GA4 sees the click
It can attribute a visit from an assistant. It cannot see an AI answer that recommended the firm and produced no click.
For legal marketing agenciesOr put your firm name and dates in the header first ↓
Give clients an AI visibility report you can defend.
A free, copy-ready report template built for law firms. It separates recommendations from referral traffic, puts the denominator beside every percentage, preserves raw answers and citations, and keeps unsupported “AI ranking” claims out of the deck.
Ungated. No email address. No invented benchmark data.
Markdown · No signup · The header fields are optional and stay in your browser
Client report
Fictional example · not measured
Recommendation share
—
Withheld until run floor is met
Successful answers
—
Never inferred from planned runs
Failed attempts
—
Shown; excluded from share
Top cited source
—
Linked to raw answers
Why the blanks matter
Unmeasured does not render as zero. The template refuses the most common reporting shortcut before a client ever sees it.
The template
Ten sections, from decision to raw evidence.
Take it blank, or fill in the six header fields — firm, who it is for, who prepared it, market, practice area, measurement window — for one client. The metrics stay blank placeholders either way; nothing here produces a number.
Ungated. No email address. No invented benchmark data.
Optional: put your firm and dates in the report headerShow fieldsHide fields
Add your firm and reporting dates, then copy or download the template. Measurements stay blank for your own evidence. No signup is needed, and what you type stays in this tab.
Press Update preview after an edit. Your entries are not saved or sent to us.
law-firm-ai-visibility-report-template.mdCopy-ready
# [Law firm] — AI recommendation visibility report Prepared for: [client / partner group] Prepared by: [agency] Market: [metro, state] Practice area: [practice area] Measurement window: [start date]–[end date] Question-set version: [version] > Directional measurement of sampled AI answers. This is not a ranking, an endorsement of any firm, or proof that an AI answer produced a signed case. ## 1. Decision this report supports [One sentence: the decision the client needs to make after reading this report.] ## 2. Executive finding - Recommended in: [recommended answers] of [successful answers] eligible answers ([share]%) - Failed or excluded attempts: [count] - Spread across repeated runs: [low]%–[high]% - Most frequently recommended competitor: [firm] — [count] of [successful answers] - Most repeated cited source: [source] — [count] citations - Next action: [one evidence-linked action] ## 3. Scope and collection conditions | Field | Value | |---|---| | Client-style hiring questions | [count and link to frozen set] | | AI surfaces checked | [consumer surface or API named precisely] | | Runs per question and surface | [count] | | Locale and language | [location / language] | | Signed-in or session state | [state] | | Successful answers | [count] | | Failed / refused / empty attempts | [count, excluded from denominator] | | Material method changes | [none, or list] | ## 4. Recommendation evidence | AI surface | Recommended | Successful answers | Recommendation share | Spread | |---|---:|---:|---:|---:| | [surface] | [count] | [count] | [share]% | [low–high]% | Recommended means the answer puts the firm forward as one to contact. A mention, citation, or description alone does not count. ## 5. Competitors named instead | Firm | Recommended answers | Share of successful answers | Example answer | |---|---:|---:|---| | [competitor] | [count] | [share]% | [evidence link] | ## 6. Sources shaping the answers | Source | Times cited | Firms associated with it | Evidence | |---|---:|---|---| | [directory / article / firm page] | [count] | [firms] | [answer links] | ## 7. Findings and actions ### Finding 1 — [plain-language observation] - Observation: [what the stored answers show] - Interpretation: [what the evidence may mean, labeled as interpretation] - Action: [specific owner and change] - Completion evidence: [what will prove the work was done] - Rerun date: [date] ## 8. What changed since the prior report | Item | Prior | Current | Method comparable? | Evidence | |---|---:|---:|---|---| | [metric or finding] | [value] | [value] | [yes / no and why] | [links] | ## 9. Limits - AI answers vary between identical and paraphrased runs. - This sample does not estimate all questions potential clients may ask. - Recommendation visibility and referral traffic are different measurements. - GA4 records attributable visits; it does not record an unclicked recommendation. - No change in this report proves that a particular marketing action caused an AI answer. - This report measures AI outputs, not attorney quality or expected case outcomes. ## 10. Evidence appendix Attach or link every question, raw answer, cited URL, collection timestamp, surface label, classification, exclusion, and calculation used above.
Write the executive finding last. Every material sentence should trace to a question, answer, source, calculation, or explicitly labeled interpretation.
The reporting gap
Three familiar reports miss the event the client asked about.
Rank tracking sees the result page
It does not record which firms an assistant put into a private hiring shortlist.
A screenshot sees one answer
It does not show variance, coverage, failed attempts, or whether the same result survives a rerun.
The receipt test
Six checks before a visibility number reaches a client.
A report that fails one of these checks may still be useful internally. It is not ready to support a client claim.
01
Surface named precisely
Consumer ChatGPT and the OpenAI API are not interchangeable labels.
02
Frozen question set
The same client-style questions are preserved for a comparable rerun.
03
Complete denominator
Successful answers and excluded failures appear beside every share.
04
Recommendation separated from mention
Being named as a firm to contact is not the same as appearing in passing.
05
Raw answers attached
A strategist can reconstruct every material claim from stored evidence.
06
Action separated from inference
The report says what happened before suggesting why or what to change.
From blank template to evidence
See the shape with one public law-firm market.
Run a small, public benchmark before discussing a portfolio. The check asks client-style hiring questions, shows which firms were named, and keeps the answers visible. It is a snapshot—not the monthly measurement this template is designed to report.
- 1
Choose a public firm
Use a client, prospect, or your own agency site only with appropriate permission.
- 2
Inspect the answers
Read the firms and cited sources instead of stopping at a score.
- 3
Decide whether recurrence matters
If the snapshot changes a client decision, define the frozen monthly protocol.
Questions agencies ask before using the template
Can we put our agency branding on it?
Yes. The template is free to reuse and adapt. Keep the method, denominator, limits, and evidence links attached to the claims they qualify.
Does this replace GA4 or Search Console?
No. Those tools measure different parts of the journey. Use them alongside recommendation evidence, not as substitutes for it.
Does downloading the template measure anything?
No. It is an empty reporting structure. Filling in the optional header fields does not change that: it writes the firm, market, practice area and window you typed, leaves every metric as a blank placeholder, and stores nothing. Populate the rest only from a documented collection protocol and stored answers.
How often should we rerun it?
Use the same frozen question set and collection conditions when the purpose is comparison. Change the cadence only when it matches the client decision and provider cost.
Can a client publish “top AI-ranked law firm” from this?
The template does not create an AI ranking or endorse a firm. Public comparative claims need their own legal and bar-advertising review, with the sample and method attached.
For the underlying calculation, see Case Recommendation Share. For the full collection rules, see the methodology.