Begin with a business question

Agree what useful progress would look like.

Choose the services, customers and markets the work is intended to support. Then identify a manageable set of research questions: discovering a provider, comparing an approach, checking a capability or confirming business information.

Keep branded questions separate from unbranded discovery. “What does this company do?” tests factual representation; “Which company could help with this project?” tests another situation. Combining them into one headline percentage can obscure the difference.

Document the starting position before substantial implementation. Record the priority pages, known technical issues and available measurement access. That baseline gives later reporting a meaningful reference point.

Record observations that can be reviewed

Keep the source, conditions and denominator.

For each sampled observation, record the exact prompt, platform, date, country or language where relevant, search mode where known, answer and cited URLs. Note whether the business is mentioned, whether one of its pages is cited and whether the important facts are correct.

Repeat a stable core of questions under comparable conditions. Keep new exploratory questions in a separate group. If a report says the business appeared in a proportion of observations, show the number of tests and explain what counts as an appearance.

For example, a hypothetical result of six mentions in twenty recorded answers describes that sample. It does not mean the company appears in thirty per cent of all customer searches. A small sample is useful for investigation when its limits remain visible.

Connect visibility with customer activity

Different measures answer different questions.

Work completed

Record the pages changed, technical issues addressed and evidence added, with publication dates.

Platform data

Use the Search Console or Bing reporting available for the client's verified account. Name the report and metric rather than implying access to every question asked in an AI assistant.

Website activity

Review identifiable referral visits and the pages they reach. Some journeys lose their referrer or return later through another route, so this will not capture every AI-assisted visit.

Enquiries and commercial outcomes

Where tracking is available, report enquiry events separately from client-confirmed lead quality, proposals and won work. An optional enquiry question about how someone found the company can provide useful context.

Next priorities

Explain which findings support the next content, technical or evidence improvements. A report should help decide what to do, rather than just display a score.

How Byte approaches reporting

A clear record within the agreed engagement.

We agree the scope of reporting around the business and the data available. The process brings together completed work, technical findings, sampled visibility observations and measurable website or enquiry outcomes.

An example reporting structure is: objectives and period; changes completed; access and indexing findings; observed mentions and sources; website and enquiry activity; evidence gaps; next priorities. Data that is unavailable is identified, not treated as zero performance.

This reporting supports our AI SEO and GEO work and helps review an audit roadmap over time. Request an initial review to discuss your objectives and existing measurement.

Sources and further reading

Official documentation for the details.

The AI search measurement method documents the observation records and monthly review structure behind our reporting.