AI visibility tracking for PR agencies: separate the evidence

A defensible PR reporting model that keeps answer samples, citations, search demand, human referrals, conversions, and crawler evidence separate.

Share
AI Visibility Tracking for PR Agencies: What Evidence Can Prove

PR agencies need reports that a client can reproduce. AI visibility is not one measurement. Keep answer samples, citations, search demand, human referrals, conversions, and crawler evidence separate because each answers a different question.

Six evidence types to report

  • Answer samples. The response returned for an exact prompt, provider, model, date, location, and browsing state.
  • Citations. The URLs or domains shown in that recorded answer, with the prompt and timestamp attached.
  • Search demand. Queries, impressions, clicks, click-through rate, and index state from a search console or equivalent source.
  • Human referrals. Observed sessions arriving from an AI product or campaign marker, with the referrer and landing page preserved when available.
  • Conversions. Defined goals and recorded revenue in analytics. A conversion is an analytics event, not a citation label.
  • Crawler evidence. Requests from identified agents, requested paths, source coverage, timestamps, and HTTP outcomes when the infrastructure records them.

Why a blended score is hard to defend

A score can summarize a defined prompt study, but it depends on the prompt set, providers, model versions, locations, sampling schedule, and weighting formula. It should not be presented as a client's position across every AI system.

Do not add unlike measures together. A citation is not a click. A human referral is not proof of which answer influenced the visit. A crawler request is not proof of indexing, citation, recommendation, or conversion. Unknown coverage is not zero activity.

Build the client report around the business question

  • If the question is what an assistant said, attach the exact answer sample.
  • If the question is what source it used, attach the cited URL list.
  • If the question is whether demand changed, attach the search export and window.
  • If the question is whether people arrived, attach referral and session evidence.
  • If the question is whether the work produced a business event, attach the goal or revenue definition and attribution rule.
  • If the question is whether a site was reached, attach crawler evidence and the covered sources.

Put the method, date range, source coverage, and limitation beside every metric. Keep answer screenshots, citation exports, search data, analytics outcomes, and infrastructure observations in separate sections or appendices.

Where SearchSeal fits

SearchSeal provides privacy-friendly website analytics for human sessions, acquisition sources, goals, revenue, and human AI referrals, alongside a separate crawler evidence module. It does not run prompt campaigns, calculate share of voice, or turn sampled answers into a universal visibility score.

A defensible PR deliverable makes clear what was observed, what was sampled, what was attributed, and what remains unknown. It never claims that crawling caused a citation, referral, recommendation, or conversion.

Simple website analytics with AI visibility built in