Page Deep Audit (Vision + AI Render)

Page Deep Audit is the deepest analysis Rankion runs on a single URL. Instead of content signals like Content Audit (Site Crawl), Page Deep Audit looks at the landing page as a visual product: 3 screenshots (Desktop, Tablet, Mobile with real UA emulation), Lighthouse metrics, an Opus 4.7 multimodal analysis on layout, trust, CTA, persona fit, critical issues — plus up to 3 gpt-image-2 AI renders that show what the optimal variant could look like visually. This is the tool for CRO and landing-page iteration, not for SEO bulk work.

What it can do

• 3 source screenshots — Desktop, Tablet, Mobile with real user-agent emulation.
• Lighthouse metrics — Performance, SEO, Accessibility, Best Practices as numbers.
• Opus 4.7 vision analysis — user_intent, persona_fit, trust_score, layout_score, cta_score, problem_solution_clarity, above_the_fold_quality, mobile_friendliness_visual.
• Personas + pain points — derived from the visual, not from assumptions.
• Critical issues — prioritized by severity (high / medium / low) with evidence snippet.
• Improvement suggestions — per area (headline, cta, trust, layout, copy, visuals, forms, navigation, seo, accessibility) with a before/after example.
• 3 AI renders — gpt-image-2 generates the ideal Desktop, Tablet, Mobile version as a visual reference.
• Headline rewrite — a ready-to-paste H1 drop-in suggestion.

When to use

• You want to CRO-audit a landing page before pouring money into ads.
• You want to give a designer an objective visual reference ("this is how it should look").
• You're iterating a variant: Audit → Adjust → Re-audit → compare score diff.
• You need persona-driven copy suggestions based on the real page, not on briefing theory.

Workflow

1. Start audit — POST /page-audit with {url, tracking_project_id?, persona?}. Returns 202 + {id, status:"pending", url}.
2. Poll until completed — main flow ~1–2 minutes (scraping → screenshotting → analyzing → completed). AI renders run +6–9 minutes in a separate background job.
3. Read reports — GET /page-audit/{id} returns 6 image URLs (3 source + 3 AI render), an analysis block with scores, personas, issues, suggestions, headline rewrite, and Lighthouse metrics.
4. Iterate — apply suggestions in priority order, deploy the page, run a re-audit, compare the score diff.

Polling example:

ID=$(curl -s -X POST $BASE/page-audit \
  -H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" \
  -d '{"url":"https://example.com/landing"}' | jq -r '.id')

while true; do
  R=$(curl -s -H "Authorization: Bearer $TOKEN" $BASE/page-audit/$ID)
  STATUS=$(echo "$R" | jq -r '.status')
  IDEAL_M=$(echo "$R" | jq -r '.ideal_mobile_url // "null"')
  [ "$STATUS" = completed ] && [ "$IDEAL_M" != null ] && break
  sleep 30
done

API

Pipeline status: pending → scraping → screenshotting → analyzing → completed. AI render fields (ideal_screenshot_url, ideal_tablet_url, ideal_mobile_url) populate one by one after completed — Desktop first, then Tablet, then Mobile.

Credits & Limits

• Main audit: 30 credits.
• AI renders: up to 3×15 credits (Desktop, Tablet, Mobile).
• Full audit with all renders: up to 75 credits per run.
• Async — main flow ~1–2 min, renders +6–9 min after that.
• url is required and capped at 500 characters; 422 on validation fail.
• Cross-team and cross-project access returns 403.

Related modules

• Content Audit — site-wide inventory instead of single deep audit.
• Content Optimizer — optimize the content layer; Page Deep Audit targets visual + UX.
• AI Content Editor — apply the headline rewrite suggestion directly in the editor.