# Velasco Lawyers: a longitudinal GEO study.

A protocol for investigating a multilingual legal-services website: the questions, collection method and intervention log.

## Study status

Protocol draft · Baseline collection has not started

The protocol specifies how responses, sources and website changes will be recorded and compared.

Relationship disclosure: Luca Vicini has an ongoing marketing and website role with Velasco Lawyers. This is an affiliated practice setting, not independent validation of a consulting outcome.

## Research question

How does Velasco Lawyers appear in AI answers to relevant legal-services enquiries across defined languages and markets, and how do those observations change after documented website interventions?

## Collection protocol

- Agree the buyer-intent prompt universe and competitors before the first run.

- Record language and market separately; distinguish app observations from API runs.

- Record engine, model or configuration where available, timestamp and search/grounding settings.

- Repeat observations and retain raw answers, citation URLs, errors and collection costs.

- Freeze the baseline before interventions; log each change with its date and rationale.

- Publish counts, denominators, trends and limitations together.

## Measurement definitions

Metric | Definition for this protocol

Mention rate | Valid responses mentioning the brand ÷ all valid responses in the defined panel.

Citation rate | Valid responses linking to the brand’s site ÷ all valid responses in that panel.

AI share of voice | Responses mentioning the brand ÷ the sum of response-level mention counts for all brands in the fixed comparison set. Count each brand once per response.

Prominence | Record first mention position and recommendation context separately; do not imply a stable search ranking.

Business outcomes | Track identifiable referrals and qualified enquiries separately from visibility observations.

## Baseline & intervention log

The reporting will pair counts and denominators with an intervention timeline. Changes will be assessed alongside product updates, competitor activity and sampling variation.

## Reusable protocol files

Download the field definitions and empty observation / intervention templates. They contain no client data or sample results.

## CSV field definitions

`run_id`: unique collection batch. `prompt_id`: stable prompt identifier. `language`: ISO language code. `market`: country/market, separate from language. `surface`: consumer_app or API. `model_configuration`: observed product/model/settings; use unknown when unavailable. `repeat_number`: observation number within the batch. `competitor_set_id`: versioned fixed comparison set. `valid_response`: whether a usable answer was received. `brand_mentioned` and `brand_site_cited`: true/false only for valid responses; leave blank on failures. `first_mention_position`: character offset in the retained answer (zero-based), blank if absent. `citation_urls`: all cited URLs, JSON array. `source_domains`: distinct cited hosts, JSON array. `raw_response_file`: relative path to retained original answer. `error`: failed collection reason. `cost_currency`: ISO currency. `cost_amount`: recorded collection cost; leave blank if unknown.

`baseline_run_ids` and `followup_run_ids`: JSON arrays of collection IDs. `confounders`: concurrent changes or external events. `status`: planned, applied or reviewed. Do not put client enquiry details or other personal data in these templates.