September 2026 field note
Measuring search and AI visibility without pretending it is a ranking.
Search positions, AI citations, referral visits, and point-in-time answers are different signals. A useful report names exactly what it observed.
“AI visibility” is not one universal result. A credible observation starts by naming the exact search or AI surface, the method used, and the limits of what the evidence can establish.
Start by naming the surface
Google Search, Google’s generative features, Bing, Microsoft Copilot, ChatGPT search, Gemini, Perplexity, and other products use different indexes, interfaces, models, and retrieval methods. Their outputs can vary by time, query context, location, and account state.
“The business ranked third in AI” is not meaningful. “The exact business name appeared in one dated ChatGPT search observation for this question” is specific enough to review. When an API is used, name the API or benchmark rather than presenting it as a consumer-product result.
Separate search metrics from AI observations
Google Search Console can report impressions, clicks, click-through rate, and average position for Google Search, alongside reporting for supported generative-AI surfaces. Those metrics do not measure every search journey or every AI product.
Bing’s AI Performance reporting describes citation activity across supported AI-generated answers. Microsoft says those citations do not represent ranking, authority, importance, clicks, or a page’s role in an individual response. Grounding-query groupings are not complete user prompts.
OpenAI says ChatGPT referral URLs may include utm_source=chatgpt.com. A referral visit is different from a citation, and neither automatically proves a lead or sale.
Record a reproducible observation
- The exact question tested
- The named platform or API surface
- The date, time, and timezone
- Relevant market or location conditions
- The number of repeated observations
- The result depth examined
- Whether the exact business name appeared
- An observed position only when the surface presents an ordered list
- Cited source domains when the interface exposes them
- Important account, personalization, or method limitations
Use the same method when checking again. Even then, a different answer does not prove that one website edit caused the change. Models, indexes, demand, and competing sources also change. A single check is a snapshot, not a forecast.
Report findings with boundaries
A useful monthly report separates what was observed, what was not established, and what the business can improve next. For example:
Observed: a page received visible citations in a supported surface during the selected range.
Not established: that the page held a ranked position or that a recent edit caused those citations.
Controllable next step: review whether the cited page has current facts, a focused topic, relevant internal links, and complete source support.
Use denominators for repeated checks: “appeared in two of three dated observations” is clearer than “appears consistently.” Do not call source order a ranking unless the product explicitly presents an ordered rank. Do not imply that a citation caused traffic or traffic caused a conversion without supporting analytics.
Turn observations into practical work
Review whether the website clearly states the business name, services, scope, contact details, and evidence. Check crawlability, canonicals, internal links, metadata, structured data, and page experience. Strengthen thin or ambiguous passages with verified facts and original explanation.
Then repeat the observation with the same question panel and record the new result. Report completed corrections and verified page changes separately from platform-controlled outcomes. Vloom can stand behind the written scope and transparent reporting; search engines and AI platforms control rankings, answers, recommendations, and citations.
Request Vloom’s free visibility snapshot, review the website foundation behind visibility, or discuss ongoing visibility support.