What Answer Source Intelligence measures
Begin with a fixed portfolio of commercially relevant buyer questions and a preselected provider set. Preserve every verified response, timestamp, exact returned URL, target position and provider failure. A source network should be built only from successful checks. An unavailable response is not a negative citation and should never be replaced by an estimate.
The output connects questions, providers and returned domains. It shows how many appearances each source earned in the observed set, how many questions and providers returned it, whether it persisted across comparable runs, and whether its count rose or fell. These are bounded observations, not a claim about the entire web.
Use a disclosed source-authority score
A useful authority score can combine citation appearances, provider reach, question reach, run persistence and recent movement. Each component and its maximum contribution should be visible. The score helps prioritize investigation inside the monitored set; it is not backlink authority, traffic, editorial quality, expertise or proof of trust.
Repeated return is a pattern worth inspecting. It is not automatic proof that the source is accurate, influential or responsible for the answer.
Separate source types before choosing an action
An owned page, government record, academic study, technical documentation, community discussion and independent publisher require different responses. Strengthen an existing owned answer when it lacks a self-contained passage. Put a primary public record beside a jurisdictional claim. Expose version history and examples for documentation. Address a recurring community question with tested first-party evidence rather than imitating personal posts.
For independent publishers, create a genuinely useful evidence asset: original data, a documented method, named expert context and downloadable material. Make it easy to review, but never buy, fabricate or imply editorial endorsement. Ethical earned media is an opportunity to make evidence independently assessable, not a recipe for manufactured citations.
A source gap is not a backlink gap
A backlink gap asks which sites link to competitors. An answer-source gap asks which sources were returned for a particular fixed question and provider response while the monitored owner was absent, lost or trailing. The system should show the exact leading domains, categories, appearance counts, example returned URLs, target coverage and mapped owner.
Configured competitors come from a separate public-site benchmark supplied by the customer. A returned competitor can therefore be labelled explicitly without treating every commercial domain as a competitor. Compare its public passage, evidence, freshness and format. Do not infer its traffic, private rankings or commercial success.
Turn each pattern into a controlled decision
- Owned visibility loss: preserve the canonical owner, compare before-and-after returned URLs, repair any evidence or passage regression, then rerun the identical test.
- Primary-source lead: cite the current original record beside the exact claim and explain the first-party contribution separately.
- Documentation lead: publish stable, versioned instructions with examples, limitations and change history.
- Competitor source lead: close the measured evidence or answer-format gap without copying the competitor.
- Publisher lead: create stronger independently reviewable evidence and use ethical outreach only where editorially relevant.
Verify without rewriting the denominator
Every action needs an owner, target URL, current citation count, expected completion state and repeat test. Rerun the same questions against the same providers, preserve failures, compare exact returned URLs and keep crawl evidence separate from provider evidence. If the question or provider set changes, begin a new baseline rather than presenting the movement as comparable.
What a professional export should contain
Export one source row with domain, role, category, authority score, current appearances, change, question reach, provider reach, persistence, example URLs and appropriate response. Export one question row with state, checks, target citations, owner URL, unsupported owner claims, source leaders, source-type mix, lost and gained providers, action, acceptance test and evidence class. The report should also state plainly that a returned domain does not prove trust, backlinks, traffic, endorsement or causal influence.