AI Visibility Case Study Framework

Why most case studies are weak

Screenshots of one favourable answer do not prove sustained visibility. Credible evidence needs a baseline, repeated observations and clear disclosure of what changed.

The minimum case-study structure

Record the market, prompt set, platforms, baseline period, interventions, observation window and outcomes. Include limitations and any external factors that could have influenced results.

Metrics worth showing

Use mention rate, recommendation rate, citation rate, factual accuracy, referral activity and qualified enquiries where available. Avoid presenting a proprietary composite score without explaining its meaning.

Protecting client value

Publish enough detail to show the problem and outcome, but not confidential data, internal source maps, exact scoring weights or the full implementation sequence.

How Cited applies this

Our case studies are designed to demonstrate evidence and commercial relevance. The diagnostic process and prioritisation logic remain within client engagements.

Show the baseline, not just the win

A credible case study includes unfavourable starting results and the limits of the available data. Publishing only the strongest prompt or best screenshot creates a success story, but not reliable evidence.

Distinguish contribution from causation

AI visibility can change because of website improvements, new external sources, model updates or shifts in retrieval. Unless the design controls for those factors, describe the work as contributing to the observed change rather than proving sole causation.

Connect exposure to commercial value

Where privacy allows, pair visibility metrics with referral visits, branded search, qualified enquiries or sales feedback. This helps readers understand whether increased presence in AI answers influenced real customer behaviour.

What credible evidence looks like

A strong case study separates observed change from claimed causation. It records the starting point, work completed, measurement dates and other factors that may have influenced the result. Screenshots, prompt logs, analytics annotations and source records provide stronger support than an isolated percentage or testimonial.

Commercial confidentiality can be protected without weakening the evidence. Sensitive tactics may remain private, but the timeframe, measurement method, limitations and nature of the outcome should still be clear enough for a reader to assess the claim.