1. Agree on the service and market before testing
Write down the firm's actual service area and the matters it wants to discuss with prospective clients. Ask a designated attorney to confirm the scope. A firm that handles divorce, custody and mediation should not automatically describe itself as handling every family-related matter.
Keep a short intake sheet: verified firm name, office details, current lawyers, matter types, languages actually supported and the correct inquiry destination. These facts become the reference for the audit. Do not publish private client information in the sheet or use it in public AI test prompts.
2. Preserve the baseline
Choose a small set of unbranded discovery questions that represent genuine client needs. Separately test questions that name the firm to assess factual accuracy. Save the full answers, sources and test conditions before changing pages.
An example discovery prompt is “Which family-law firms should someone consider for a divorce in [verified service area]?” Treat the resulting list as one observation in that product and session. Do not describe it as a permanent ranking.
For each observation, save the prompt, date, product, visible model or mode, location context, whether search was enabled, mentioned firms and cited URLs. Record unavailable tests rather than silently dropping them. Explain the denominator used in any summary. Our guide to how AI assistants mention law firms explains the distinction between access, accuracy and selection.
3. Check whether important pages can be reached
Have the website team inspect the main family-law page, lawyer biographies, office pages and contact page. Record the response status, canonical URL, crawl directives, indexing status and mobile behavior for each. A sitemap should contain the intended public URLs, and normal links should lead to them.
Google's official guidance connects eligibility in its generative search features to indexing and snippet eligibility. Other products have their own access rules. Check OpenAI's search-crawler documentation and Anthropic's crawler guidance separately. A successful Google live test does not establish access for every provider.
4. Correct contradictions before writing more
Build a page-by-page correction log. Give each issue a source, owner, proposed correction and verification date. Prioritize facts that could confuse a person contacting the firm.
- Conflicting lawyer roster: confirm current roles with the firm; retain the approved roster and corrected URLs.
- Different contact numbers: identify the intended destination; retain the verified number and route test.
- Old service description: obtain attorney-reviewed wording; retain the approved text and public readback.
- Unsupported award or result: verify context or remove the unsupported claim; retain original evidence and permission.
- Important page absent from navigation: add a relevant internal link; record the source page and working destination.
This register is our recommended operational method. It is not a claim that each correction is a known AI ranking factor.
5. Answer a useful question completely
Start with a question the firm repeatedly receives and can answer responsibly. “What happens when I first contact the firm?” is often suitable for a process page. Confirm who responds, what information is requested, how scheduling works and what the firm can accurately say about the next step.
Give the page a descriptive heading and a clear answer. Link to relevant service and lawyer pages. Have the firm review any legal explanation for the jurisdictions involved. Avoid producing dozens of nearly identical location or question pages merely to capture wording variations.
6. Turn approved content into a supporting video
After the written answer is approved, decide whether the message benefits from a real attorney, an illustrated explainer or a clearly labeled AI concept. The choice should serve the audience and the subject.
Keep the approved script, review owner, asset permissions and final transcript together. Explain synthetic presenters clearly. Check the player on mobile, provide useful surrounding text and retain a working contact route. Our family-law AI video examples demonstrate creative formats only. For production scope, see AI video for family-law firms.
7. Make inquiry handling part of the release
A page that attracts attention still needs a functioning next step. Test its buttons and form routing with an authorized internal test. Verify the CRM record, assignment and notification destination. Record whether the notification arrives in the inbox or Junk.
Do not count an automated notification as a qualified lead. Keep form submissions, qualified inquiries and retained matters distinct, and restrict access to personal inquiry information.
8. Retest and explain what changed
Repeat the same baseline questions under comparable conditions. Log changes to the site and to the testing environment. Summarize mentions, factual accuracy and citations separately from website visits or inquiries.
A useful review answers four questions: what was changed, what was checked publicly, what the comparable observations show, and what remains uncertain. If several changes occurred together, do not attribute a result to one change without supporting evidence. If visibility does not improve, report that plainly and use the cited sources and access checks to choose the next investigation.
What a useful engagement should leave behind
Ask for a verified fact register, dated baseline evidence, a prioritized correction log, approved content, a public release checklist and a clear retest record. These deliverables make the work reviewable even when recommendations fluctuate.
Kid in a Hoodie's family-law AI visibility service follows that evidence-first approach. The combined visibility and video program connects reviewed pages with supporting production. Neither guarantees that an AI assistant will recommend a firm.
Editorial note: This is Kid in a Hoodie's recommended marketing workflow, informed by the linked technical documentation. It is not legal advice, a universal AI ranking formula or a report of measured client outcomes.