Changelog

PlayHero^° (formerly PlayGTM) is the public brand. WebUX is how we talk about scores, pillars, and monitoring — not a separate company. Existing product URLs under /webux stay the same.

This release extends Lumen from structural readiness — llms.txt, crawlers, schema — into live monitoring of how AI assistants talk about your brand, whether their descriptions match yours, and how you rank across the assistants your buyers actually use.

Citation Scout — five platforms

Citation Scout now queries ChatGPT, Perplexity, Gemini, Grok, and Claude with buyer-intent prompts to track whether AI recommends your product. Results are stored per prompt and per platform, with response excerpts and proof screenshots. Share of Model — your mention rate across AI platforms — is tracked on your monitoring cadence and surfaced on the Lumen history page and in the Monitoring drawer.

Framing accuracy alerts

A new Lumen Rallycard type fires when AI describes your product materially differently than your page does. The check compares Citation Scout response excerpts against your headline and value props using a secondary LLM pass. Cards surface at 70% confidence or higher.

Ten AI readiness signals

Lumen now checks ten signals on every scan, up from five. Four new checks follow Google's published AEO guidance:

• ■Content freshness — pages not updated in 12+ months lose citation share. Lumen checks the Last-Modified header and flags stale pages.
• ■Snippet access controls — nosnippet and max-snippet=0 block AI from excerpting your content. Lumen detects both and surfaces a high-severity Rallycard.
• ■Schema validation — Lumen now checks that your structured data matches your visible page content, not just that it exists.
• ■Multimodal readiness — product images without alt text are invisible to AI multimodal search. Lumen counts them and generates a Rallycard.

Share of Model in the UI

The Lumen history page now shows Share of Model across all five platforms — mention rate, citation rate, and per-prompt results. Scan history rows show per-platform citation status inline. The Monitoring drawer adds a visibility timeline and prompt-level breakdown with trend signals.

Monitoring screen

The monitoring empty state now surfaces Lumen AI Discoverability as a distinct capability alongside WebUX monitoring — its own card, its own CTA, and plan-aware gating — so you can see what activates before you add URLs.

Agency view

Each client workspace card in the All Clients view now shows that client's Lumen grade, llms.txt status, and Share of Model percentage — a single-screen read on which clients need AEO attention without opening every workspace.

Schema recommendations that understand what kind of site you are

The Agent Visibility pillar previously recommended SoftwareApplication schema to every site it scanned. A SaaS product and a professional services firm received the same recommendation. That changes today. The engine now classifies site type before generating schema findings, reading GTM motion, CTA behavior, positioning themes, and audience signals from the scan result. Software products continue to receive SoftwareApplication recommendations. Service businesses — agencies, consultancies, professional firms — now receive ProfessionalService schema recommendations instead. FAQPage is conditional on detected FAQ content regardless of site type. Getting this wrong undermines trust in the cards that are correct. Getting it right makes Agent Visibility actionable for a much wider range of sites.

Agent attribution is now accurate on every Rallycard

Scan results were attributing every finding to UX Sentinel — including findings that Page Vitals Enforcer and Lumen detected. The root cause was a hardcoded agent value being set before the pillar-based fallback logic could run. The fix resolves attribution from the finding's pillar directly: load and speed findings surface under Page Vitals Enforcer, agent visibility findings surface under Lumen, everything else under UX Sentinel. The three-agent model now reads correctly across the full card list on every scan result.

Lumen panel: detected vs. not detected

The Lumen widget previously showed the same "No llms.txt detected" state regardless of whether the site actually had an llms.txt file. Sites with a strong llms.txt were being offered a Generate button for something they'd already done. The panel now reads llmsTxtQuality from the Agent Visibility pillar signals directly. When a valid llms.txt is detected, the panel shows a confirmed state and a link to enroll in Lumen monitoring instead of the Generate CTA. The Generate button appears only when llms.txt is genuinely absent.

Pillar status labels respect benchmark delta

A pillar scoring 75 with a positive delta — meaning the site is above benchmark — was displaying "Needs work" because the label logic only looked at absolute score. Labels now account for both score and delta. A site above benchmark threshold reads "On track" even if the absolute score hasn't cleared 80. Critical risk still fires on significant negative delta regardless of score level.

Agency workspace tool navigation preserves client context

Navigating to Compare or SuperScan° from inside a client workspace was dropping the workspace context, resetting the sidebar to the all-clients view. All tool navigation now carries the active workspace forward. Switching tools no longer loses your place in the client hierarchy.

Yesterday we calibrated the scoring engine. Today we made what the engine surfaces actually usable — connecting Agent Visibility findings to the Heroboard, fixing how agents are attributed across the product, locking the pillar display order, and adding buyer journey context to every pillar in the dashboard.

Agent Visibility is now a first-class Heroboard pillar

Previously, Agent Visibility signals were computed on every scan but never became Rallycards. The score existed. The gap was visible. But there was nothing to approve, assign, or track.

That changes with today's update.

Every authenticated WebUX scan now generates Agent Visibility Rallycards automatically when gaps are detected:

• ■No llms.txt detected — HIGH priority. AI crawlers have no discovery file to parse and must reconstruct your product from raw HTML, reducing citation accuracy and confidence.
• ■AI crawlers blocked in robots.txt — HIGH priority. Your robots.txt blocks GPTBot, ClaudeBot, or PerplexityBot — the bots powering the AI assistants your buyers use to research vendors.
• ■og:image missing — LOW priority. Reduces richness of AI-generated citations when visual context is included.

These cards appear in the Rallycard list alongside Load Intelligence and Script Intelligence findings — sortable by impact, approvable to the Heroboard, with the Approve button wiring the finding into your governance workflow.

The llms.txt generator in the Lumen widget remains a one-click action for the most common gap.

Lumen is now a monitoring agent, not a separate tool

Lumen scans now appear in your authenticated Lumen history regardless of whether they were triggered from the public `/lumen` page or from a full WebUX scan. Every scan has Agent Visibility data — the Lumen history page now surfaces all of it in one place, tagged by source.

Public Lumen scans no longer generate Rallycards on the Heroboard. That was creating duplicate findings — the same load speed issue appearing twice from two different scan paths. Heroboard findings now come exclusively from authenticated WebUX scans.

Agent attribution is correct throughout the product

A long-standing inconsistency: Script Intelligence findings were attributed to a fifth agent badge that didn't match the four agents in the Agent Status widget. Script Intelligence is a scan sub-capability — its findings belong to UX Sentinel.

All Script Intelligence findings now display the UX Sentinel badge. The underlying data is preserved — the Script Intelligence tab, the detection method label, and the database values are unchanged. Only the display attribution is corrected.

Agent colors are now locked to canonical values shared across the dashboard, scan results, and Heroboard. One source of truth. No drift.

UX Sentinel#F43F5EPage Vitals Enforcer#00FF99Lumen#06b6d4Drift Detector#f59e0b

Pillar display order follows the buyer journey

Pillars now render in a fixed sequence across the dashboard breakdown, scan results chips, and pillar score strips — in the order a buyer experiences your page:

1. Load Intelligence — Can users even get in?

2. Signal Hierarchy — First visual scan

3. Messaging Clarity — Do users understand what you do?

4. Content Clarity — Does the copy resonate?

5. CTA Clarity — Can users take action?

6. Conversion Friction — Is the path frictionless?

7. Agent Visibility — Can AI find and cite you?

This order is fixed and documented. It will not change between scans or sessions.

Dashboard pillar breakdown now shows buyer journey context on hover

Hovering any pillar row in the dashboard breakdown opens a tooltip showing:

• ■The stage number and pillar name
• ■The buyer journey question for that stage
• ■A description of what the pillar measures and why it matters
• ■The current score and benchmark for your site

The tooltip overlays without shifting any layout. All seven pillars work the same way — no special cases.

Scan results boilerplate reflects actual agents

The summary line on scan results previously hardcoded "UX Sentinel found N Rallycards" regardless of which agents actually contributed findings. It now derives the agent count dynamically from the findings on the scan.

A scan with findings from Lumen, Page Vitals Enforcer, and UX Sentinel now reads: "3 agents found 6 Rallycards."

Additional fixes

• ■Scan list NEW badge now shows the actual count of unreviewed Rallycards per scan, and clears when all cards are approved or dismissed. Previously always showed "1 NEW."
• ■Lumen history page empty state button now matches the header button style and opens the public Lumen check in a new tab.
• ■Script Intelligence cards on authenticated scan results now show a benchmark estimate range instead of "available after signup."
• ■Baseline badge in the Scan Timeline is now legible — bordered pill with visible contrast instead of near-invisible dim text.
• ■Agent colors, pillar display order, and agent attribution rules are now enforced as product constants — preventing regression across future releases.

No API changes. No schema migrations beyond the addition of Agent Visibility Rallycard generation.

We spent today's sprint doing something unsexy but necessary: auditing every pillar in the scoring engine against real scan data, tracing the math, and fixing the gaps between what the algorithm was supposed to measure and what it was actually measuring.

What changed

• ■Messaging Clarity: GTM motion now comes from conversion architecture (CTA text, pricing presence, free-tier language, and form field density), not tech-stack detection. The nav complexity penalty was removed from Messaging Clarity.
• ■Conversion Friction: `analyzeUxSignals()` now populates real DOM-based friction flags. Field count scopes to the first form in main content, thresholds are GTM-motion-aware, and conversational interface detection (Drift, Intercom, Qualified, Chili Piper, Typeform, Fillout) can reduce friction and suppress demo-gating flags where appropriate.
• ■Content Clarity + Signal Hierarchy: both now map and store real signals. Content Clarity surfaces audience/differentiation/value-prop signals and applies a double-generic penalty when needed. Signal Hierarchy now uses heading nesting data and can score above benchmark for strong structure.
• ■Agent Visibility: signal mapping now includes structured data, OG image, llms.txt quality, crawler blocking, and citation-block signals. `aiVisibilityScore` now includes `agent_visibility` at 20% weight.

Tooltip and description accuracy pass

• ■Messaging Clarity copy now reflects deterministic scoring (LLM-generated Rallycards do not set pillar scores).
• ■Conversion Friction is now the canonical label everywhere; Form Friction references were removed.
• ■Friction thresholds now reflect motion-aware limits instead of one fixed field-count rule.
• ■Load Intelligence wording now references LCP, FCP, CLS, and TBT (INP removed).
• ■Seven pillars is now consistently reflected across product and docs references.

Scores are now stable

These changes align the engine with its intended design. Scan-to-scan variance remains normal for real pages, but algorithmic bands no longer collapse to similar values regardless of quality. Strong pages score above benchmark and weak pages score below it for the right reasons.

No API changes. No schema migrations. Existing scan history is unaffected.

What Lumen monitors

• ■llms.txt — whether your site has published a structured AI discovery file, and whether it has enough content for AI systems to generate accurate citations.
• ■robots.txt AI directives — whether GPTBot, ClaudeBot, PerplexityBot, and other AI crawlers are permitted to access your site, or blocked by Disallow rules.
• ■JSON-LD structured data — whether your key pages have schema.org markup that AI research agents can parse into structured citations.
• ■OpenGraph completeness — whether og:title, og:description, and og:image are present across conversion pages.
• ■Canonical URL configuration — whether AI systems following links to your content land on the correct canonical page.
• ■Script blocking weight — whether third-party script load prevents AI crawlers from parsing your page content.
• ■Blocking signal detection — whether bot protection systems are preventing machine access entirely.

Agent Visibility is the 7th pillar in every WebUX Score. Lumen findings surface as Rallycards on the Heroboard with qualitative pipeline impact framing — AI buyer shortlist exclusion risk, reduced citation accuracy, and crawler access failures.

Generate llms.txt from your scan

When Lumen detects a missing or incomplete llms.txt, a Generate button appears in scan results. One credit synthesizes a ready-to-deploy file from your scan data — product description, key pages, key terminology, and citation guidance. Download, copy, or forward to your developer in one click. Deploy to your domain root at /llms.txt and re-scan to see your updated Lumen score.

Free AI buyer readiness check

playhero.ai/webux/lumen — no login required. Enter a URL and get a Lumen score in 30 seconds: a grade (A+ through F), sub-scores for Structured Data, Crawler Access, and llms.txt, and a prioritized list of findings. Share your score with one click — pre-filled copy includes your grade, score, and link.

In the app

• ■Lumen in the sidebar under Tools — scan history with grade tiles, average score, and click-through to full results.
• ■Agent Status on the dashboard shows Lumen as the fourth active agent.
• ■Agent Visibility appears in the pillar breakdown on the dashboard.

For agencies

Every URL in an agency workspace is now scanned for Agent Visibility signals. Lumen Rallycards appear on the Heroboard alongside UX Sentinel and Page Vitals Enforcer findings. AI readiness can be included in standard client reporting.

agent@playhero.ai

Send any URL to agent@playhero.ai and receive an automated Lumen + WebUX scan result by email — overall score, top Rallycards, and a link to the full report. No login required. Works for any machine or AI agent that can send email.

Feature — Compare Mode

Enter two URLs at playhero.ai/webux/compare. PlayHero^° scores both in parallel and returns a side-by-side pillar breakdown, a script attribution note when load differences come from technical changes rather than design work, an estimated pipeline delta, and a plain-language verdict. If both variants score identically — as LeanData's current homepage and their new variant do — PlayHero^° says so directly rather than manufacturing a winner. The intended use is pre-validation: before you expose traffic to a new variant, or when you are not running a formal A/B test and need an objective conversion-readiness score to inform the rollout decision. The report answers a specific question — which page is more structurally capable of converting the traffic it receives — without claiming to predict actual conversion outcomes. That distinction matters.

Improvement — Pipeline estimates recalibrated

Growth stage profiles used in pipeline calculations have been updated with 2025 benchmark data across all tiers. ACV and monthly visitor baselines now come from Optifai (N=939 companies, Q1–Q3 2025), KeyBanc Capital Markets 2024, and TheDigitalBloom Pipeline Benchmarks 2025. Enterprise classification now requires multiple strong signals — a single analyst badge is no longer enough to push a mid-market site onto enterprise-scale numbers.

Improvement — GTM motion detection

GTM motion — hybrid, PLG, sales-led — now resolves correctly for the majority of scans. The rule-based scorer is used as a fallback when the LLM returns low-confidence results, significantly reducing unknown classifications across sites like Workato, JETSTREAM, and Alkira.

Improvement — Dashboard

The pipeline widget now shows a Confirmed or Estimated label depending on whether findings have been approved, a purple accent to distinguish it as the primary metric, and a single contextual action link. The WebUX Score widget shows B2B SaaS benchmarks alongside your score — industry average 67, top quartile 81. Score history charts no longer show colored fill bands bleeding across the chart when multiple URLs are tracked.

Fix — Scan workspace pinning

Scans initiated from the Scans page now correctly associate with the active workspace. Previously, scans triggered outside a workspace context landed with no workspace attribution, causing pipeline estimates to not appear on the workspace dashboard. A permissions scoping issue affecting pipeline display for certain account structures has also been resolved.

Fix — Workspace avatars

Sidebar workspace avatars now load favicons for all workspaces. When a favicon is white, transparent, or unavailable, the workspace initial letter on an accent-colored background is always visible underneath.

Improvement — Impact now scoped to the affected audience

Previously, a UX finding on a pricing page would apply its conversion benchmark across all monthly visitors to the site. That produced technically derived but practically indefensible numbers — a site with enterprise-scale traffic would show impact estimates that no CMO would act on. PlayHero now scopes each finding to the fraction of visitors who actually reach the affected page and who are in an active buying evaluation. A pricing page finding is scoped to the ~18% of visitors who reach pricing, further narrowed by the buying motion rate for the site's GTM profile. The same finding on a demo request page scopes to a smaller but higher-intent audience.

Improvement — Step-level conversion, not full-funnel

The formula previously applied a full-funnel conversion rate — visit-to-close — to a benchmark lift that was measured at the page level. These don't compose correctly. PlayHero now uses a step-level pipeline entry rate for each page type, sourced from the same CXL Institute and Unbounce Conversion Benchmark research. This measures the probability of a visitor taking the next action from that page — not the probability of them becoming a closed deal. The result is a pipeline opportunity estimate, not a revenue guarantee.

Improvement — Pipeline language throughout

Impact estimates now use pipeline-accurate language: Pipeline opportunities / mo, Avg opportunity value, Pipeline entry rate. The prior wording implied closed revenue. PlayHero has no CRM access and never will at the anonymous scan tier — these are pipeline estimates, always labeled as such.

Fix — Methodology tooltip shows the full scoping chain

Every Rallycard methodology tooltip now shows the complete calculation: monthly visitors → page reach rate (with page type) → buying motion rate (with GTM motion) → affected audience → pipeline entry rate → pillar lift → incremental opportunities → ACV → GTM motion weight. Every assumption is labeled and sourced. Nothing is hidden in a footnote.

Fix — Agency client preview shows agency branding

The client Heroboard preview (/webux/client/preview/[token]) was showing the PlayHero^° mark instead of the agency's logo. Fixed — the preview now correctly resolves and displays the agency logo when one is on file.

Pipeline impact estimates on all future scans will reflect these changes. Existing Rallycards retain their original estimates — re-scanning a URL applies the updated methodology. Assumptions remain editable after signup.

Today we're opening the doors to the PlayHero^° private beta. If you're on this page, you're one of the first people to get access — thank you for being here.

PlayHero^° is a WebUX monitoring platform purpose-built for mid-market GTM teams. It sits between the moment a visitor arrives on your site and the moment they convert — watching for the UX issues that kill pipeline before your team even knows they exist.

The Problem We're Solving

Revenue teams spend enormous energy driving traffic. But when a form breaks, a CTA disappears on mobile, or a page regresses in performance, no one finds out until a rep asks why demos dropped. PlayHero^° closes that gap — automatically, continuously, without requiring a dedicated engineering resource.

How It Works

• ■Run a Free Scan on any URL — no account required to start.
• ■AI agents evaluate your pages for conversion-impacting UX issues across accessibility, performance, layout, and messaging.
• ■Issues are surfaced as Rallycards on your Heroboard — prioritized, human-approved, and ready to act on.
• ■Each Rallycard links to a Resolution Brief — a structured fix recommendation with business impact context.

What's Included in Private Beta

Who It's Built For

What's Coming Next

• ■ElevenLabs TTS narration on Resolution Briefs (Growth: standard voice · Scale: voice cloning).
• ■Weekly Heroboard Digest email — KPI pills, top Rallycards, and a single-CTA action summary.
• ■Drift Detector Phase 3 — proactive regression alerts when a page drifts from its last-known-good baseline.
• ■Public roadmap and upvote system.
• ■Agent-journey monitoring — track conversion health across both human visitor sessions and AI-driven buyer journeys.

We're building this in the open, shipping fast, and listening closely. If you hit something broken, something confusing, or something missing — we want to know. Use the feedback button in the sidebar or reply directly to your invite email.