Intelligent Content Transformation
Our AI engine rewrites your pages as enriched markdown -- Schema.org metadata, FAQs, entity definitions, and semantic context, in the format LLMs tokenize best -- and suggests per-page optimizations you can enable, auto-apply and verify against Search Console.
Key metrics
Up to 12× EnrichmentMarkdown OutputPage OptimizationsBrand VoiceEdge-CachedSub-50ms CachedHow AI optimization transforms your content
The AI pipeline identifies what's missing and fills the gaps automatically.
Feature 01
Content Gap Analysis -- AI-Powered Enrichment
Before (origin HTML · thin content)
<div class="product-page">
<h1>Running Shoes</h1>
<p>Great shoes for running.</p>
<div class="price">$149.99</div>
<button>Add to Cart</button>
</div>
<!-- ~200 tokens. No schema,
no FAQs, no context. -->After (enriched markdown)
---
title: Running Shoes
schema:
type: Product
offers: { price: 149.99, currency: USD }
# only facts found on the page
---
# Running Shoes
Running shoes are specialised
footwear designed for forward
motion, with cushioned midsoles
and breathable uppers...
## Specifications
| Feature | Detail |
|---------|----------|
| Midsole | EVA foam |
| Drop | 10mm |
## FAQ (14 questions)
- Are these good for trail running?
- What sizes are available?
- How do they compare to Ultra Boost?
- ... + 11 more
## See also
- [Trail running shoes]
- [Marathon-grade footwear]
# ~2,400 tokens · up to 12× enrichedRatings, prices and claims are never invented — only facts on the page are structured.
Our AI engine analyzes your page content with deterministic settings to identify what's missing for AI systems to truly understand your content. The content gap analyzer then rewrites the page as an enriched markdown document with schema metadata, FAQs, and entity context baked in -- served only to AI crawlers, never to human visitors.
Enrichments injected
- Schema.org LD+JSON structured data (Product, Article, FAQ, etc.)
- FAQ sections with 10-15 AI-generated questions and answers
- Entity expansions with Wikipedia-style definitions
- Semantic relationship mapping (alternatives, related items)
- "What is X?" contextual paragraphs for disambiguation
- Credibility markers and factual statistics
Transformation pipeline
Crawler Detected
GPTBot / ClaudeBot / PerplexityBot
Content Fetched
Origin HTML retrieved
Gap Analysis
AI identifies missing enrichments
Rewritten as Markdown
Enriched .md with schema + FAQs + context → edge cache
Feature 02
Template Optimizations -- Path-Pattern Content Rules
Template Optimizations apply to every page matching a path pattern. Set custom AI prompts for different sections of your site, configure cache TTLs, scope rules to specific domains, and control exactly how the AI transforms each page type -- always in your Brand Voice.
Configuration options
- Path Patterns -- Match URLs with glob patterns like /products/* or /blog/**
- Custom AI Prompts -- Tell the AI exactly how to enrich specific content types
- Cache TTL -- Configure per-rule caching from 5 minutes to 30 days (default 1h)
- Domain Scoping -- Apply rules to specific domains or tenant-wide
Content rule configuration
{
"name": "Product Pages",
"path_pattern": "/products/*",
"domain_id": "d_abc123",
"is_active": true,
"custom_prompt": "Focus on product specs,
sizing guidance, and buyer FAQs.
Add Schema.org Product markup from
the price and availability on the page.",
"cache_ttl": 3600,
"priority": 10
}AI model parameters
Model: Large-scale instruction-tuned LLM Temperature: 0.4 (deterministic) Max Tokens: 4,000 Pipeline: Content Gap Analyzer Cache: Edge-cached (configurable TTL) Fallback: Secondary AI provider (auto-failover)
Crawler-only activation
AI optimization only activates for recognized AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Anthropic-AI, Google-Extended, etc.). Human visitors always receive the original, unmodified page. Zero impact on user experience.
Feature 03
Page Optimizations -- Suggest, Enable, Verify
Where Template Optimizations act on a path pattern, Page Optimizations work one URL at a time. The engine reads the page's own content, your Search Console queries for that URL, and the AI fan-out queries around it, then proposes specific changes -- title, meta description, schema, FAQ block, entity definitions. You enable each suggestion with one click, or let approved types auto-apply.
How it works
- Suggest -- Per-page proposals grounded in the page, Search Console and fan-out queries -- nothing invented
- Enable -- One-click enable per suggestion, or optional auto-apply for the types you trust
- Verify -- After 14 days, each change is checked against Search Console impressions and clicks for that URL
- Revert -- Every optimization is reversible -- switch it off and the original is served immediately
Per-page suggestions
/products/trail-runner-x · suggestions
sampleFeature 04
Brand Voice -- Your Style Guide, Every Rewrite
Every AI rewrite follows your per-property voice guide. Define tone, vocabulary, terms to prefer and avoid, and how formal the copy should be. Template Optimizations, Page Optimizations and generated FAQs all read like you wrote them -- and each property (or each client, for agencies) keeps its own guide.
- One voice guide per property -- clients and sub-brands stay distinct
- Applies to markdown rewrites, FAQ generation and page suggestions alike
- Preferred and banned terminology enforced in every output
- Preview any page in the Content Preview & Diff view before it goes live
Brand Voice -- shop.example.com
tone: confident, plain-spoken, no hype
audience: runners choosing their next pair
formality: conversational, second person
prefer: "trail shoe", "heel drop", "stack height"
avoid: "game-changing", "revolutionary", "best-in-class"
claims: only what the page states -- never invent
ratings, awards or pricesFour steps to AI-optimized content
From crawler detection to enriched response -- fully automated at the edge.
Step 01
AI crawler detected
Wrenda identifies the incoming request as an AI crawler by matching user-agent patterns. GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and dozens more are recognized in under 1ms.
Step 02
Origin content fetched
The original HTML is fetched from your origin server. Static assets (images, CSS, JS) are automatically detected and skipped -- they pass through directly with CDN caching.
Step 03
AI analyzes content gaps
Our AI engine runs the content gap analyzer with deterministic settings. It identifies missing Schema.org markup, thin content sections, absent FAQs, and areas lacking semantic context.
Step 04
Enriched markdown served & cached
The enriched markdown document is served to the crawler with Content-Type: text/markdown and cached at the edge with a configurable TTL. Subsequent requests for the same URL serve from cache in single-digit milliseconds.
Optimized for every content type
AI optimization adapts to your content -- whether you sell products, publish articles, or host documentation.
E-Commerce Product Pages
Product pages get Schema.org Product markup built from the price, availability and specs on the page. AI-generated FAQs cover sizing, shipping, comparisons, and care instructions. Ratings and claims are never invented.
- Schema.org Product + Offer from on-page facts
- Buyer FAQ sections (10-15 questions)
- Spec tables, entity definitions and alternatives
Publisher Articles
Articles are enriched with Article schema, entity definitions for key terms, topical context paragraphs, and comprehensive FAQ sections for featured snippet eligibility.
- Schema.org Article + Author + DatePublished
- Entity expansions for technical terms
- Topical authority signals and citations
SaaS Documentation
Documentation pages receive TechArticle schema, code example annotations, API reference enrichments, and troubleshooting FAQs that AI assistants can surface directly.
- Schema.org TechArticle + HowTo markup
- API parameter and endpoint descriptions
- Troubleshooting FAQs for common issues
See the enrichment in action
Send a request as GPTBot and see how Wrenda enriches the response automatically.
# Request as GPTBot -- triggers AI optimization curl https://yoursite.com/products/running-shoes \ -H "User-Agent: GPTBot/1.0" # Response headers show optimization was applied HTTP/2 200 X-AI-Optimized: true X-Crawler-Action: optimize X-Content-Enrichment: 12.4x X-Cache: HIT Content-Type: text/markdown; charset=utf-8 # Same URL as a regular browser -- no optimization curl https://yoursite.com/products/running-shoes \ -H "User-Agent: Mozilla/5.0 Chrome/120" # Response: original page, untouched HTTP/2 200 X-Crawler-Action: pass-through Content-Type: text/html; charset=utf-8
~200 → 2,400
Token enrichment (up to 12×)
16K tokens
Per-page output limit
< 50ms
Cached response time
AI Optimization FAQs
Common questions about how Wrenda transforms content for AI crawlers.
Wrenda matches the request User-Agent against a continuously-updated list of AI crawler patterns: GPTBot, ClaudeBot, anthropic-ai, PerplexityBot, OAI-SearchBot, Google-Extended, Bytespider, CCBot, Applebot-Extended and more. Detection is per-request — the same URL can be served as enriched markdown to GPTBot and the original page to a regular browser.
For pages routed through the optimize action, Wrenda rewrites the page as a richly enriched markdown document: schema metadata in YAML front-matter, expanded entity definitions, 10–15 AI-generated FAQs from page context, and semantic context paragraphs. Markdown tokenises far more cleanly than HTML, so AI engines parse and cite the content more readily — and humans and Googlebot see the original page unchanged.
No. Optimization is gated by user-agent, so only AI crawlers receive the enriched markdown response. Browsers, Googlebot and Bingbot get pre-rendered or pass-through traffic — whichever you configure per-rule. Google does not penalise serving different content to AI crawlers.
Schema markup is one piece of what Wrenda generates. The full pipeline also adds entity expansion, factual enrichment, FAQ blocks and relationship maps — written specifically for the way LLMs tokenise and weight content. It runs at request time on the edge, so changes to your origin propagate immediately without rebuilding.
Transformation runs on the edge close to your visitors. First-request latency for a long page is typically under 2 seconds; cached responses return in under 50ms. The default cache TTL is 1 hour, so repeat requests within an hour return immediately.
Each cache miss counts as one uncached optimization against your plan's monthly allowance; beyond that nothing extra is billed: crawlers get cached copies or your original pages until you upgrade. Cache hits and pass-through traffic are free. In our testing, hit rates typically exceed 90% once the cache is warm. Plan allowances are listed on the Pricing page.
Get Started
Make AI systems understand your content
Enable AI optimization in minutes. No code changes required -- just point your domain and let our AI engine do the rest.
Start Optimizing14-day free trial -- No credit card required