Page-level backlink evidence

Domain-level referring data tells you which domains are connected in the crawl graph. It does not automatically identify the current source page that contains a link. Use the evidence workflow when a user asks for exact URLs or asks which pages link.

Start with referring domains

Call referring_domains for the target. Select the returned linkingDomain values that need source-page research.

Create research tasks

Pass up to 50 source domains to get_backlink_research_tasks. The result gives the agent a structured next step for fresh web research.

Use referring_domains for example.com. Pass the first 25 linkingDomain values to get_backlink_research_tasks, then find and verify the exact source pages.

Verify the live page

For each source domain:

1. Search for likely source pages.
2. Open the candidate page.
3. Confirm that the page contains an href to the target domain.
4. Capture the actual source title, linked target URL, visible anchor text, and nearby text.
5. Record confidence based on the strength of the verification.

Do not treat an empty cache lookup as proof that no backlink exists. It only means evidence has not been saved for that target and source pair.

Save confirmed evidence

Call save_backlink_evidence after verification. Saving confirmed evidence is unmetered and remains available even when metered lookup quota is exhausted.

Use list_backlink_evidence to retrieve evidence that was previously saved. It is a cache lookup, not a fresh web search.

Report uncertainty clearly

Separate these statements in any result:

• the source domain appears in Common Crawl domain-level data;
• a live source page was found;
• the page contains a confirmed link;
• the target URL, anchor, and surrounding text were captured.

This prevents a domain-level edge from being presented as a verified live backlink.