How to Audit Your Law Firm's Visibility Across ChatGPT, Gemini and Google

Part of our law-firm search visibility guide. For the workflow behind this advice, see our approach to discovery research.

Define the panel before collecting answers

Choose questions from real intake themes, then remove confidential detail. Label each question by practice, jurisdiction, intent, and whether it names the firm. Include questions about situations the firm can actually serve.

Freeze the first panel and assign version numbers to later additions. Otherwise a better-looking result may simply reflect easier questions.

Record enough context to repeat the check

For each observation, capture the exact question, date, engine, product mode, location context, whether search was used when known, named firms, citations, and a saved answer. Keep failed checks out of the denominator but report their count.

Do not treat every absence as equivalent. The answer may contain no business recommendations, a list that excludes the firm, or an ambiguous name. Mark the distinction.

Separate scores from diagnosis

Calculate a mention rate as answers naming the firm divided by valid observations in that defined panel. Name the denominator and sampling period. It is not market share or the probability that every customer will see the firm.

For each missing area, inspect public information. Record a possible explanation and a next check, not an asserted ranking cause. Compare competitors only within the same panel.

Produce a decision, not just a report

Prioritize a few findings by importance to the client, strength of the evidence, effort, and risk. Assign an owner and review date. Save a work log beside the observation log.

Repeat the panel under comparable conditions. Changes in answers, sources, or product behavior are themselves useful findings. Keep them visible instead of smoothing them into a single upward chart.

Questions worth asking

How many prompts are enough?

There is no universal threshold. Cover the important services and intents, repeat observations, and disclose the panel size and limits.

Is this a controlled experiment?

Usually it is monitoring. A stronger causal study needs controlled conditions and a design that addresses confounding changes.

Sources & scope

Platform documentation below supports the technical context. The suggested workflows and examples are our editorial recommendations, not claims of measured customer results or access to private ranking systems.

Google: AI features and your website OpenAI: crawler controls

Continue the investigation

Make the method concrete: explore an example visibility report. The tool is a demonstration, not a live audit.