Evidence-oriented technical SEO review

Soft 404 Evidence Builder

Compare suspected pages with your real error page and healthy examples. Review status behavior, content sufficiency, phrase hits, similarity, redirects, indexing conflicts, and page importance without an opaque probability score.

Browser-only processingNo live URL fetchingEvidence, not a Google verdict
Private by default: Processed locally in your browser. Your files are not uploaded. A normal refresh clears the working data.
  1. 1Add data
  2. 2Map or configure
  3. 3Analyze
  4. 4Review issues
  5. 5Export

Step 1

Add page evidence

Use crawl mode, HTML mode, or both. You need at least one suspected page and one known-good or true error comparison.

Crawl CSV or TSVBest for status, redirects, indexability, inlinks, and large audits

CSV or TSV, up to 30 MB and 50,000 rows. Include a Group/Role column identifying suspected, good, or error rows.

Saved HTML filesBest for template, text, phrase, and empty-shell evidence

Add individual .html or .htm files, or unencrypted ZIP archives containing HTML. Maximum 5,000 expanded HTML files or 50 MB total.

Step 2

Configure extraction and phrases

One phrase per line. Multilingual examples are heuristics, not authoritative classification.

Simple tags, IDs, and classes only, such as main, #content, or .product-main.

Status provenance: crawl statuses are labeled “from crawl export.” HTML statuses are labeled “user-entered.” Saved HTML without a supplied status remains unknown.

Run the evidence comparison

Analysis is deterministic and runs in a Web Worker so the page remains responsive during large jobs.

How to interpret the evidence

Separate signal families

Status, phrase hits, content size, template similarity, healthy counter-evidence, redirects, and importance remain independent. No single low word count or phrase decides the result.

Verify live behavior

Crawl exports and saved HTML are evidence snapshots. Recheck the live response, rendered output, canonical, robots directives, links, and redirect chain before implementation.

Choose an intended outcome

Use the checklist to investigate valid content, an equivalent redirect, 404, 410, noindex, or a page that still needs product and SEO review.

Frequently asked questions

Does this tool reproduce Google soft-404 classification?

No. It compares user-supplied pages and reports transparent evidence such as status behavior, error language, content size, template similarity, redirects, and indexing conflicts. Google may use other signals and systems.

Are crawl files or saved HTML uploaded?

No. Files and pasted data are processed locally in your browser. The tool does not send page content, URLs, titles, or issue examples to the application server or third parties.

What comparison data is required?

Add at least one suspected page and at least one known-good or true error page. Results are stronger when you supply both comparison sets and status provenance from a crawl export or manual verification.

Can saved HTML prove the original HTTP status?

No. A saved file contains markup, not the response that delivered it. Add user-entered status metadata or use a crawl export when status evidence matters.

Why is a short page not automatically labeled as a problem?

Valid contact pages, media pages, product pages, and utility pages can be short. The analyzer keeps content size separate from core fields, healthy-page similarity, status, phrase hits, and other evidence.

Which HTML and archive formats are supported?

The tool accepts individual .html and .htm files plus unencrypted ZIP archives containing HTML. Live URL fetching remains deliberately unsupported.

Reviewed Jul 2026 · Sources and limitations

Review details: 2026-07-13 · Marc LaClear · v1.0

Reference sources:

Known limits:

  • This evidence profile does not reproduce or predict Google soft-404 classification.
  • Saved HTML cannot prove the original HTTP status unless status metadata is supplied separately.
  • Source HTML can omit content created after JavaScript runs, and selector extraction supports simple tag, ID, and class selectors only.
  • Encrypted or damaged ZIP archives, unsupported ZIP compression methods, live URL fetching, CSS rendering, screenshots, authenticated pages, and automatic site changes are outside this MVP.
  • Low word count, error phrases, and similarity are contextual signals rather than universal commands.

Report an issue with this tool