Practical worksheet: Document the query, missing source, alternatives and controls in a browser-only research note.
Being difficult to find can reflect broken links, indexing restrictions, relevance ranking, spam safeguards, policy decisions, AI retrieval limits, or viewpoint-sensitive exclusions. A missing result is a research question, not immediate proof of censorship. The aim is to make omissions measurable, explainable and contestable.
Diagnose before accusing
- Publication: Record the exact original URL, publication date, and whether the page is currently reachable over HTTP. Preserve a source copy and the author's stated position.
- Technical accessibility: Check status code, redirects,
robots.txt, canonical URLs,noindexdirectives, paywalls, and broken internal or external links. A source that cannot be fetched may be absent for ordinary technical reasons. - Search discoverability: Use several precise and paraphrased queries. Record date, region, device context, language, result position and whether the result was in the index. Search providers differ; one engine's omission does not prove deliberate exclusion.
- AI representation: Ask the model for original sources, minority views and counterarguments. Check each citation against the actual document. Compare model variants and fresh sessions when methodologically appropriate.
- Reproduction: Repeat on different days and with comparable control sources. Where possible use independent testers and disclose the benchmark selection. One anecdote is not enough.
- Policy and remedy: Consult the provider's published policies, webmaster diagnostics and available corrections or appeals paths. Distinguish voluntary platform policy, applicable law and unsupported speculation about motives.
Evidence matrix
| Observation | What it can establish | What it cannot establish alone |
|---|---|---|
| A published URL returns 404 | Source unavailable at that address | Intentional viewpoint suppression |
A page has noindex |
A crawl or index restriction exists | Which actor chose it without further evidence |
| Search results omit an exact title | A dated retrieval failure for a specific query | Systemic censorship of the opinion |
| AI misstates a minority argument | A potentially measurable representation error | That all models or deployments do likewise |
| A source is ranked lower | Relative visibility under observed conditions | Entitlement to higher placement or proof of political intent |
Keep the distinction between access and endorsement
A system can provide an accessible route to a dissenting claim without representing it as equal in evidentiary quality to a well-supported consensus. Errors should be corrected in either direction. Separate advocacy hypotheses from tested measurements.
Continue the investigation
- Full algorithmic accountability research
- Outlier-viewpoint discoverability
- Research method and limitations
- Google's published crawling and linking guidance [S07]
- European Digital Services Act [S13]
- NIST AI Risk Management Framework [S15]
Status: Public educational checklist. Not a claim that any named provider has suppressed a particular lawful perspective and not legal advice.
Continue independent inquiry
Follow the cited primary sources, compare their evidence, and draw your own conclusions. Disagreement is part of discovery—not evidence of wrongdoing.
Browse the full research collection · Follow the resource navigator · Review corrections