This guide is built from Google's current GA4 documentation, Avinash Kaushik's September 2026 analysis of the new Google Analytics features (Occam's Razor, TMAI #521), published AI-traffic conversion research from Adobe and Shopify, and configurations observed across law firm analytics accounts. Platform details reflect GA4 as of October 2026. Where a recommendation comes from practitioner observation rather than documentation, the text says so.
Before anything else, one warning that pays for the whole read: the single most common analytics defect in law firm accounts is the same lead counted twice, which silently doubles your reported conversions and halves your reported cost per lead. Chapter 6 covers every version of it and the fix.
What This Guide Covers
- Why Law Firm Analytics Breaks: The Measurement Gap
- GA4 Structure in Plain English
- The Clean Setup Checklist
- One Tag, One Truth: Installing GA4 Without Duplicates
- Key Events for Law Firms: Tracking What Becomes a Case
- The Double-Counting Chapter: Five Ways GA4 Conversion Tracking Inflates
- Call Tracking and CRM: From Key Event to Signed Case
- Channels That Matter: Reading Acquisition Like an Operator
- The AI Assistant Channel: Measuring ChatGPT Traffic
- Reports That Answer Partner Questions, Including Ask Advisor
- GA4 Attribution and Assisted Conversions
- Cross-Channel Budgeting: Predictive Planning in GA4
- The Governance Cadence and Common Mistakes
- Frequently Asked Questions
Why Law Firm Analytics Breaks: The Measurement Gap
Most law firm analytics accounts fail before a single report is opened, because nobody decided what the numbers are supposed to prove. The firm installs Google Analytics, traffic accumulates, and a year later a partner asks which marketing actually produces cases. The honest answer is that the account was never built to know.
The fix is a measurement plan, and it takes an hour. Start from the business objective and work down. For a consumer law firm, the macro conversion, the action that directly fulfills the business objective, is a qualified consultation request. The micro conversions are the steps that lead there: a phone call, a contact form, a live chat started, an intake booking, a guide downloaded. This macro and micro structure comes straight from Google's own measurement planning framework, and it matters because it forces a hierarchy. A firm that counts newsletter signups and consult requests as the same "conversion" has built a dashboard that cannot answer the only question that matters.
Write the plan as four lines before touching any settings: the objective (signed cases from the web at or below a target cost), the macro conversion (qualified consult request), the micro conversions (call, form, chat, booking), and the targets by channel. This is the same discipline behind the complete law firm revenue system: measurement follows strategy, never the reverse.
Analytics tools report what happened on the website. A law firm runs on what happened after: the consult, the retainer, the fee. The entire job of GA4 in a law firm is to connect the first world to the second with numbers you can trust. Every chapter in this guide serves that connection.
GA4 Structure in Plain English
GA4 has three levels, and most firms only ever need one of each. The account is the firm. The property is the measurement container where reports live. The data stream is the source feeding it, which for nearly every law firm is one web stream covering the site. Firms with a separate brand per practice area or state typically still run one property per real business entity, because fragmenting data across properties is the fastest way to make every report too thin to read.
The deeper shift from the old Universal Analytics: everything in GA4 is an event. A page view is an event. A call click is an event. A form submission is an event. Some events are collected automatically, some arrive through a feature called enhanced measurement, and the ones that matter most, your lead actions, you define and then mark as key events. Key events are what GA4 used to call conversions, and the renaming matters when reading older guides: anything written about "conversions" in GA4 before 2024 means key events now.
Two structural facts save hours of confusion later. First, GA4 sessions are stitched differently than the old tool, so year-over-year comparisons against Universal Analytics numbers are directional at best. Second, GA4's default data retention for user-level detail is short: extend it to 14 months in the property settings on day one, because the setting is not retroactive. Several of the 2026 features in this guide are also forward-only, collecting from the day you activate them, which makes a correct setup today the cheapest analytics investment the firm will ever make.
The Clean Setup Checklist
These nine settings separate a trustworthy property from a noisy one. A competent developer or marketer completes the list in an afternoon, and each item names what it protects.
Practitioner note on baselines. Keep one unfiltered view of reality. In GA4 terms: before adding filters, know what raw looks like, and annotate the dates when filters and key events change. When a chart moves months later, the first question is always "what did we change," and the annotation log answers it in seconds.
One Tag, One Truth: Installing GA4 Without Duplicates
Every GA4 defect in Chapter 6 descends from the same root: more than one thing on the page sending the same signal. So the installation rule is absolute. The site carries exactly one implementation of the Google tag, either hardcoded or through Google Tag Manager, never both.
Law firm sites accumulate tags the way garages accumulate boxes. A theme that bundles an analytics field, a plugin that injects its own snippet, an old agency's Tag Manager container, and a developer's hardcoded gtag can all coexist, each one innocently installed, together sending two or three of everything. The audit is simple: open the site with Google's Tag Assistant and count how many times the GA4 measurement ID fires on one page load. The answer must be one.
The second thing to understand is that one Google tag can serve several destinations: the GA4 property (an ID starting with G-), a Google Ads account (AW-), and a combined Google tag (GT-). Seeing several IDs in the tag settings is normal and is not double counting. Double counting is the same destination receiving the same event twice. Firms that run serious paid search programs live and die on this distinction, because the Ads account both receives its own tag signals and can import key events from GA4, and Chapter 6 covers how those two paths collide.
Choose Tag Manager when the firm runs call tracking, multiple ad platforms, or a marketer who iterates on events. Choose the plain hardcoded tag when the site is simple and nobody will touch events for a year. Both are correct; running both is the defect.
Key Events for Law Firms: Tracking What Becomes a Case
A law firm needs five key events, maybe six. Not twenty. Every event you promote to a key event dilutes the signal the reports and the ad platforms optimize against, and a conversions list stuffed with trivial actions tells you everything works, which is the same as learning nothing.
| Key event | What it captures | Type |
|---|---|---|
| phone_call | Click-to-call taps on mobile, plus calls from a call tracking platform where connected | Micro |
| form_submit_lead | The contact or case evaluation form, fired once on successful submission | Micro |
| chat_started | A live chat or AI intake conversation opened by the visitor | Micro |
| booking_completed | A consultation scheduled through the calendar tool | Macro |
| guide_download | A lead magnet downloaded with contact details | Micro |
Name events deliberately and document them, because event names are forever in your historical data. Where the action implies a value difference, attach one: a booked consultation is worth more than a guide download, and giving GA4 that hierarchy improves everything downstream, from reports to the smart bidding that your marketing budget ultimately funds.
One philosophical rule keeps the list honest: a key event should mark a human raising their hand, not a page rendering. Thank-you page views, scroll depth, and time on page are useful engagement signals, but promoting them to key events is how accounts drift into vanity. The quality of this list is also what decides whether the intake conversion work in our law firm conversion optimization guide can be measured at all.
The Double-Counting Chapter: Five Ways GA4 Conversion Tracking Inflates
Double counting is the most expensive defect in legal analytics because it flatters everyone. Reported leads go up, reported cost per lead goes down, the agency report looks great, and the firm makes budget decisions on numbers that are roughly half real. Here are the five versions found in law firm accounts, each with its fix.
1. Two tags, one page
A hardcoded gtag and a Tag Manager container both firing the GA4 configuration. Every page view, every event, every key event doubles. Diagnose with Tag Assistant as in Chapter 4; fix by removing one implementation entirely, not by trying to referee between them.
2. The form event and the thank-you page, both marked as key events
The form fires form_submit_lead on submission, the visitor lands on a thank-you page, and a second key event fires on that page view. One lead, two key events. Pick one source of truth per action. The submission event is usually the better choice, and the thank-you page stays as a plain destination for remarketing audiences.
3. Enhanced measurement colliding with your own form tracking
GA4's enhanced measurement includes automatic form interaction events. If a marketer also builds a custom form submission event, the account now holds two similar-looking signals, and sooner or later somebody marks both as key events. Decide which one is canonical, and leave the other unmarked, or switch the automatic form tracking off.
4. Google Ads counting both its own tag and the GA4 import
The subtle one. Google Ads can record conversions from its own conversion tag and import key events from GA4. A firm that sets up both for the same action reports double conversions inside Google Ads, which corrupts smart bidding, not just reporting. Rule: each conversion action lives in exactly one system of record. Import from GA4, or tag natively in Ads, never both for the same action, and check the Ads conversion settings after any agency transition, because this is where handovers break.
5. Call tracking counted twice
A call tracking platform pushes a call event into GA4 while the click-to-call tap that started the same call also fired phone_call. One conversation, two key events. Decide whether the platform or the on-site tap is the counter, and demote the other. Firms reconciling against the CRM, as Chapter 7 recommends, catch this in the first month because web leads exceed intake's count.
The test for a clean account is one sentence: every lead action increments exactly one key event, exactly once, in exactly one system of record. Audit it quarterly by submitting a test form and placing a test call, then counting what shows up. Five minutes of testing beats a year of inflated reports.
Call Tracking and CRM: From Key Event to Signed Case
GA4 ends at the lead. The firm's economics begin at the signed case, and connecting the two is what turns analytics from a traffic report into a management tool.
Call tracking comes first, because for most consumer practices the phone is the majority lead path. A call tracking platform with dynamic number insertion shows each visitor a unique number, ties the call to the session and its source, and pushes the call into GA4 as an event. Configured well, this is how "calls" stop being a blind spot. Configured carelessly, it creates defect five in Chapter 6 and, worse, numbers cached by Google Business Profile that misattribute calls for months. Keep the tracked number pool for the website, keep the firm's canonical number everywhere identity matters.
Then the CRM closes the loop. Tag every lead at intake with its source and bring the outcome back: consult held, retained, case value. The reconciliation runs monthly and asks one question: for each channel, what did a signed case cost? That number, cost per signed case, is the only metric that lets GA4 data stand next to Local Services Ads, referrals, and every other channel on the same scorecard, the same reconciliation discipline the LSA audit checklist enforces, and it is the backbone of the personal injury law firm KPIs the most disciplined firms run on. One caution from the field: channels that generate calls which never touch the website, like LSA, will always look smaller in GA4 than they are. The CRM, not GA4, is the final scoreboard; GA4 is the diagnostic layer underneath it. We rebuilt exactly this pipeline in our marketing attribution case study.
Channels That Matter: Reading Acquisition Like an Operator
The Traffic Acquisition report is where managing partners and law firm owners should start every month, because it answers the operating question: where do the people who become leads come from, and is that mix changing?
GA4 groups traffic into default channels: Organic Search, Paid Search, Direct, Referral, Organic Social, Email, and, since 2026, AI Assistant, which gets the next chapter to itself. The groupings are only as honest as your tagging. Everything Google can detect automatically, it does; everything else, email sends, newsletter links, campaign links from Go High Level or any other platform, needs UTM parameters or it lands in Direct, the junk drawer of analytics. The discipline is boring and non-negotiable: every link the firm distributes carries a source, medium, and campaign, built once in a shared sheet, reused forever.
Two upgrades worth activating. First, Source Group Consolidation, new in 2026, merges the spelling chaos of facebook, fb, and meta-facebook into one coherent source group, retroactively, which ends years of regex workarounds in channel reporting. Second, read channels with engagement columns, not session counts: engagement rate and session key event rate are what separate a channel that sends readers from a channel that sends clients. The same lens applies when newer sources appear, whether that is ChatGPT advertising or a local directory experiment: judge entrants by key events per session, not by traffic.
The AI Assistant Channel: Measuring ChatGPT Traffic
In mid-May 2026, Google Analytics quietly started answering the question every marketing meeting now asks: is ChatGPT sending us business? A channel called AI Assistant now appears by default in Reports, under Acquisition, in Traffic Acquisition, with the medium set to ai-assistant. No regex, no custom channel group, no setup. Avinash Kaushik's analysis of the 2026 GA features (Occam's Razor, September 2026) is the best public breakdown of how it behaves, and the law firm translation follows.
| Channel | Sessions | Engagement rate | Session key event rate |
|---|---|---|---|
| Organic Search | 4,812 | 58.1% | 2.4% |
| Paid Search | 2,140 | 61.3% | 4.1% |
| Direct | 1,966 | 49.7% | 1.8% |
| AI Assistant | 187 | 71.2% | 4.9% |
| Referral | 645 | 55.4% | 2.1% |
Recreation of the GA4 Traffic Acquisition report. The AI Assistant row is small, and that is normal. The quality columns are where the story lives.
Expect the number to be modest: for most businesses, answer engine traffic runs 1 to 5 percent of sessions. The value is in the quality. Adobe's March 2026 research found conversions from answer engine traffic 42 percent better, with 48 percent more time spent on product pages; Shopify's May 2026 data showed 50 percent higher conversion on product pages and 14 percent higher order value. The legal translation: someone who asked an AI assistant a question, got an answer citing your firm, and still clicked through is further down the decision path than the average searcher. Compare the AI Assistant row against Organic Search on engagement rate and session key event rate in your own property; most firms will find the same pattern.
Now the four caveats that keep your reporting honest, all four from Kaushik's analysis and verifiable in your own account. The channel has no history before mid-May 2026, so year-over-year comparisons do not exist yet. The list of assistants Google counts is a moving target, with platforms added and removed month to month, so check the channel definition before shipping a report. Traffic from LLM mobile and desktop apps often carries no referrer and lands in Direct, undercounting the channel. And Google classifies clicks from its own AI Overviews and AI Mode as Organic Search, not AI Assistant, which means the row systematically understates how much of your traffic is really answer-engine driven. Label the number a floor, not a total.
Sources: Avinash Kaushik, Occam's Razor, September 2026; Adobe (March 2026) and Shopify (May 2026) AI traffic research as cited there.
What should a firm do with the row? Three things. To track AI traffic in GA4 beyond the headline row, open the landing pages dimension for the AI Assistant channel to see which pages answer engines send people to; those pages are what you rank for in AI answers, and they tell you where your authority actually lives. Feed that insight into your answer engine strategy, which we built out in the AEO and SEO guide for law firms and tracked across the broader law firm marketing trends: the firms being cited are the firms publishing the clear, sourced, quotable content assistants prefer. And brief leadership correctly: this is a margin story, not a traffic story. A channel that is 2 percent of sessions converting substantially better than organic justifies content investment on value per visit, not on visit count.
The AI Assistant channel is the first default report in any mainstream analytics tool that measures the answer engine era. It arrived understated and imperfect, and it is still the clearest early signal a law firm has for whether its content earns citations inside ChatGPT, Perplexity, and the rest. Watch the quality columns, treat the volume as a floor, and invest accordingly.
Reports That Answer Partner Questions, Including Ask Advisor
A law firm needs four reports and a question box, not forty dashboards.
Traffic Acquisition (Chapter 8) answers "where do leads come from." Landing pages answers "which pages earn them," and read it with key events per page, because a blog post that quietly produces consult requests deserves budget that a high-traffic, zero-lead page does not. Funnel exploration answers "where do we lose people," built once as homepage or landing page, then contact or booking page, then key event; the step with the biggest drop is next quarter's conversion project. Path exploration answers "what do people actually do before contacting us," which routinely demolishes assumptions about how prospects navigate practice area pages.
The question box is new. Ask Advisor, GA4's conversational analysis feature, sits behind the magnifying glass icon at the top of the interface and takes questions in plain English: which channels drove consultation requests last month, how did this month's leads compare to last, which landing pages improved. For the partner who will never learn the GA4 interface, this is the honest recommendation: start with Advisor, let it do the navigation. Two cautions travel with it. Its answers are bounded by what GA4 contains, so it cannot see your CRM or your case values, and on causal questions, why something happened, treat its answers as hypotheses to verify, not findings.
One step further for firms with technical help: Google ships an official Analytics MCP server that connects GA4 data directly to an AI assistant like Claude or ChatGPT. That setup can cross-reference analytics with your promotion calendar, intake data, or media plan in one conversation, the kind of analysis that used to take an analyst a day. It pairs naturally with the operational side of AI intake for law firms: the same firms automating intake are the ones automating the analysis of it.
GA4 Attribution and Assisted Conversions
Last click lies by omission. A prospect reads your guide from an organic search in March, sees the firm's name twice more, and in April clicks a paid ad and books. Last click hands the whole case to paid search and tells you content did nothing. Multiply that by a year of budget meetings and firms end up cutting the channels that start journeys to feed the channels that end them.
GA4's 2026 Conversion Attribution Analysis, in the Advertising section, brings assisted conversions back as a first-class report. Set the model to last click to see assists honestly: for each channel, how many conversions it closed versus how many journeys it appeared in without closing. The companion funnel view sorts channel touchpoints into early, mid, and late journey stages, which is the clearest picture GA4 has ever offered of which channels open demand versus close it. Expect surprises; organic search routinely turns out to be a far bigger assist engine than its last click numbers suggest, which changes how you value the content program behind a scaling firm's growth system.
Three rules keep attribution reporting honest. Assists sum to more than 100 percent by design, so use them to rank channels, never to total them. Attribution is not incrementality: it reorders credit among observed touches, it does not prove a channel caused the case, so pair it with judgment and, at larger budgets, holdout tests. And the Advertising section requires a linked Google Ads account, one more reason the Chapter 3 links matter. Note also that GA4's reporting attribution defaults to data-driven; when your numbers differ from an agency's, the model and the attribution window are the first two places to look.
Cross-Channel Budgeting: Predictive Planning in GA4
The newest GA4 capability moves the tool from describing last month to projecting next quarter. Cross-Channel Budgeting, in the Advertising section, offers two instruments: project plans, which track whether current spend is pacing toward its projected outcome, and scenario plans, which model what happens to results if budget moves between channels. The models train on your own history and account for seasonality.
For a law firm, the honest use case is the budget meeting. Instead of arguing feelings, bring a scenario: what does the model project if $5,000 moves from paid social into paid search next quarter? Sometimes the most valuable discovery is a flat line, learning that moving 20 percent of the budget barely moves projected results, which tells you the real constraint is not allocation but intake conversion or coverage, exactly the pattern we see when firms work through how much a law firm should spend on marketing.
Cautions in proportion: these are models trained on your past, so present outputs as projections, never promises. The feature carries eligibility requirements, including linked ads accounts and sufficient conversion history, and its projections are only as honest as the cost data connected, which is why importing cost data from non-Google platforms matters for firms spreading spend across Meta or TikTok. And a projection is not a substitute for the 90-day channel reconciliation against signed cases; it is the planning layer on top of it.
The Governance Cadence and Common Mistakes
Analytics accounts rot without a calendar. The maintenance load is small, but it is not zero, and the firms with trustworthy numbers are simply the ones that scheduled these checks.
Monthly: 30 minutes
- Read Traffic Acquisition with engagement and key event columns; note the AI Assistant row and check its channel definition, because the assistant list changes.
- Reconcile GA4 key events against the CRM's lead count. A gap over 10 percent means a tracking defect or an attribution leak; find it now.
- Review the landing pages report for pages earning key events, and feed winners to the content plan.
- Annotate anything that changed: site launches, new forms, tag changes, filter edits.
Quarterly: one hour
- Run the double-counting test from Chapter 6: one test form, one test call, count the key events.
- Review the key events list against the measurement plan; demote anything that crept in.
- Check Google Ads conversion sources for duplicate systems of record.
- Review attribution and budgeting reports before the quarterly budget conversation.
And the mistakes that account for most of the damage we find in law firm properties: measuring without a plan, so every number is an orphan; two tag implementations running at once; a key events list with fifteen entries; UTM anarchy, so half of email and campaign traffic reports as Direct; call tracking bolted on without deciding the single counter; agencies grading their own homework on last click; and nobody assigned to the monthly reading, which is how a quietly broken tag goes unnoticed for two quarters. Every one of these is cheap to prevent and expensive to discover late.
A law firm does not need more data. It needs a short list of trustworthy numbers connected to signed cases, read on a schedule, by a named owner. GA4 in 2026, with the AI Assistant channel, assisted conversions, and conversational analysis, is finally a tool that rewards that discipline instead of burying it.
Want Your Numbers to Mean Something?
A clean GA4 property, honest conversion counting, and a monthly reconciliation to signed cases is infrastructure every growing firm needs. If you want this built and audited around your firm's channels, start a conversation.
Schedule a Free ConsultationFrequently Asked Questions
Is Google Analytics 4 free for law firms?
Yes. GA4 is free at the scale any law firm operates, including the 2026 features covered in this guide. The real costs are configuration time and, optionally, a call tracking platform and the specialist hours to set everything up cleanly.
What are key events in GA4?
Key events are the actions you designate as meaningful outcomes, what GA4 called conversions before the 2024 renaming. For a law firm they should be lead actions only: phone calls, contact form submissions, chats started, and consultations booked. Keeping the list short protects every report and every ad platform optimizing against it.
What is the AI Assistant channel in Google Analytics?
A default channel, live since mid-May 2026, that groups traffic arriving from AI assistants like ChatGPT under Reports, Acquisition, Traffic Acquisition, with the medium ai-assistant. It requires no setup. It understates true AI traffic because LLM app visits often land in Direct and Google counts its own AI Overviews clicks as Organic Search, so treat the row as a floor.
How much traffic should a law firm expect from ChatGPT and other AI assistants?
Typically 1 to 5 percent of sessions in 2026. The quality runs higher than the volume suggests: Adobe's March 2026 research measured 42 percent better conversion from answer engine traffic, and Shopify found 50 percent higher conversion on product pages. Judge the channel on key event rate and value per visit, not session count.
How do law firms track phone calls in GA4?
Two layers. Click-to-call taps are tracked as an event on the site. A call tracking platform with dynamic number insertion adds the calls themselves, tied to source and session, pushed into GA4 as events. Pick exactly one of the two as the key event counter for calls, or every conversation counts twice.
Why are my GA4 conversion numbers higher than the leads my intake team sees?
Almost always double counting: duplicate tags, a form event plus a thank-you page both marked as key events, enhanced measurement colliding with custom form tracking, or call tracking counted alongside click-to-call. Run a test lead end to end and count the key events it creates; the answer should be one.
How do I stop Google Ads from double counting conversions with GA4?
Give each conversion action exactly one system of record. Either import the GA4 key event into Google Ads or use a native Ads conversion tag for that action, never both. Audit the Ads conversion settings after any agency change, because duplicated sources corrupt both reporting and smart bidding.
Should a law firm use Google Tag Manager or the plain Google tag?
Either works; running both is the defect. Tag Manager suits firms with call tracking, multiple platforms, or an active marketer iterating on events. The hardcoded Google tag suits simple sites that rarely change. The site must carry exactly one implementation.
What is Ask Advisor in Google Analytics?
GA4's conversational analysis feature: a question box that answers plain-English questions like "which channels drove the most key events last month" without navigating reports. It is bounded by what GA4 contains and should be double-checked on causal questions, but it is the fastest way for non-analysts to get real answers from the tool.
Does GA4 show assisted conversions?
Yes, again. The 2026 Conversion Attribution Analysis report in the Advertising section shows, per channel, conversions closed versus journeys assisted, and sorts touchpoints into early, mid, and late stages. It requires a linked Google Ads account, and assists intentionally sum past 100 percent, so use them to rank channels rather than total them.
Can GA4 tell me what a signed case costs by channel?
Not alone. GA4 ends at the lead; the signed case lives in your CRM. The working system tags every lead with its source at intake, brings outcomes back monthly, and reconciles cost per signed case by channel. GA4 supplies the behavioral and channel detail underneath that scoreboard.
How often should a law firm review its analytics?
Monthly for 30 minutes: channels with quality columns, CRM reconciliation, landing pages, annotations. Quarterly for an hour: the double-counting test, key events review, Ads conversion sources, and attribution before budget decisions. The cadence matters more than the sophistication.