30d
ResolvedHero CTA invisible on mobile viewportsverified by scan #48 · +8 pts · Jul 2
ApprovedCLP landing drops 19 daysS. Kim · brief drafted · Jun 28
acme.com/Journey Score 81 · steady
DismissedStale hero messagingM. Ortiz · reason: rebrand scheduled Q3 · Jun 25
PendingComparison table fails on 768pxDrift Detector · awaiting review

See your first Rallycard.

Free scan. No login, no credit card.

How PlayHero Works | PlayHero