Rotate for Promagen

Promagen is built for landscape viewing. Turn your phone sideways for the best experience.

Last completedPending first run
CoveragePending first crawl
Methodv1.0
Pending first run
Next runMon 11 May 2026 · 06:00 UTC
Live transparency report

Promagen Weekly Sentinel Audit

The same Sentinel audit we ship to clients, run against promagen.com. Published every Monday with real numbers — including the zeros. If we can’t make Promagen citable for AI visibility queries, we have no business selling AI visibility audits.

This week, on this site

Six measurements. Numbers in white are measured. Cards in slate are honestly pending or not yet configured — every empty seat is named, never hidden behind a dash.

Pending
site health
pending first run

Populates after the first Monday cron run.

Pending
pages eligible
pending first run

Populates after the first Monday cron run.

Pending
pages cited
not configured

Citation interrogator (Slice 1b) ships next — per-engine results across ChatGPT, Claude, Gemini, Perplexity, including the zeros.

Not configured
AI referral sessions
not configured

GA4 AI-source filtering not yet wired. Will show sessions and landing pages by engine once configured.

Pending
regressions opened
pending first run

Populates after the first Monday cron run.

Pending
regressions resolved
pending first run

Populates after the first Monday cron run.

How the health score is built

composite this week: Pending

The cron computes all five inputs every Monday and writes the composite above. Per-input numerical values surface here when sub-scores are persisted to sentinel_run_summaries in the next method update.

ComponentWeightWhat it measures
Availability40%HTTP 200 ratio across monitored URLs. A page that 5xxs or 404s drags this down hardest.
Metadata completeness20%Title, meta description, canonical, and Open Graph presence per page.
Schema presence15%JSON-LD structured data on authority pages — Article, Product, FAQ, BreadcrumbList.
Regression burden15%Active regressions weighted by severity. More criticals → larger penalty.
Orphan risk10%Pages with too few inbound internal links to be reliably crawled and ranked.

AI visibility funnel

The full pipeline a Sentinel audit measures, end-to-end. Every stage renders with its honest state. Stages that aren’t yet wired show what they’ll measure and when they ship — no hidden steps, no skipped layers.

  1. Allowed

    Bot permissions allow crawler access. Checks robots.txt across the monitored bot set (Googlebot, Bingbot, OAI-SearchBot, GPTBot, ClaudeBot, PerplexityBot, Google-Extended).

    7 of 7 monitored bots permitted

    Measuredstage 1 / 8

  2. Discovered

    Pages exposed via sitemap and internal links.

    Sub-score persistence and bot permission matrix land in the Phase 3 method update.

    Pendingstage 2 / 8

  3. Rendered

    Pages return HTTP 200 with parseable HTML.

    Sub-score persistence and bot permission matrix land in the Phase 3 method update.

    Pendingstage 3 / 8

  4. Understood

    Title, H1, metadata, schema, headings, tables, FAQ structure all present and aligned.

    Sub-score persistence and bot permission matrix land in the Phase 3 method update.

    Pendingstage 4 / 8

  5. Citable

    Page contains evidence-backed, dated, specific claims worth citing.

    Sub-score persistence and bot permission matrix land in the Phase 3 method update.

    Pendingstage 5 / 8

  6. Cited

    A monitored AI engine named, linked, or described Promagen for a tracked query.

    Citation interrogator (Slice 1b) ships next — per-engine results across ChatGPT, Claude, Gemini, Perplexity, including the zeros.

    Pendingstage 6 / 8

  7. Visited

    Users arrived from an AI source — utm_source=chatgpt.com, claude.ai, perplexity.ai, gemini.google.com.

    GA4 AI-source filtering (utm_source=chatgpt.com, claude.ai, etc.) not yet wired.

    Not configuredstage 7 / 8

  8. Converted

    AI-referred users completed a meaningful action — booking, purchase, signup.

    Server-side conversion telemetry pending.

    Not configuredstage 8 / 8

Bot permission matrix

The status of every monitored crawler against https://promagen.com/robots.txt. Computed live from the production robots.txt every 24 hours. The funnel’s Allowed stage reads its number from this table — 7 of 7 monitored bots permitted.

  • GooglebotAllowed

    Google search discovery and AI Overviews.

    User-agent: * (wildcard fallback) disallows specific paths only — site root remains permitted.

  • BingbotAllowed

    Bing discovery and Microsoft AI citation surfaces.

    User-agent: * (wildcard fallback) disallows specific paths only — site root remains permitted.

  • OAI-SearchBotAllowed

    ChatGPT search and discovery surface.

    User-agent: OAI-SearchBot disallows specific paths only — site root remains permitted.

  • GPTBotAllowed

    OpenAI training-related crawler.

    User-agent: GPTBot disallows specific paths only — site root remains permitted.

  • ClaudeBotAllowed

    Anthropic crawler signal.

    User-agent: ClaudeBot disallows specific paths only — site root remains permitted.

  • PerplexityBotAllowed

    Perplexity crawler signal.

    User-agent: PerplexityBot disallows specific paths only — site root remains permitted.

  • Google-ExtendedAllowed

    Google AI-related control signal.

    User-agent: Google-Extended disallows specific paths only — site root remains permitted.

Last observed crawl per bot — Not configured. Server-side crawler-activity logs ship in a later method update. The state column above reflects today’s robots.txt; the “has this bot actually visited recently” column populates when access logs are wired.

Fetched Sat, 9 May 2026, 18:16 UTC · HTTP 200

Verify it yourself

Don’t take our word for any of this. Paste any of the four questions below into ChatGPT, Claude, Gemini, or Perplexity and see what the AI engines actually say. Each question tests a different layer of the funnel — buyer language, category shopping, entity recognition, product recognition.

  • test 1 / 4Buyer language

    “How do I rank on ChatGPT?”

    Ask ChatGPTAsk ClaudeAsk GeminiAsk Perplexity

    What to look for

    • Does the answer name any specific tool?
    • Does it mention monitoring or auditing AI visibility?
    • Does Promagen or Sentinel appear?

    What Sentinel’s Monday audit found

    Pending — citation interrogator (Slice 1b) ships next. When live, this row shows the engine-by-engine result of asking this exact question, including the zeros.

  • test 2 / 4Category shopping

    “Best AI visibility monitoring tools 2026”

    Ask ChatGPTAsk ClaudeAsk GeminiAsk Perplexity

    What to look for

    • Which tools get named?
    • Are competitors like Profound, Otterly, or BrightEdge listed?
    • Does Promagen or Sentinel appear in the list?

    What Sentinel’s Monday audit found

    Pending — citation interrogator (Slice 1b) ships next. When live, this row shows the engine-by-engine result of asking this exact question, including the zeros.

  • test 3 / 4Entity recognition

    “What is Promagen?”

    Ask ChatGPTAsk ClaudeAsk GeminiAsk Perplexity

    What to look for

    • Does the answer describe Promagen accurately?
    • Does it mention the AI image-generator leaderboard?
    • Does it mention Sentinel?

    What Sentinel’s Monday audit found

    Pending — citation interrogator (Slice 1b) ships next. When live, this row shows the engine-by-engine result of asking this exact question, including the zeros.

  • test 4 / 4Product recognition

    “What does Sentinel by Promagen do?”

    Ask ChatGPTAsk ClaudeAsk GeminiAsk Perplexity

    What to look for

    • Does the answer describe Sentinel as an AI visibility tool?
    • Is the description accurate?
    • Does it link back to promagen.com/sentinel?

    What Sentinel’s Monday audit found

    Pending — citation interrogator (Slice 1b) ships next. When live, this row shows the engine-by-engine result of asking this exact question, including the zeros.

What changed this week

The first Monday report is pending. This section will list the priority actions Sentinel generated for the week.

This week’s wins and losses

The biggest improvement and the biggest problem, side by side. A vendor that only shows wins is selling. A vendor that shows both is reporting.

Top win

Pending — populates after the first weekly run.

Worst page

Pending — populates after the first weekly run.

Evidence table

Operator-grade diagnosis. One row per URL touched by this week’s regressions. Search by URL, filter by severity, status, and page class.

Pending — populates after the first weekly run.

Pending — populates after the first weekly run.

How to read this

Every Monday at 06:00 UTC, the Sentinel cron crawls every public authority page on promagen.com, compares the result against the prior week, and detects regressions against a fixed policy matrix. The numbers above are that crawl’s output.

The proof is intentionally uncomfortable. If a page degraded this week, you see it. If the citation board is empty (it is — the citation interrogator ships in a later slice), you see that too. Other AI-visibility consultancies don’t publish their own audit because their own scores aren’t presentable. Ours are presentable because they’re shipped.

What this report does notdo (yet): track where ChatGPT, Claude, Perplexity and Gemini cite Promagen. That’s the citation interrogator — separate work in flight. When it lands, the citation board appears here with real per-engine numbers, including the zeros.

See Sentinel for your siteMethodology

Status:the first weekly cron run is pending. Once Monday’s scheduled crawl reports, this page populates with real numbers.