Brand mentions versus website citations: read an AI visibility report

A mention and a citation answer different questions

A brand mention means the observed answer names your organisation, product or service. An own-domain citation means the interface displays a link or source attribution to a page on your website. One can occur without the other: an answer may name a brand while relying on a third-party source, or it may link to a page without naming the brand in the visible sentence.

Treat factual accuracy as a separate outcome too. A named brand can still be described with the wrong service, market, ownership or contact information. For procurement, ask suppliers to report mentions, own-domain citations, third-party citations and accuracy in distinct fields rather than collapsing them into one visibility score.

• Brand named in the answer
• Own website displayed as a source or link
• Third-party source displayed
• Material facts correct
• User can take a relevant next step

Define one observation before calculating a rate

The unit should normally be one completed answer to one recorded question under stated conditions. Keep the question, date and time, market and language, engine, interface, device or account state where observable, exact answer, displayed URLs and capture evidence. Mark any unavailable setting as unknown instead of guessing it.

Separate completed answers from failed requests, blocked sessions and checks that never ran. A completed answer with no mention is valid evidence of no observed mention. A failed or unrun check is missing data. This distinction prevents availability problems from being reported as poor visibility.

Use explicit denominators

For each metric, show both numerator and denominator. Mention rate can be stated as completed answers that mention the brand divided by completed answers. Own-domain citation rate can be stated as completed answers displaying a link to the organisation's domain divided by completed answers. Accuracy should use an agreed set of material facts and record which facts were checked.

Do not compare percentages built from different question sets, engines, languages, dates or completion rules without showing the change. If 24 of 30 planned checks complete and six mention the brand, the observed mention rate is 6/24 and completion is 24/30. This is a calculation example, not a SALIENS client result.

• Planned checks
• Completed answers
• Answers with a brand mention
• Answers with an own-domain citation
• Answers with a material factual issue

Reconcile manual observations with platform reports

A platform's reporting definitions may measure something narrower than a manual answer review. Google says its Generative AI performance report covers AI Overviews and AI Mode, and counts impressions when links to a site are shown. Its page dimension groups data by the final linked URL, generally assigning performance to the canonical URL. That is evidence about displayed links to your property, not a count of every textual brand mention across all AI systems.

Record platform-reported impressions and clicks separately from manually observed mentions and citations. Note the product coverage and reporting date. Google also says recent data can be preliminary and that reporting limitations apply. A monitoring supplier should preserve the raw observation record so the buyer can understand what each number includes.

Interpret movement without claiming causation

Before-and-after comparison requires a stable core question set and documented conditions. Report content changes, technical fixes, source changes, platform updates and incomplete checks alongside the results. An increase in observed citations does not by itself prove that one intervention caused the change, and a citation or impression does not establish a visit or enquiry.

Use the evidence to decide the next action. Inaccurate mentions call for a fact and source review. Relevant mentions without an own-domain citation may justify strengthening an authoritative page and its internal links. Citations to a weak page may call for clearer service scope and contact routes. Low completion first requires an execution or access diagnosis.

Ask for this reporting table

Before buying an audit or monitoring service, request a sample row structure and definitions. Confirm who retains the answer evidence, how unavailable checks are treated, whether URLs are normalised to the correct domain and how factual issues are reviewed. Agree which decisions the report should support before choosing a dashboard or headline score.

SALIENS helps brands become discoverable and accurately represented in search and AI-generated answers, giving potential customers the information they need to make informed enquiries. We provide GEO audits, AI visibility monitoring, content and technical improvements, and ongoing GEO delivery for UK and international organisations, including English and Chinese requirements. Email info@oxfordintelligence.co.uk with your website, target market and language, and the questions you want measured. Rankings, citations, visits and enquiries are not guaranteed.

• Question and observation conditions
• Completion status and exact answer
• Brand mention and accuracy fields
• Own-domain and third-party cited URLs
• Platform metrics kept separate from manual observations