Measure answers, not a technical readiness score
A technical scan tells you whether a public page exposes crawlable, structured information. AI visibility measurement asks a different question: when an AI provider answers a selected customer prompt, does it mention your brand, cite your pages, or recommend alternatives?
Keep those results separate. A strong technical score does not mean the brand will appear in a generated answer. A useful measurement report records the prompt, market, provider, run date, successful-answer coverage, brand mentions, position, citations, and competitors.
Define the denominator before comparing rates
For a prompt set, mention rate should use successful answers as its denominator. If 6 of 12 successful answers mention the brand, the mention rate is 50%. If 3 additional provider requests failed, they are coverage failures and should not silently become three more no-mention answers. Report both the number of successful answers and the failures so readers can judge how much evidence the rate represents.
The same rule matters for a chart over time. A week with no successful answers is no data, not a 0% mention-rate point. A week with successful answers and no brand mentions is a genuine zero for that measured sample. Distinguishing these states prevents an outage, model error, or rate limit from looking like a visibility decline.
Read the metrics together
Mention rate estimates how often the brand appears among successful answers for a selected prompt set and period. It does not show where the brand appeared or whether it was cited.
Position describes the brand's relative place among the brands named in an answer. Make the measurement rule explicit because an answer can mention brands in prose, a list, or a table. A provider may not return an ordered ranking at all.
Citations identify source URLs or domains attached to a successful answer. A cited page can support one response without guaranteeing future citations. Review the exact prompt and answer to understand whether the citation supports a product fact, comparison, implementation detail, or broader explanation.
Competitor presence shows which other brands appear for the same question. The most useful comparison is prompt-level: a competitor might be visible for a category question but absent from a pricing question. A blended average can hide that distinction.
Check answer coverage and provider differences
AI providers can return different answers to the same prompt. Track the provider and configured model with each observation, along with market and language, and do not compare two time periods as if the setup were identical when it changed.
When a provider fails, keep its error visible and excluded from mention calculations. Investigate whether the failure is transient or related to configuration, availability, quotas, or the request itself. If coverage is low, fix measurement reliability before making a content decision from the rate.
Use citations as evidence for the next action
Start with a prompt where the brand is absent or an alternative is recommended. Read the full answer and inspect the cited sources. Ask what evidence is missing from your public pages: a clear product definition, a detailed comparison, an implementation guide, pricing boundaries, a regional use case, or a credible reference.
Improve the page that can answer that need. State the answer near the relevant heading, include verifiable product details, show examples or limitations, and link to supporting documentation. If the cited competitor page contains materially better evidence, address the evidence gap rather than copying its keywords or format blindly.
Compare periods carefully
For a useful historical comparison, keep the prompt wording, market, provider/model, and cadence stable. Record when the site or product changed. A sampled response can vary between runs, so look for repeated patterns across several successful observations and inspect the source material before attributing a change to one edit.
LLM Scan shows query-level answer history, successful coverage, mentions, cited sources, competitors, and provider failures in the domain dashboard. The AI visibility tracking guide explains the setup and its limits.