# 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:** /whyonto
**Extracted:** 2026-03-25T16:51:29.387Z

---
Manifesto

# The web wasn't built for AI.  
That's not an opinion — it's a format problem.

The\_Widening\_Gap

## Optimized for humans.  
Invisible to machines.

Every day, billions of AI agent requests hit websites designed exclusively for human eyeballs. The result is predictable: wasted tokens, hallucinated facts, and an invisible tax on the entire AI ecosystem.

Modern web applications ship hundreds of kilobytes per page. Inside that payload, only a few kilobytes of actual business-critical content — prices, descriptions, documentation, policies — matter for AI consumption.

When AI crawlers ingest these pages, their token budgets burn on layout divs, Tailwind utility classes, tracking scripts, and animation wrappers. The semantic signal drowns in visual noise.

Signal Detection

~2%

Noise Ratio

Extreme

95%+

SYSTEM\_CORE :: CRITICAL

of typical page bytes are invisible to AI understanding

10-100x

SYSTEM\_CORE :: OPTIMIZED

payload reduction possible with clean Markdown extraction

3x

SYSTEM\_CORE :: TRENDING

increase in AI agent traffic to content sites in the past year

$0.00

SYSTEM\_CORE :: NULL

what most sites invest in AI-readability optimization today

Substance

## Ontology: the study of being, abstracted from appearance

Onto borrows its name from ontology — the philosophical study of being, abstracted from appearance. Just as ontology asks what something fundamentally is rather than how it looks, Onto separates what your website is from how it appears.

“The same content, faithfully served in two formats: rich HTML for human browsers, clean Markdown for AI agents. We don't create a shadow web for bots. We faithfully represent the same underlying content in multiple formats — HTML for humans, Markdown for agents — rather than cloaking or injecting deceptive content.”

Onto Design Philosophy

WHAT'S AT RISK

## What happens without AI optimization

Risk\_ID: 01

### Hallucinated Facts — attributed to your brand

AI models fill gaps with plausible-sounding fiction. Wrong prices, outdated features, fabricated policies — all presented to users as your official information. You don't find out until a customer complains or a deal falls through.

SYSTEM\_DIAGNOSTICS

\[FAIL\]"Nike offers free returns on all items" → actual policy: only within 30 days, with exceptions.

Risk\_ID: 02

### Invisible in AI-powered search

AI search engines and shopping assistants can't extract your value proposition from React noise. Your competitors with cleaner markup get cited. You don't. This is happening right now, and most companies don't know it.

SYSTEM\_DIAGNOSTICS

\[FAIL\]Perplexity answers "best project management tool" → cites your competitor's clean docs, skips your React-heavy marketing page entirely.

Risk\_ID: 03

### Token cost inflation for your customers

Every API call that includes your page burns 10-100x more tokens than necessary. Your customers pay more for worse AI integrations of your content. Over time, this makes you the expensive option to integrate — even if your product is better.

SYSTEM\_DIAGNOSTICS

\[FAIL\]596KB HTML page → 148,000 tokens per query → $1.48 per call instead of $0.01.

INDUSTRY\_CONSENSUS

## We're not the only ones saying this.

PLATFORM\_VERIFIED

SYSTEM\_SOURCE

### Vercel

99% reduction

“

Serving Markdown instead of HTML to AI agents reduces payload sizes by ~99%. Complex React pages drop from hundreds of KBs to a few KBs of pure Markdown.

Checksum: ONTO\_VERC\_09X2B

Data\_Integrity\_Locked

PLATFORM\_VERIFIED

SYSTEM\_SOURCE

### Cloudflare

10x token savings

“

Their Markdown for Agents feature converts HTML to Markdown at the edge when requests carry Accept: text/markdown — positioning AI crawlers as first-class citizens.

Checksum: ONTO\_CLOU\_19X2B

Data\_Integrity\_Locked

PLATFORM\_VERIFIED

SYSTEM\_SOURCE

### Anthropic Research

Higher hallucination

“

Token noise — the ratio of irrelevant boilerplate to semantic text — directly degrades a model's ability to extract accurate facts.

Checksum: ONTO\_ANTH\_29X2B

Data\_Integrity\_Locked

## The dual-representation web is already here.

PHASE\_01 // INFRASTRUCTURE\_SHIFT

Large infrastructure providers now expect a web where the same URL yields multiple formats, chosen via the Accept header and User-Agent detection. Early adopters gain better AI-driven discovery, richer citations, and more accurate summaries.

PHASE\_02 // IDENTITY\_STANDARDS

Brands that move now will be accurately represented in AI interfaces as they become the primary way people find and evaluate products. Brands that wait will be misrepresented — or omitted entirely.

PHASE\_03 // STRATEGIC\_WINDOW

The window to move is open.  
It won't stay open indefinitely.

## Don't wait for the shift

See where your site stands right now. It takes 10 seconds.

[Check Your Score

](/)[Review Technical Specs](/how-it-works)

Protocol check takes 10 seconds.

End\_Of\_Manifesto