AI transparency <how this site uses AI>

<How I Use AI>

Every article on this site starts with me. I bring the idea, the topic, the outline, and the opinions — drawn from 20+ years working in SEO, PPC, analytics, and marketing infrastructure. Before any draft is written, I answer a structured set of interview questions about the topic in my own words. My answers become the raw material. AI takes those answers and shapes them into a structured, publishable article.

This is a collaboration: I supply the thinking, the observations, and the experience. AI supplies the structure and the editorial discipline. Neither works without the other. The ideas are mine. The draft is ours.

<What AI Does and Does Not Do>

• Does: turns my answers into structured prose, applies consistent editorial standards, formats content for readability, SEO, and AI retrieval.
• Does not: invent opinions I have not expressed, fabricate data or statistics, manufacture experience I have not had, substitute generic examples when I gave a specific one.
• Every article contains at least one sentence that could only have been written by someone who has done this work firsthand — that line always comes from my answers, never from AI.
• If I did not say it, it does not appear in the article.

<Data and Analysis Methodology>

The following tools on this site perform automated analysis. This section documents what each tool checks, how results are scored or classified, and what the criteria mean.

URL Scanner

What it checks: HTTP status codes, redirect chains, redirect type (301/302), response time, canonical tags, and indexability signals.

How results are classified: Pass / Warning / Fail based on status code and redirect depth.

• Pass: 200 status, no redirect chain longer than 1 hop, response time under 2s.
• Warning: Redirect present but resolves correctly, or response time between 2–4s.
• Fail: 4xx or 5xx status, redirect loop, response time over 4s, or blocked by robots.

Results reflect the state of the URL at time of scan and are not cached.

Feed Master

What it checks: Product feed structure, required field presence, field format validity, image URL accessibility, price format, and GTIN/MPN format.

How errors are classified: Error / Warning / Info.

• Error: Missing required field, malformed price, or invalid GTIN format — feed will be disapproved.
• Warning: Missing recommended fields (color, size, material), low-quality image URL — feed may underperform.
• Info: Fields present but not optimized for performance — optimization opportunity.

Redirect Checker

What it checks: Full redirect chains for a given URL — each hop's HTTP status code, destination URL, and latency per hop.

How results are classified: Pass / Warning / Fail based on chain length and terminal status.

• Pass: Resolves to a 200 in 1–2 hops, total latency under 1s.
• Warning: Chain of 3 or more hops, or mixed redirect types (e.g., 301 followed by 302).
• Fail: Redirect loop, terminal 4xx/5xx, or chain does not resolve.

Latency is measured as wall-clock time per hop using live HTTP requests at time of scan.

Image Tools

What it checks: Image format, file size, presence and quality of alt attributes, and general optimization signals (e.g., use of next-gen formats).

How results are classified: Pass / Warning / Fail per image.

• Pass: Modern format (WebP/AVIF), descriptive alt text, file size within acceptable range.
• Warning: Legacy format (JPEG/PNG without optimization), missing or generic alt text, oversized file relative to display dimensions.
• Fail: Missing alt attribute entirely, broken image URL, or file over a threshold that would materially affect page performance.

Typography Tools

What it checks: Font families referenced on a page, font weights loaded vs. used, CSS @font-face declarations, and rendering flags that affect legibility.

How results are classified: Pass / Warning / Notice.

• Pass: Fonts are declared, weights are used, no orphaned font loads detected.
• Warning: Weights loaded but not used (potential render bloat), or font stack missing a fallback for CJK or system rendering.
• Notice: Informational flags — e.g., using variable fonts, loading from external CDN, or declared font not present in CSS computed styles.

Revenue Audit

What it checks: Revenue dataset integrity across channels — normalization quality, field completeness, channel attribution consistency, and anomaly signals.

How results are classified: Confidence score (0–100) per channel row, aggregated to a dataset-level integrity score.

• High confidence (80–100): All required fields present, attribution is consistent, no statistical anomalies detected in channel totals.
• Medium confidence (50–79): Minor gaps in attribution, some fields missing or inconsistent, channel totals within acceptable variance.
• Low confidence (0–49): Missing required fields, attribution conflicts, or totals that do not reconcile across sources.

The audit does not access live payment processors or financial systems — it evaluates the structure and internal consistency of the data you provide.

AI Visibility

What it checks: Signals that affect how AI crawlers and language model retrieval systems access and index a page — robots.txt directives, sitemap presence, structured data, canonical signals, and content clarity markers.

How results are classified: AI Visibility Score (0–100) composed of sub-scores per signal category.

• High (75–100): AI crawlers permitted, sitemap present and valid, structured data declared, canonical is self-referencing, content is clearly segmented.
• Medium (40–74): Crawling permitted but sitemap absent or partial, structured data missing or incomplete, canonical ambiguous.
• Low (0–39): AI crawlers blocked by robots.txt, no sitemap, no structured data, or page structure impedes retrieval.

This tool reflects the technical signals AI systems use — it does not predict whether any specific AI product will surface the content.

Content Sentiment

What it checks: Tone, sentiment polarity, and emotional register of page content — positive, negative, neutral, or mixed at both the page level and by content section.

How results are classified: Sentiment label + confidence score per segment.

• Positive: Language is constructive, affirming, or solution-oriented.
• Negative: Language signals friction, criticism, or problem framing (may be appropriate in context — e.g., a technical troubleshooting guide).
• Neutral: Informational tone with no strong directional signal.
• Mixed: Competing signals within the same section — e.g., a pros/cons structure.

Sentiment classification is based on model inference and should be treated as a directional signal, not a definitive editorial judgment.

<Accuracy and Limitations>

AI-assisted content reflects my knowledge at time of writing. Tool analysis reflects the state of the data at time of scan. Neither constitutes professional advice.

Results from automated tools should be verified before making infrastructure changes, publishing corrections, or drawing conclusions from aggregate data. The tools are diagnostic aids — final judgment belongs to the practitioner using them.

<Feedback>

If you find an error in methodology, a factual inaccuracy in an article, or a tool producing unexpected results, use the contact page to let me know.

Methodology is updated when criteria change, tools are revised, or a documented error is confirmed. When that happens, the "Last updated" date at the top of this page will reflect the revision.