Byte Digital measures AI search work through recorded visibility observations, accurate business information, website activity and commercial outcomes. Reporting connects the improvements delivered with the evidence available and the next priorities. The measures retain their individual meaning so that clients can see what has changed.
The AI search measurement guide introduces the approach for prospective clients. This article documents the method used to scope and organise the reporting.
What each measure means
| Measure | What Byte records | What it tells the client |
|---|---|---|
| Public access | A page request and the content returned | Whether that test reached the expected information |
| Search indexing | Evidence from the relevant search platform | Whether the platform reports the URL as indexed |
| Brand mention | The business appears in a recorded answer | The brand was included in that observation |
| Citation | A recorded answer links to a specific source | The answer referenced that source |
| Recommendation | The business is suggested for the stated need | The business was presented as an option in that answer |
| Factual accuracy | The answer agrees with checked business information | Whether the representation is useful and correct |
| Referral visit | Analytics identifies a referring AI service | A measurable visit reached the website |
| Enquiry or purchase | A defined business action is recorded | The visitor progressed to the agreed conversion |
| Qualified opportunity | The client confirms fit and progression | The enquiry has commercial relevance |
These measures describe successive parts of discovery and decision-making. They are reported with their source, period and definition, rather than combined into an unexplained score.
Establish the starting position
At the start of an engagement, Byte agrees the business goals, priority services or products, target markets and relevant URLs. It checks the available website measurement and defines what counts as an enquiry, order or qualified opportunity.
The baseline includes a manageable set of customer questions. Branded questions check the representation of known company facts. Unbranded questions examine discovery for a service or buying need. The same distinction is retained in later reporting.
Website changes, campaigns, tracking changes and other relevant activity are recorded as context. This helps explain a change in the results without assigning every movement to one piece of work.
Repeatable AI observations
An observation record includes the exact prompt, platform and surface, date, language, market, search mode, response evidence and cited URLs. Model or version details are recorded when available. Byte also records whether the answer mentioned or recommended the business and whether its factual description was accurate.
A stable core question set is repeated under comparable conditions. New exploratory questions form a separate group. Repeated observations show the frequency of an outcome within the sample, with the numerator and denominator stated together.
For example, a report format can show citations in 3 of 5 observations. This is an illustration of the reporting format, not a Byte or client performance result. It makes the evidence more interpretable than a percentage without its underlying test count.
Platform reporting and website analytics
Google's Generative AI performance report in Search Console reports impressions from AI Overviews and AI Mode. Byte reviews it where the verified property has sufficient data and authorised access, alongside the other available search reports. Its figures remain labelled as Google platform data. Google's report documentation (opens in a new tab) explains coverage and dimensions.
Bing Webmaster Tools and other platform reports are used according to the information available in the client's account. The report names the source and metric. Sampled assistant answers remain a separate observation set.
Website analytics can identify visits carrying a recognisable AI referrer. Visits without that referrer are not retrospectively assigned to AI. An enquiry's optional “how did you find us?” response can add useful self-reported context alongside measured traffic.
Connect enquiries with business outcomes
The website event definition establishes whether a conversion represents a submitted form, a telephone-link click, a completed purchase or another action. The client's sales or enquiry records can then establish qualification, proposals, bookings and won work.
For a local service business, useful qualification may include service area, service required and availability. For ecommerce, order value, returns and margins provide commercial context. For public-sector campaigns, the agreed outcome may be participation or service engagement. The wider measurement guide covers those distinctions.
What the monthly review contains
- The business objective, reporting period and work included.
- Content and technical changes completed, with affected URLs.
- Access and indexing findings, with the scope of the checks.
- Sampled mentions, citations, recommendations and accuracy findings.
- Available native platform metrics and their definitions.
- Identifiable referral visits, tracked enquiries and client-supplied quality information.
- Evidence and public-profile improvements completed.
- Prioritised next actions and the decisions needed to deliver them.
The proposal agrees the reporting scope and review arrangements. Byte checks totals against the source records and notes changes in coverage or definitions. Missing information is marked as unavailable so it remains distinct from a measured zero.
Using Byte's own implementation as evidence
Byte's public knowledge base is an inspectable implementation. Its publication record establishes the preserved files and verified Bitcoin timestamp. Those records demonstrate the delivery and provenance of the resource; search and enquiry results are measured through their own evidence.
The dated effectiveness record in Byte's website case study begins with the 20 October 2026 discovery and access checks. The linked AI search observation log records the first six ChatGPT answers, their citations and selected answer-accuracy checks. Referral and enquiry outcomes are recorded separately when available. Branded factual questions and unbranded agency-discovery questions are tracked separately.
Applying the measures to different businesses
The local-business guide connects observations with service-area fit and enquiry quality. The ecommerce guide connects product research with visits, orders and commercial context. The Byte website case study records delivery checks and publication evidence for Byte’s own implementation.
References
- Byte Digital: how to measure AI search visibility.
- Google: Generative AI performance report (opens in a new tab), reviewed 19 October 2026.
- Byte Digital's approved reporting workflow, 19 October 2026.