# Onto — AI is reading your website wrong. > Onto serves clean Markdown to AI agents instead of HTML noise. Same content, right format. One line of code. No changes to your site. **Source:** /how-it-works **Extracted:** 2026-03-25T16:51:29.496Z --- System Architecture # How Onto Works One edge layer. Two representations. AI agents get clean, accurate content. Humans get the full experience. Zero code changes to your existing site. Status: Analysis\_Critical ## Here's what AI agents actually receive from your site today. When GPTBot or ClaudeBot crawls a modern Next.js app, they download the full page — 400 to 800KB of HTML, CSS framework classes, JavaScript bundles, inline SVGs, and tracking scripts. Inside all of that: your product description, your pricing, your documentation. Maybe 2 to 5% of the total payload. The rest is visual infrastructure that means nothing to a language model. The agent reads everything. Burns tokens on everything. And when the signal-to-noise ratio is this low, it starts filling gaps with plausible-sounding fiction. That's a hallucination — and it's attributed to your brand. Waste Factor 92-98% Token Fragility High Deep\_Packet\_Inspection // Agent\_View // PACKET\_DUMP\_STREAM :: GPTBot\_Crawl\_2026.log 001
003
... \[400+ lines of structural noise\] ... \[Tailwind utility classes\] ... \[Tracing & Analytics scripts\] Fragment Detected 642

Our product costs $49/mo

... \[Remaining DOM hierarchy\] Total Payload 596KB Ingested Signal Detection ~0.8% SEMANTIC THE SOLUTION — THREE STEPS ## Three steps. Zero friction. 01DETECTION ### Request arrives Any visitor hits your URL. Onto's edge middleware inspects the Accept header and User-Agent string in under 1ms. Human browsers send Accept: text/html. AI agents send Accept: text/markdown, or use known crawler User-Agents like GPTBot, ClaudeBot, or PerplexityBot. 01 02ROUTING ### Edge Middleware Routes Based on who's asking, the middleware rewrites the request. Humans get the full HTML experience — nothing changes for them. AI agents get rewritten to your pre-compiled .onto payload. Sub-15ms overhead at p99. No network calls for human traffic. 02 03SERVING ### Clean Payload Served The AI agent receives compact, semantic Markdown with preserved heading hierarchy, tables, code blocks, and structured metadata. 10-100x smaller. Accurate pricing. Correct feature descriptions. Zero hallucination noise. 03 Protocol Architecture ## Two paths. One URL. Request\_Ingress Path\_Type :: Standard ### Human Path Status: Unoptimized Browser RequestAccept: text/html Middleware CheckNo agent signal → pass through ResponseFull React/HTML experience Core Payload596KB (unchanged) Path\_Type :: Optimized ### Agent Path Status: ONTO\_ACTIVE Agent RequestAccept: text/markdown Middleware CheckAgent detected → rewrite to .onto ResponseClean Markdown + JSON-LD Core Payload8-12KB (98% reduction) Infrastructure\_Mode :: Multi\_Representation Onto Protocol v1.0.4-Edge Processing Engaged Transformation Example ## What Onto extracts Protocol :: Semantic\_Extraction\_v1.0 Format: .onto/mdIntegrity: Validated What an AI agent receives from your pricing page with Onto: \# Acme SaaS — Pricing \> Enterprise-grade automation for modern teams \## Plans \-\*\*Starter\*\*: $29/mo — 5 users, 10K events \-\*\*Pro\*\*: $99/mo — 25 users, 100K events \-\*\*Enterprise\*\*: Custom — unlimited everything \## Key Features \-Real-time analytics dashboard \-SOC 2 Type II certified \-99.99% uptime SLA Total\_Payload12KB total Protocol\_Confidence100% signal. Zero hallucination risk. The Stack ## Four components. One system. LIVE ### AIO Score Calculator Free public tool. Enter any domain, get an instant AI-readiness report. The fastest way to see what AI agents currently receive from your site. ALPHA ### @ontosdk/next Extractor Open-source Next.js plugin. Compiles your routes into clean Markdown at build time. npm install @ontosdk/next. One line in your next.config. ALPHA ### Edge Middleware Bot-aware traffic routing at the CDN edge. Serves .onto payloads to AI agents, full HTML to humans. Under 15ms overhead. COMING Q3 2026 ### Control Plane (Pro) SaaS dashboard for analytics, per-route context injection, and agent policy management. See which AI agents crawl your site and control exactly what they see. ## Ready to index for the agent web? Scan any website and see exactly what AI agents currently receive — and what they could get with Onto. [Scan Your Site ](/)[Read the Scoring Methodology](/scoring) ONTO\_PROTOCOL :: DEPLOYMENT\_READY