---
optiview_delivery: edge_optimized
content_type: semantic_markdown
source_url: https://optiview.ai/ai-audit-methodology
brand_content: included
publication_version: "9.0"
published_at: 2026-10-16T20:36:07Z
composition_version: composition-v2.1.0
presentation_version: presentation-v1.2.3
brand_content_source: narrative_graph
delivery_publication: graph:33
narrative_versions:
  - 33
narrative_version: 33
---

# Your website. The evidence behind your audit.

The AI Audit · Method and references

AI Audit captures one public page, measures what it contains, and compares the captured HTML with standard text extraction and Optiview’s prepared content. The report keeps every result tied to source you can inspect and download.

## Three representations. One captured page.

**Captured HTML** contains the returned markup, scripts, styles, data and text. **Standard text extraction** uses whole-body Turndown Markdown where available. **Optiview prepared text** is the prepared Markdown returned by the Optiview capture service. Older reports may identify Mozilla Readability as their baseline; its article selection differs from whole-body extraction.

Every representation uses `cl100k_base` tokenization. Percentage differences between standard and prepared text use the standard extraction as the denominator. HTML-versus-text is a comparison of formats. If a capture is clipped, it is labeled a sample throughout the report.

## The reading budget

The default 10,000-token reading budget is an adjustable example. Choose 2,000, 5,000, 20,000 or a custom value to compare all three representations with the same amount of room. The first-window disclosures contain the actual decoded opening tokens. The token map uses 256-token inspection blocks and offers access to the entire captured representation.

Scrunch recommends keeping primary content within approximately 10,000 tokens in its Site Audit documentation. Actual retrieval budgets depend on the system and task. [Read Scrunch’s published recommendation ↗](https://helpcenter.scrunchai.com/en/articles/11833527-understanding-the-site-audit-tool-in-scrunch)

## Why test a reading allowance?

A page can share model context with instructions, a question and other sources. Anthropic describes context as a finite resource; OpenAI exposes retrieval controls; Exa offers query-relevant excerpts. These support comparing the material a workflow receives.

The room-left comparison subtracts the full standard or prepared text count from the selected page allowance. Negative values are displayed as overflow. Opening-window previews show prefixes; query-based retrieval can select later passages directly.

Read [Your Website Has to Share the Room](https://optiview.ai/resources/ai-reading-budgets/) for the primary-source review, the real Apple example and the reasoning behind this audit.

## Where the source weight comes from

The component bar partitions the received HTML into UTF-8 bytes for executable script elements, embedded app data, structured data, inline styles, markup/layout and text. Individual component token counts tokenize each category separately, so their sum can differ from the whole-document token count. The map color identifies the dominant source component in each block.

Script references describe the received source; external JavaScript file contents, transfer sizes and execution costs are not counted. Structured data, useful links and legal conditions can be important content. A large packaging share describes construction; the actionable findings identify specific items to inspect or fix.

## Does the site serve Markdown when asked?

The audit makes two anonymous requests to the entered page: one with `Accept: text/html` and one with `Accept: text/markdown`. A successful Markdown response must return HTTP 200, a text/markdown content type and a nonempty response that is not an HTML document. The audit gives the site credit regardless of which service provides it; provider headers are not required.

The response details distinguish a switch from HTML to Markdown from a page that serves Markdown to both requests. Blocked requests, timeouts and unverifiable bodies remain inconclusive. An HTML response means only that this request did not confirm Markdown delivery. Samples are limited to 64 KiB, marked when partial and retained with the response headers and check time in the report. Old saved reports are not silently retested. Native Markdown is separate from the audit-generated Optiview preview, and does not certify answer quality.

## From an observation to an evidenced finding

Supported findings are organized under code and clutter, navigation and customer links, business explanations, and internal labels or repeated instructions. Only areas with supported observations appear; the report does not require a fixed number of findings. The Optiview text is a preview generated for this audit, not an installed service on the audited website. Claim-to-condition differences introduced in that preview are excluded from customer-facing findings and summaries; the original measurements and comparison evidence remain in the audit download. The code-and-clutter observation reuses the existing source-byte measurement and is an opportunity to provide text-focused content, not a claim of poor AI visibility. Each supported finding states what we found, why it matters for understanding the business, and how to fix it. Exact source excerpts remain inspectable. Duplicate findings, unconfirmed matches and observations without a meaningful recommendation are excluded from the findings list; original measurements and downloadable analysis remain available. Known attribution-only URL differences are not presented as changed customer destinations. Problems remaining in Optiview’s prepared text are attributed to Optiview preparation or fidelity.

Link-list counts require at least three consecutive Markdown links without prose. Each representation’s share uses its own full text token count. The largest groups are ranked independently, so the two example columns need not contain identical links. Repetition checks preserve punctuation and numeric distinctions; internal-label and joined-form-label checks show literal instances. Prepared heading transitions are checked in the returned Markdown. Useful links, legal conditions and structured facts are not automatically removable waste.

## Passage evidence · version 2

`source-sections-v2` tracks headings and principal wording separately from numeric values, qualification markers, reference markers and destination checks. Visible-word matching normalizes Unicode and case, excludes Markdown destinations and explicit note markers, then maps matches back to the original character and token ranges. Currency and percentage symbols remain in the evidence and reading-window boundaries.

Similar wording is a review candidate, never a semantic-preservation pass. Candidates require a matching heading and at least 55% shared visible-word overlap, or an otherwise matching numeric-text shape. Changed wording, values and qualifiers are shown explicitly. Multiple possible pairs remain unresolved; the test does not choose the occurrence with the most favorable reading position.

Up to five heading-and-support samples are shown in the context cards; up to 40 standard Markdown sections are examined for additional fidelity observations. A nearby-match search examines up to 6,000 characters and the first eight candidate paragraphs. A support sample clipped at 600 characters is identified as a sample; unsampled tails are not treated as invented additions. Deserialized saved reports are reanalyzed from their stored text using version 2.

Located anchors are tested across six configurations: 512 tokens with 64 overlap at offsets 0 and 128; 1,024 with 128 overlap at offsets 0 and 256; and 1,500 with 150 overlap at offsets 0 and 375. Chunk inclusion does not certify unchanged wording or retained conditions. Starting-position counts describe each text representation.

## Claims need their conditions

An explicit note reference is checked separately from its principal claim. The audit can identify a missing marker while locating the claim in both versions, compare a uniquely resolved note body, or find that body elsewhere. Inline conditions and verifiable renumbering are handled separately from lost associations.

When a number is reused for several notes, those conditions are candidates rather than a verified association. Candidate ordering uses literal shared words. The displayed source ranges allow inspection without assigning a condition arbitrarily. The first 20 reference-bearing paragraphs are examined, with bounded candidate excerpts. Link checks compare up to 500 original links using unique visible labels and resolved destinations; they do not test destination availability over the network.

## Initial and rendered responses

New audits retain the actual returned initial and rendered text when the capture service supplies them, together with provenance and recorded completeness. The report shows selected rendered passages not located in the returned initial text. It compares received responses; it does not isolate JavaScript from other response differences.

Older reports may contain only initial/rendered token counts. Those counts cannot identify missing products, eligibility rules or conditions. The audit requests a new capture for passage-level evidence rather than reconstructing an initial response. Missing information in a returned sample is distinguished from information absent from the website.

## Example input cost

The calculator applies your input price per million tokens to the full selected representation and the number of reads. It excludes output, caching, search, retrieval, rendering and network costs. Changing the example leaves the measured source counts untouched.

## Research references

*   [Anthropic: effective context engineering](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents)
*   [Anthropic: preserving context in retrieved chunks](https://www.anthropic.com/news/contextual-retrieval)
*   [OpenAI: retrieval size, latency and relevance](https://developers.openai.com/api/docs/guides/tools-file-search)
*   [OpenAI: search context controls](https://developers.openai.com/api/docs/guides/tools-web-search)
*   [Cohere: model-specific reranking limits](https://docs.cohere.com/v2/docs/reranking-best-practices)
*   [Liu et al.: Lost in the Middle (TACL 2024)](https://aclanthology.org/2024.tacl-1.9/)
*   [Cloudflare: Markdown for agents and representation size](https://blog.cloudflare.com/markdown-for-agents/)
*   [Scrunch: Site Audit categories and content-size guidance](https://helpcenter.scrunchai.com/en/articles/11833527-understanding-the-site-audit-tool-in-scrunch)
*   [OpenAI gpt-oss browser: retrieval backends](https://github.com/openai/gpt-oss/blob/main/gpt_oss/tools/simple_browser/backend.py)
*   [OpenAI gpt-oss browser: content processing](https://github.com/openai/gpt-oss/blob/main/gpt_oss/tools/simple_browser/page_contents.py)
*   [Mozilla Readability: fallback and legacy baseline](https://github.com/mozilla/readability/tree/0.6.0)
*   [Exa: rendered content extraction](https://exa.ai/docs/contents/quickstart)
*   [You.com: retrieval reference](https://you.com/docs/quickstart)

These references inform the approach. AI Audit runs Optiview’s capture and the documented local measurements; it does not call Exa or You.com.

## Your evidence stays inspectable.

Local downloads stay on your device. Publishing creates an unlisted report containing captured HTML, extracted text and retained initial/rendered response text where available, available for seven days. Use public, nonsensitive URLs. [Privacy policy](https://optiview.ai/privacy/)

[Run your free AI Audit →](https://optiview.ai/ai-audit/)


---

# Published brand narrative

# Site-wide brand information

## Who we are

### Optiview — Parse, Retrieve and Publish for AI

Optiview helps businesses and developers make website information easier for AI to use. Parse converts supplied HTML into clean Markdown and source metadata. Retrieve fetches an eligible public HTTPS webpage and prepares its content. Publish helps a verified business prepare its own website content and make approved Narrative Graph context available through a supported connection.

Parse and Retrieve use application keys and a prepaid developer wallet. Publish has separate website setup, ownership checks and a subscription. Optiview does not control whether an outside AI service discovers, uses or cites the resulting content.

Applies to: Optiview

## Why we matter

### A clearer source, from website content to business context

Webpages mix useful information with navigation, layout and application code. Businesses may also have important qualifications, answers and context that are not clearly published.

Optiview gives developers a way to prepare source content for their applications, and gives businesses a way to publish the context their teams approve. Your application chooses how to use Parse and Retrieve results. Your business controls what it publishes through Publish.

Clearer publishing makes information available; it does not determine an AI system’s answer. Each outside service chooses its own retrieval, recommendations and citations.

## What we offer

### Three products, with separate jobs and clear pricing

- **Optiview Parse:** Supply complete HTML and a source URL. Receive clean Markdown, source and output hashes, parser details and processing warnings. Parse does not fetch the URL or execute JavaScript. A successful page uses 1 credit.
- **Optiview Retrieve:** Supply one eligible public HTTPS URL. Optiview checks supported access and publisher policies, fetches anonymous static HTML and parses it. It does not search the web, execute JavaScript, bypass a login or access private networks. A successful page uses 5 credits total, including parsing.
- **Developer credits:** USD $20 buys 20,000 credits shared by Parse and Retrieve. There is no developer subscription, automatic recharge or scheduled credit expiration. Failed jobs consume no credits. Applicable taxes and the total appear at checkout. Source size, quotas and retention limits apply.
- **Optiview Publish:** USD $99 per website per month plus applicable tax. Connect a verified website, review the prepared source and optionally maintain approved Narrative Graph context in Identity, Story, Offerings, Trust, Knowledge and Actions. Choose site-wide or exact-page scope. Saving drafts does not publish them. Eligible AI tools can receive prepared Markdown or HTML through supported delivery, while ordinary visitors keep the normal website.

Developer credits do not purchase Publish subscriptions. Parse and Retrieve do not add a customer’s private Narrative Graph to arbitrary third-party pages. See https://optiview.ai/pricing/ and https://optiview.ai/developers/ for the current offer and limits.

## Why trust us

### Inspect the output and understand what the evidence establishes

Developers can inspect the prepared Markdown, job status, source metadata, processing warnings and credit ledger. Retrieve records the public source response it observed; Parse identifies supplied HTML rather than claiming to verify the current origin.

Publish administrators can review parser output, the combined published-output preview, source checks and publication history. A preview is separate from a recorded live delivery. Saved edits remain separate from the active publication until explicitly published.

Our experiment pages contain selected first-party AI conversations and source comparisons. Delivery records establish what was served, not downstream ingestion, recommendations, citations or revenue. Observed language matches do not establish causation or a consistent visibility improvement. Explore the evidence and limitations at https://optiview.ai/proof/.

## Helpful knowledge

### Common questions about the three products

**Which product should I use?**
Use Parse when you already have HTML. Use Retrieve when you need Optiview to fetch one eligible public page first. Use Publish when your business wants to prepare its verified website and publish approved context.

**Do Parse and Retrieve require website ownership or DNS changes?**
No website connection is needed for these developer products. You remain responsible for source rights and permitted downstream use. Public availability does not grant a reuse license.

**How do developer accounts and keys work?**
Create a developer workspace and save the personal sign-in key and offline recovery code. Create scoped service keys for server-side applications. Email is an unverified contact identifier, not account-ownership proof or an email-only recovery method. Keep all credentials private.

**Can my agent use Optiview?**
The authenticated developer API returns prepared content for your application or model. A downloadable JavaScript client and local MCP adapter are available. Your host must authorize sources and any credit expenditure. Compatibility does not establish adoption or a partnership with an AI provider.

**Does the API include my private Narrative Graph?**
Parse and Retrieve are source-only. Verified website publishing and approved publisher context belong to the separate Publish workflow.

**Does Publish replace our website or CMS?**
No. Your website stays in its existing system. Your web team establishes and tests a supported connection. Your team reviews and explicitly publishes any Narrative Graph additions. Ordinary visitors retain the normal website experience.

**Does every AI tool get the prepared version?**
No. Delivery depends on the website connection, source eligibility and the request. Some tools explicitly request Markdown; supported AI requests may also receive prepared Markdown or HTML through eligible delivery paths. Traditional search crawlers retain the original website response. A claimed tool name does not prove the requester’s identity.

**Does the demo publish or save to our workspace?**
No. The public demo is a separate preview. Its AI-generated examples require review and do not transfer automatically into an approved publication.

**Will Optiview guarantee citations or recommendations?**
No. Optiview prepares and publishes sources. Outside AI services choose what to retrieve, use and cite.

## What to do next

### Choose your starting point

- [Compare Parse, Retrieve and Publish](https://optiview.ai/products/) to choose the right workflow.
- [Create a developer workspace](https://app.optiview.ai/developers) for self-service Parse and Retrieve, then add credits and create an application key.
- [Read the API quickstart and agent-tool setup](https://optiview.ai/developers/) for request examples, supported sources, errors, quotas and downloads.
- [Review pricing and terms](https://optiview.ai/pricing/) for prepaid credits and the separate Publish plan.
- [Preview your website](https://optiview.ai/demo/) to inspect prepared content and explore unapproved narrative examples without connecting or publishing.
- [Plan Publish setup](https://optiview.ai/onboarding/) with your web team. Publishing requires ownership checks, review and a supported connection; contact [hello@optiview.ai](mailto:hello@optiview.ai) for assisted setup.

Already using Publish? Review your Narrative Graph, publish approved changes explicitly, inspect the combined source in Markdown and check recorded delivery in Activity. Keep your existing visibility tools to study downstream AI answers.
