Audit layer
Source Trust Readiness
Answer engines have started classifying sources at the domain level. This audit measures the seven governance layers that classification can read — and tells you which ones your domain currently fails.
Why domain-level classification changes the problem
Perplexity has begun labelling domains with source categories such as Government, Academic, and Trusted, and describes its source review as operating at the domain level rather than page by page, weighing signals such as whether a site identifies its authors and corrects its errors.
If classification is applied to a domain, a single well-optimized page cannot inherit trust the rest of the domain has not earned — and a single unattributed, unsourced section can pull down pages that had nothing wrong with them.
That moves authority away from a vague ranking concept and toward a maintained source-reputation layer: a set of publisher behaviours that are machine-readable, checkable, and therefore engineerable.
The seven layers
- 01
Authorship
Whether every substantive page names a human author, whether that author resolves to a stable profile with credentials and a contact address, and whether the byline in the visible page matches the author in the structured data and the feeds.
- —Named human author on every article, answer, and research page
- —Author profile page with credentials, affiliation, and contact
- —Byline, Person JSON-LD, and feed author fields all agree
- —One canonical spelling of the author's name across the domain
- 02
Editorial policy
Whether the domain publishes standing rules for how it sources claims, what it refuses to publish, how AI assistance is used, and who holds editorial responsibility — and whether those rules are discoverable at a stable URL and declared in schema.
- —Public editorial policy at a stable, linked URL
- —publishingPrinciples declared on the Organization node
- —Stated sourcing standard for statistics and third-party claims
- —Stated position on model-assisted drafting and human review
- 03
Corrections
Whether errors are reported, verified, corrected in place, and logged with dates — the single strongest behavioural signal separating a maintained source from an abandoned one.
- —Public corrections policy with a reporting channel and response time
- —Dated correction log describing what changed
- —correctionsPolicy declared on the Organization node
- —Retraction handling for unsupportable claims
- 04
Ownership and funding
Whether the operating entity behind the domain is identifiable, whether its funding and commercial relationships are disclosed, and whether sponsored or affiliate content is separated from editorial.
- —Named legal or operating entity with address and contact
- —Ownership and funding disclosure at a stable URL
- —Disclosure of commercial relationships on pages that discuss them
- —Sponsored, affiliate, and paid-inclusion policy stated
- 05
Citations and evidence
Whether claims carry primary sources with canonical links and observation dates, whether figures are traceable to the underlying study rather than a summary, and whether inference is labelled as inference.
- —Primary-source links on statistics and third-party claims
- —Observation dates on competitor and category assertions
- —Inference distinguished from documented platform behaviour
- —No uncited figures carried over from older drafts
- 06
Entity consistency
Whether the organization, its people, and its products resolve to one stable identity across the domain, the structured data, the feeds, and the third-party record — or fragment into near-duplicates an engine has to guess between.
- —One canonical @id per organization, person, and product
- —Name, address, phone, and email identical across every surface
- —sameAs pointing only to controlled, live, first-party profiles
- —Legacy names retained as alternateName, never as live variants
- 07
Provenance
Whether an engine can establish when content was written, when it was last substantively changed, what version of the site produced it, and whether dates are honest rather than refreshed for appearance.
- —Machine-readable published and modified dates on every page
- —Modified dates derived from real version history, not page load
- —Sitemap lastmod matching the source of truth for each page
- —Material rewrites recorded rather than applied silently
What the audit delivers
- A dimension-by-dimension readiness rating across all seven layers, with the specific evidence found or missing for each check
- The domain's declared trust signals as an engine can currently read them — schema properties present, absent, or contradicted by page copy
- Entity consistency map across page copy, structured data, feeds, and third-party profiles, with every divergence listed
- Provenance audit of published and modified dates against their real source of truth
- A prioritized remediation list separating declarations that can ship immediately from behaviours that have to be established over time
- Optional re-measurement after remediation, run against the same checklist so the two readings are comparable
Applied to this domain first
BackTier runs the checklist against BackTier.com before selling it. The editorial policy, the ownership and funding disclosure, and the dated correction log below are the visible output of layers 02, 03, and 04; the canonical entity IDs and the version-derived last-modified dates carried in the sitemap are the output of layers 06 and 07.
Common questions
What is a Source Trust Readiness audit?
It is a domain-level review of the publisher behaviours answer engines can check: authorship, editorial policy, corrections, ownership and funding, citations and evidence, entity consistency, and provenance. It rates each dimension on the evidence actually present on the domain and produces a prioritized remediation list.
Why does domain-level source classification matter for AI visibility?
Perplexity has begun labelling domains with source categories and reviews sources at the domain level rather than page by page, weighing signals such as whether a site names its authors and corrects its errors. When classification attaches to a domain, page-level optimization cannot compensate for governance the rest of the domain lacks.
How is this different from an AI Visibility Baseline?
A Baseline measures outcomes: whether a brand appears in answers to a frozen prompt set across named platforms. Source Trust Readiness measures inputs: whether the domain carries the governance signals an engine can use to classify it as a source. They are usually run together, because a low readiness rating explains baseline results that content changes alone will not move.
Can a Source Trust Readiness rating guarantee a platform label?
No. Platforms do not publish their classification thresholds and do not accept submissions for them. The audit establishes whether the checkable signals are present, consistent, and machine-readable. What an independent platform does with them is outside any provider's control.