1. The firm and the diagnosis
The client is a regional plaintiff-side personal injury practice at the small-to-mid tier. Between five and fifteen attorneys, one metro plus adjacent counties, a mixed matter book leaning heavily on motor-vehicle work with a smaller book of premises, dog bite, and one focused catastrophic-injury vertical the firm had built a real reputation in. The site sat on WordPress with a respectable technical SEO baseline. Core Web Vitals were passable, canonicalization was clean, sitemap coverage was intact, the practice-area pages ranked in the top ten for the primary head terms in the metro, and the Google Business Profile hovered around the map pack on category-level queries. This was not a firm that had ignored marketing. It was a firm that had done marketing competently under an older model that had quietly stopped producing the pipeline it used to.
The pre-engagement stack looked like most firms in this tier. A retained SEO vendor on a fixed monthly fee, a Google Ads relationship inside a legal-vertical paid-media shop taking a percentage of media spend, a Google Business Profile the office manager updated when she remembered, an intake team of three trained on scripts written five years earlier, and CASEpeer as the case-management system with reporting that pulled matter counts and settlement values but did not tag source at intake in a way the marketing dashboards could reconcile. Marketing spend ran roughly $28,000 to $42,000 a month across paid search, LSA, retained SEO, a small local sponsorship line, plus a two-market cable and streaming TV buy that dialed up and down with the calendar.
The competitive landscape was the normal shape. Two national brands dominated paid search and LSA position ranking, with Morgan and Morgan running in-market at a spend level no local firm was going to match. Three local mid-market competitors held down the map pack most weeks. Two aggregator brands (a large ranking-directory site and a large lawyer-matching service) were absorbing citation share on the informational queries the whole metro's buyer set was researching. The firm had a plausible identity in the market, a strong catastrophic-injury reputation among referring attorneys, and a signed-case book that had held up better than raw inquiry volume suggested.
What was not working was the pipeline. The managing partner had watched intake calls trend down for twelve straight months while paid budget held steady and Google Ads impression share stayed inside its usual band. Signed cases were softer, cost per qualified consult was creeping, and the referring-attorney network was holding but not compensating for the drift. The instinct in the room when we walked in was that the market had gotten tougher or that the national spend had leaned in harder. Neither was quite right. Impression share on paid was intact. Rankings on the head terms were intact. The map pack position was intact. Everything the classical dashboards measured looked normal, and yet the phone was ringing less and the qualified consults sitting at the intake desk were smaller.
The real diagnosis was a discovery-layer shift. The proportion of consumers resolving early-stage legal research inside Google AI Overviews, ChatGPT, Perplexity, and Claude had crossed the threshold where a firm invisible to those surfaces started losing traffic that never showed up in the classic Google Search Console dashboards. Traditional blue-link SEO was still working. It was working on a shrinking share of the total question volume. Paid channels were not going to compensate for the erosion because paid does not solve the research-phase visibility problem that answer engines were quietly eating. The path forward was not to spend harder into the failing surface. It was to move the site's center of gravity onto the surfaces the buyer was actually using to shortlist the firm. That reframing is the whole engagement in one sentence, and the seven workstreams that follow are the operational shape of how it was executed. Readers who want the underlying framework should see AEO, GEO, and the shape of modern search and the 19 ranking surfaces, both of which the retrofit was organized against.
2. The discovery audit, in detail
Every retrofit starts with a baseline that the managing partner can hold up as evidence that the current investment is under-producing. Without that baseline the conversation is a philosophical debate about whether answer engines matter. With the baseline in the room, the conversation shifts to which workstreams ship first and how fast. Two weeks were carved out at kickoff for a discovery audit that ran across five parallel tracks: an answer-engine citation baseline, a technical SEO crawl, an entity audit across the open web, a Google Business Profile audit against the operational hygiene checklist, and a competitive citation analysis for the top four firms holding position in the metro.
The answer-engine baseline is the most novel piece and the one that reframes the room. We built a fixed prompt set of 92 queries: roughly a third head terms (jurisdiction plus practice area, jurisdiction plus injury type, jurisdiction plus catastrophic sub-vertical), a third long-tail buyer questions (statute of limitations, comparative negligence, adjuster calls, stacking policy limits, when to fire the first attorney), and a third specific sub-vertical language the firm wanted to own. Each prompt was run weekly across Google AI Overviews, ChatGPT with search, Perplexity, Claude with web access, and Gemini. Results were logged capturing citation presence, position, quotation vs source-list linking, and anchor text. The baseline showed zero citations to the firm across the full prompt set. Nolo, FindLaw, an aggregator, and two national brands were absorbing the citation share the firm should have been earning on its own home turf. That was the number that reframed the room.
The technical SEO crawl ran through Screaming Frog and Ahrefs against the full site. It surfaced a few clean-up items (orphaned pages left over from an older migration, a broken redirect chain, inconsistent canonicals on the practice-area subfolder) but nothing catastrophic. The organic performance pull from Ahrefs and Semrush confirmed the pattern: keyword-level rankings had held but organic sessions had drifted down 18 percent year over year, with the drop concentrated on informational queries and long-tail question phrases rather than head terms. That drift was the signature of an answer-engine takeover. Head terms still returned a blue-link SERP the firm could compete on. Informational queries increasingly returned an AI Overview that answered the question in-line and reduced click-through rate on every organic position below it.
The entity audit inventoried every place the firm was mentioned on the web outside its own domain: Google Business Profile, Avvo, Martindale-Hubbell, Super Lawyers, Best Lawyers, state bar directory, county bar association, state trial lawyers association, LinkedIn firm and attorney profiles, Facebook, YouTube, guest-post placements from earlier link-building work, and roughly forty local directory citations of varying quality. NAP consistency was 82 percent, a passable baseline but not the 100 percent that entity confidence models reward. Attorney bios across off-site profiles were inconsistent in bar-admission language, notable-representation formatting, and even in whether certain attorneys were listed at all. The entity picture the answer engines were reading was blurry enough to explain why the firm was not being recognized as the authoritative local source on the queries it should have owned.
The Google Business Profile audit ran against a fifteen-point operational checklist: primary category, secondary categories, service area boundaries, service list, business description, photo count and freshness, post cadence, Q&A curation, review count and rating, review response rate and response time, review recency, hours accuracy, appointment link, attributes, and messaging. The firm passed 6 of 15. Primary category was set to a general "law firm" rather than "Personal Injury Attorney." The service area was drawn too tightly and excluded three counties the firm actually accepted matters from. Posts had not been added in seven months. The Q&A had a decision-driving question about consultation fees sitting open for eleven weeks.
The competitive citation analysis pulled the same 92 prompts against the top four competing firms. Two had begun structured AEO work six to nine months earlier and were capturing early citation share on long-tail queries. The other two matched our client. The window to move first in the metro was still open but not for another year. The 34-page baseline document from these five tracks became the reference the whole engagement pointed back to.
3. The seven workstreams
The engagement ran seven parallel tracks over roughly 120 days. Each had a specific owner, a defined deliverable, and a weekly report against a shared timeline. The seven interlock: content without schema does not get cited, schema without direct-answer content does not move rankings, thought leadership without a compliance layer trips state-bar rules, local and reputation without a measurement layer produces impressions the managing partner cannot tie back to fee revenue, and none of it compounds without a discovery audit that has told everyone what the baseline actually is. Removing any one degrades the whole; sequencing them wrong costs weeks. The order below is the sequence we ship in.
Workstream one: discovery audit and citation baseline
The two-week audit described in section two opens the engagement and gets refreshed monthly for the duration. After the baseline document lands, the prompt set stays live as the ongoing measurement instrument. Each Monday the prompt set runs against the five answer surfaces, results feed into the managing-partner dashboard, and the delta from prior week gets logged. New prompts get added as the firm opens new sub-verticals. Removed prompts get archived rather than deleted, because the historical trend on a deprecated prompt is still useful when a related query starts moving.
Tooling is deliberately lightweight. Prompt runs are executed manually for the first month so the team learns what the answer engines actually return, then transitioned to a lightweight automation layer. We avoided the venture-funded AEO monitoring platforms because they over-normalize the data and hide the citation-position nuance that matters most. The deliverable each week is a one-page snapshot showing citation count by surface, position shifts, new citations gained, citations lost, and any prompt that transitioned from zero to first citation.
The methodological lesson: baseline before you touch anything. A client who cannot see the before-picture cannot see the after-picture, and a retrofit measured only in impressions will get defunded by month three when the managing partner asks what they are paying for.
Workstream two: information-architecture rewrite
The top 25 pages by traffic and by strategic importance were restructured answer-first. Every page now opens with a 60 to 90 word direct-answer summary in a visually distinct block, followed by subheads phrased as the questions consumers actually type or speak, followed by supporting content that cites named sources and includes attributable numbered facts. The old pages were competent legal marketing copy that opened with brand throat-clearing, drifted through firm history and attorney credentials, and worked down to the practical answer somewhere around scroll-depth three. The new pages get to the answer in the first paragraph and then earn the reader's attention with real substance underneath.
A concrete before-and-after helps. The old motor-vehicle-accident practice-area page opened: "For over 20 years, [Firm] has represented injured drivers and passengers throughout the [metro] area, fighting for the compensation you deserve after a serious crash. Our experienced trial attorneys understand the physical, emotional, and financial toll of a motor vehicle accident..." The new version opens: "In this state, an injured driver has [X] years from the date of the crash to file a lawsuit against the at-fault party, and the state follows a modified comparative negligence rule that bars recovery if the injured driver is found more than 50 percent at fault. Insurance-company adjusters typically call within 24 to 72 hours of the police report becoming available. What follows is a plain-English breakdown of how the process works, what to say and not say to the adjuster, and how settlement value is calculated for the injuries this firm handles most often." The old open was true. The new open is useful, and it is the kind of thing an answer engine will quote.
Keyword-to-question mapping drove the subhead structure on every rewrite. For each target page we pulled the trailing 90 days of Google Search Console query data, cross-referenced it with the People Also Ask cluster around the head term, added the questions the intake team had heard buyers ask on the phone in the last quarter, and ranked the merged list by intent value. The top 8 to 12 questions became the H2 subheads on the page. This is boring, unglamorous work, and it is the difference between a page that ranks and one that gets cited.
URL structure was preserved everywhere. Internal linking equity was retained by keeping the anchor-text patterns the site already used and layering the answer-first structure on top of the traditional SEO foundation rather than replacing it. Two pages were consolidated where the older site had split near-duplicate coverage across separate URLs, with clean 301s in place. No page URLs were changed to protect existing rankings during the transition.
Workstream three: schema and structured data
A full schema stack shipped across the site. FAQPage on every guide and practice-area page with question-answer pairs matched to the subhead structure of the rewrite. Person and Attorney schema on every attorney bio with hasCredential exposing bar admissions, law school, notable court admissions, and years in practice. Organization at the site level with sameAs pointing to the firm's Google Business Profile, LinkedIn, Avvo, Martindale-Hubbell, Super Lawyers, Best Lawyers, the state bar directory, the county bar association, and the state trial lawyers association. LocalBusiness on the office location. LegalService on each practice-area page. HowTo on the step-by-step guides. Speakable on the direct-answer summary blocks. Every schema block was validated in the rich results test before deploy.
Two schema-specific gotchas are worth naming. First, LegalService and Attorney schema overlap in ways that trip validators if both are declared at the site level. The clean pattern is Organization at the site root, LegalService on each practice-area page, and Attorney on each attorney bio, with sameAs chains that reinforce the entity graph without duplicating it. Trying to declare LegalService on the homepage in addition to Organization produces conflict warnings on some validators and inconsistent rich-result eligibility across surfaces. Second, the Article versus BlogPosting decision matters more than it seems. The firm's blog posts had been shipping as Article since the site's founding, which is technically correct but under-signals the format. We switched to BlogPosting for the running blog cadence and reserved Article for the pillar guides and news-worthy pieces (published verdicts, notable settlements with appropriate disclaimers, published byline placements from third-party publications). The shift in schema type produced a modest but consistent lift in featured-snippet eligibility on the blog posts inside 60 days.
Deployment discipline mattered more than schema selection. Every schema change shipped in staging first, got validated in the rich results test twice (once against the raw markup, once against the live-rendered page), and deployed inside a Tuesday-Wednesday window when someone on the technical team was watching Search Console. The one-time overhead of the discipline paid for itself the first time a plugin conflict tried to break the schema stack (see section four).
Workstream four: content backfill
Twelve new practice-area pages were commissioned, drafted, ethics-reviewed, and shipped over the 120-day window, plus three refreshed pillar guides that already existed in weaker form. Each page targeted a specific sub-vertical intersected with the firm's jurisdiction, ran between 2,400 and 4,200 words, and followed the answer-first structure. Topics were chosen from the citation baseline: the sub-verticals where the firm was invisible to the answer engines but had real case history and real willingness to sign matters. The content backfill was not a blog. It was a permanent expansion of the site's answer footprint across the exact question set the firm's target buyer was researching.
Topic selection followed a two-filter method. Filter one: does the firm have real case history in this sub-vertical, and is a partner willing to be quoted as the attorney author? If no, drop it, regardless of query volume. Filter two: does the baseline prompt set show competitive citations to sources the firm can plausibly out-write? If yes, prioritize. If the query already has a dominant national source (Nolo on a general procedural question, an insurance-industry publication on a claims-process question) the local firm cannot easily displace, deprioritize unless the jurisdiction wrinkle is strong. This two-filter approach kept the twelve-page backfill focused on winnable ground rather than chasing traffic in categories the firm was never going to convert.
The attorney SME workflow was the mechanism that made the backfill possible without eating attorney calendars. For each page, the editorial lead pulled a research brief covering state-specific statutes, relevant case law, and the intake-team question log for that sub-vertical. The assigned attorney marked corrections. A staff writer drafted from the corrected brief. The attorney reviewed for legal accuracy and voice, then the ethics reviewer ran the compliance checklist. Two-week cadence per page, three pages in flight at any time. Publishing hit two pages a month for the first three months and three in the final month.
Workstream five: attorney thought-leadership pipeline
Every attorney at the firm committed to two to four bylined pieces over the engagement. The pieces surfaced under the attorney's own author schema with hasCredential populated correctly, which does real work for entity signals across the answer engines. This is the workstream that broke first and got fixed hardest. The original design asked each attorney to draft the pieces themselves, and by week five we had one piece across seven attorneys because senior attorneys with active caseloads and depositions could not carve out the writing time. Detail on the fix is in section four.
The interview-to-draft workflow that eventually worked ran as follows. The editorial lead scheduled a 30-minute Zoom with the attorney on a specific case type or a specific legal question the attorney had strong opinions about. The conversation was recorded and transcribed. A staff writer converted the transcript into a publishable draft in the attorney's voice, structured for answer-first with FAQPage subheads. The draft went back to the attorney for a 15 to 20 minute review and sign-off. Total attorney load: under 60 minutes per piece from interview to publish. The workflow produced fifteen pieces in the remaining window against a target of fourteen. The lesson is not that attorneys will not write. The lesson is that the interview format converts their existing verbal fluency into publishable text without asking them to context-switch into a mode they cannot spare the time for.
Publishing cadence was one bylined piece every ten days across the roster, rotating through the attorneys so no single attorney's author page went stale. Every piece got cross-posted to the attorney's LinkedIn and to the firm's LinkedIn page. Third-party placement strategy was deliberately narrow: we pitched two national legal publications and three state bar publications, landing four placements in the 120-day window that then linked back to the attorneys' bio pages on the firm site. Third-party placements matter for entity confidence in the answer-engine graph even when their direct referral traffic is modest.
Workstream six: local and reputation stack
Google Business Profile was tuned to the operational hygiene standard. Primary category corrected from a general "law firm" to Personal Injury Attorney, secondary categories added for the sub-verticals the firm wanted to signal (Trial Attorney, Legal Services), service area redrawn to include the three counties the firm actually pursues cases from, service list expanded from four items to fifteen with practice-area-specific language, business description rewritten to include the entity signals the answer engines need. Weekly posts on case results (with the required past-result disclaimers ethics counsel signed off on), attorney and office photos refreshed with a proper photographer rather than intake-desk phone shots, and the Q&A section populated with real consumer questions and firm-authored answers.
The review-request automation was the single largest driver of the GBP lift. A two-touch flow was deployed with the case-management system: an automated text after settlement disbursement with a direct Google review link, second touch 60 days later from the case manager as a personal note rather than an automated send. Response-to-review protocol was rebuilt: every review, positive or negative, gets a substantive response within 48 hours, with the response written to communicate to the review reader (the next prospective client) rather than to the reviewer. Review response rate went from roughly 45 percent to over 95 percent inside 30 days, which produced a ranking-signal lift the local rank tracker picked up inside the following month.
Citation cleanup ran in parallel across the fifteen must-have directories with NAP consistency verified everywhere. Whitespark ran the initial audit, corrections shipped through a combination of BrightLocal-managed submissions and manual claims for the highest-authority directories. NAP consistency moved from 82 percent to 99 percent across the target directory set inside 60 days.
Workstream seven: measurement and attribution dashboard
A managing-partner dashboard was stood up that reported on signed cases, projected fee revenue, cost per qualified consult, cost per signed case, and AI citation index. CallRail was deployed with dynamic number insertion and unique numbers per channel, including a dedicated number pool for AI-source referrals inferred from browser referrer chains. GA4 events tracked phone_click, form_submit, and chat_open. The case-management system had a required first-touch source field added at intake, with an explicit "AI assistant" option the intake team was trained to probe for when the caller said they got the firm's name from ChatGPT, Perplexity, or Google's AI answer.
The dashboard consolidated four data sources: CallRail for calls, GA4 for on-site conversions, CASEpeer for signed matters and projected fee revenue, and the AI citation tracker. A Looker Studio front-end pulled all four into a single view the managing partner opened Friday mornings. The one non-negotiable: nothing on the dashboard was an impression-only vanity metric. Every top-line number rolled up to either a signed case or a projected fee-revenue number reconcilable against the case-management system.
The intake team's training on the first-touch source field was the discipline that made the dashboard credible. Without a trained probe, the "AI assistant" tag would have stayed in single digits. With it, "AI assistant" grew to the fourth-largest tagged source by month four, ahead of both Facebook and Yelp referrals, which reframed the strategic conversation with the managing partner.
4. What broke, and how we fixed it
Every honest engagement produces friction. Six specific things broke on this one, and the response to each is more useful than the promise that nothing goes wrong. The pattern across all six: the response was operational, not heroic. The ones that turned into real fires were the ones where the rehearsal had not yet been done.
1. Attorney thought leadership stalled at week five
The original design asked each attorney to draft two to four bylined pieces themselves. On paper, achievable. In practice, senior attorneys with active caseloads and depositions to prepare for could not carve out the writing time. By week five we had one piece across seven attorneys, most of it dashed off during a Sunday evening, and morale on the workstream was starting to slip. The fix was a process change, not a discipline change. We moved to a 30-minute recorded interview with each attorney on a specific case type or a specific legal question they had strong opinions about, transcribed the interview, converted the transcript into a publishable draft, sent the draft back to the attorney for sign-off and edits, and shipped. Author load dropped to under 60 minutes per piece, which was the difference between a workstream that produced fifteen pieces in the remaining window and one that would have produced two.
2. Schema plugin conflict dropped rankings for two weeks
The schema stack was implemented in a WordPress environment with Yoast SEO Premium already outputting its own JSON-LD. The new structured-data layer conflicted with Yoast's schema on two high-value practice-area pages, producing malformed JSON-LD with duplicated @graph nodes and contradictory @id fields that Google's parser dropped rather than reconciled. Rich results eligibility disappeared on those two pages, rankings fell out of the top three for a small basket of head terms, and the drop was visible in Search Console within 48 hours. Isolation took two working days: disable our schema, confirm Yoast's baseline was clean, re-enable ours in a staged rollout until the conflict re-surfaced, and pin the exact types (LegalService and FAQPage) where the collision happened. The fix was to strip Yoast's schema output on the affected page templates (Yoast supports a filter for exactly this) and let our layer handle the full JSON-LD stack. Rankings recovered inside two weeks. The lesson we operationalized: no schema change ships without a pre-flight check for conflicting output from any other plugin in the stack.
3. State-bar compliance caught an outcome-guarantee risk
One of the new practice-area pages contained a sentence describing settlement outcomes in a specific injury category that, read strictly under the state bar's advertising rules, could have been interpreted as an outcome guarantee. The specific claim was a range language pattern ("cases like these typically resolve in the [X] to [Y] range") without the sufficient disclaimer that past results do not predict future outcomes and that each case is evaluated on its own facts. The firm's ethics counsel flagged it during the pre-publish review before the page went live. The specific wording was reworded to soften from range certainty into illustrative language that keyed on the factors driving valuation rather than the valuation itself, the past-result disclaimer at the bottom of the page was expanded to sit inline near the claim as well, and we introduced a compliance pre-publish checklist that the editorial workflow now requires every new page to pass before it can go into the deploy queue. The checklist covers comparative superiority claims, outcome guarantees, testimonial disclosure, contingency-fee disclosure, and the words the state bar treats as regulated (expert, specialist, guaranteed, best). Cost of adding the checklist to the workflow: three days of process design plus four hours of ongoing overhead per page. Cost of a bar complaint on a live page: measured in months of firm attention. Trade obvious. The related discipline is laid out in the ethical AI content workflow for law firms, which the editorial team referenced throughout the engagement.
4. Attorney turnover mid-pipeline broke two author pages
Two months into the engagement, a mid-level associate who had already published three bylined pieces gave notice and left the firm within four weeks. Standard hiring reality, but it created two structural problems for the retrofit. First, the associate's author schema was live across three pages, with hasCredential and sameAs pointing at their internal bio, their LinkedIn, and their state bar profile. Removing the bio without a redirect strategy would have broken the entity graph on those pieces. Second, one of the associate's pieces was ranking well and getting cited in Perplexity for a specific long-tail query, and losing the author would risk losing the citation because entity confidence would drop. The fix was a two-part protocol we now run at every departure. The pieces were retained on the site with the departing attorney's byline preserved and a small in-line note added ("This piece was authored during the attorney's tenure at the firm") for reader transparency. The Author schema retained the departing attorney's Person node with hasCredential pointing to publicly verifiable bar records rather than internal firm pages, and sameAs updated to point to the attorney's next professional home rather than a dead LinkedIn. Rankings and citations on the affected pieces held. The protocol was documented and now runs on every departure without ad-hoc panic.
5. GBP suspension recovery pulled the operational lead for six days
Two weeks after the primary-category correction and the service-area redraw, the Google Business Profile got flagged for review by Google's automated systems and briefly suspended. Suspensions of newly edited profiles are not unusual and are not evidence of wrongdoing. They are Google's system doing exactly what it is designed to do when a profile changes categories, service areas, and contact information at once. The recovery playbook is well-known but time-consuming. We submitted the reinstatement request through the standard form within 24 hours with supporting documentation: proof of address at the office location, state bar registration proof for each attorney listed, business license, and a written explanation of the recent category correction and service-area change with the operational reason for each. Google's review took four business days. The profile came back with primary category, secondary categories, and service area intact. The operational lead lost six working days to the recovery process, which delayed the review-flow deployment by roughly the same window. The lesson we now build into every engagement: never make category, service-area, and contact-info edits in a single sitting. Space them across two to three weeks so the system re-verifies incrementally rather than flagging the profile for full review.
6. Paid media reallocation mid-engagement required stakeholder alignment
By month three, the AI citation index and the GBP call-volume lift were producing enough organic and local inquiry that the marketing spend allocation the firm had been running for eighteen months no longer matched where the pipeline was actually coming from. The paid search vendor's incentive was preservation of the existing budget line, because their fee was a percentage of media spend. The right operational decision was to reallocate roughly $8,000 to $12,000 a month from broad-match Google Ads into a mixture of increased LSA daily budget, a modest streaming TV test in the highest-value county, and the review-generation and thought-leadership retainers that were producing the pipeline. Getting to that reallocation required a direct conversation with the managing partner about the vendor incentive misalignment, a written recommendation with the projected pipeline impact by channel, and a joint working session with the vendor to restructure the engagement onto a signed-case bonus rather than a percentage-of-spend fee. The reallocation shipped in month four. Cost-per-signed-case improved another 12 to 18 percent in the following two months. The lesson: agency incentive alignment is a marketing operations issue, not an interpersonal issue, and the fix is contractual restructuring rather than difficult conversations.
5. Results after 120 days
The engagement ran 120 days from kickoff to the final review with the managing partner. Numbers below are directional, aggregated for anonymization, and consistent with the retrofit's actual output. Where the range is honest it is presented as a range. Where a number reads like a single-source hero, treat it as the pattern we saw across engagements of this shape rather than a claim about one client's exact result.
| Metric | Baseline | Day 120 | Notes |
|---|---|---|---|
| AI Overview citations, prompt set | 0 / 92 | Consistent by day 90 | Long-tail jurisdiction-plus-practice-area queries first |
| Named in ChatGPT + Perplexity | 0 / 92 | ~2/3 of tested prompts | Reached at day 120, still climbing |
| Organic CTR, primary practice-area pages | Baseline | +60 to 80 percent | Answer-first rewrite plus schema |
| GBP call volume | Baseline | Roughly 2x by month 4 | Category correction, review flow, post cadence |
| Google review count | ~140 | ~290 | Two-touch systematic ask |
| Review response rate | ~45 percent | Over 95 percent within 48h | Ranking-signal lift within 30 days |
| Qualified consult to signed retainer | Baseline | Modest lift | Intake-flow rebuild, not the headline number |
| Cost per qualified consult | Baseline | Down 35 to 45 percent | By month 5, sustained |
The AI Overview citation curve
The citation curve had a shape worth naming precisely because it is the shape most firms should expect. First measurable citation landed at day 42, on a long-tail query pairing the jurisdiction with a specific catastrophic-injury sub-vertical the firm was locally known for. By day 60 the citation count was five across the prompt set, all on long-tail queries and all sourced back to the newly published practice-area pages. By day 90 it was 34, with the curve accelerating as the entity signals from the schema stack and third-party placements began reinforcing each other. By day 120 it was 58 across the full 92-prompt set, with roughly two-thirds of tested prompts naming the firm at least once across ChatGPT, Perplexity, and Google AI Overviews. Head terms were still not producing citations at day 120 because national brands and aggregator sites were entrenched on those queries. Head-term capture lags long-tail capture by 6 to 12 months and often requires additional editorial placements that only accumulate over time.
Organic CTR page-by-page
The organic click-through rate lift on the primary practice-area pages was uneven in a useful way. The three pages that had gotten the deepest rewrite and the tightest schema layer moved 70 to 90 percent on CTR. The pages that had gotten only a partial rewrite or a schema layer without a full IA restructure moved 30 to 50 percent. Two pages did not move at all, and both turned out to be pages where the target query had migrated so heavily into an AI Overview that the entire blue-link SERP was seeing depressed CTR across every position, ours included. Those two pages produced their signed-case lift through the AI Overview citation rather than through organic CTR, which is the shape the retrofit was designed to produce. The instructive detail: measuring only organic CTR would have called those two pages under-performers. Measuring citations and consult-source together showed them as the two highest-yielding pages in the whole content backfill.
GBP call volume, attribution decomposition
The roughly 2x GBP call volume lift by month four decomposed roughly as follows. About 40 percent of the lift was attributable to the primary category correction and the service-area redraw, which made the profile eligible for map-pack impression on the actual query set the firm needed. About 30 percent was attributable to the review-flow lift, which moved the profile from a mid-4-star baseline into the 4.8+ band that Google favors in the map ranking algorithm. About 20 percent was attributable to the weekly post cadence and the Q&A section curation, both of which are user-facing engagement signals the local ranking factors have been rewarding more heavily over the past 18 months. The remaining 10 percent was attributable to the ambient lift the whole retrofit produced by making the firm more visible on the answer-engine surfaces that fed buyer-intent traffic back into the map pack via branded searches. None of this is exotic. It is what a competently run local marketing operation does every quarter. Most firms in most metros are not running the operation at that hygiene bar, which is why the doubling happens when someone finally does.
Qualified consult to signed retainer
Conversion from qualified consult to signed retainer moved modestly, not dramatically. The larger movement was in the mix of consults arriving. Pre-engagement, roughly 60 percent of qualified consults were on straightforward soft-tissue MVA matters with expected fee revenue in the lower end of the firm's book. Post-engagement, that share moved to closer to 45 percent, with the offsetting rise coming in the catastrophic-injury and complex-premises verticals the content backfill and the thought-leadership pipeline had specifically been designed to attract. Average expected fee revenue per signed matter rose accordingly. The conversion percentage on those complex matters was slightly lower than on soft-tissue MVA (buyers with catastrophic injuries shop more attorneys before signing), but the fee-weighted value of each consult that did sign was materially higher, and the firm's book of business shifted toward the sub-vertical mix the managing partner had been trying to build for two years.
Cost per qualified consult, month-over-month
The cost-per-qualified-consult curve did not move in a straight line. Months one and two showed almost no change because the retrofit had not yet shipped enough content or landed enough schema for the answer engines to start citing, and the paid budget was doing all the heavy lifting. Month three showed a modest 8 to 12 percent improvement as the local and reputation stack matured and the intake flow tightened. Month four showed the sharp break: 22 to 28 percent below baseline, driven by the AI citation lift beginning to seed inquiry volume that arrived at zero direct-media cost. Month five sustained the improvement at 35 to 45 percent below baseline, which was the number that showed up in the closeout deck. Blended across paid, organic, and referral, the improvement was durable through the following two quarters as the compounding curve described in the next section began to work.
Three things worth naming honestly about the results. First, the AI citation number is the most novel and the least mature. The firm was one of the earliest movers in its metro on structured AEO work, and the citation share captured is directly proportional to that timing. A firm running the same retrofit 18 months from now will find citation share harder to earn. Second, the GBP lift is durable and mechanical. It responds to operational discipline rather than timing, and any firm running the local-and-reputation workstream at hygiene bar will see similar mechanics. Third, the cost per qualified consult dropped because the whole discovery-to-consult chain got more efficient, not because paid got cheaper. Paid captured the compounding benefit the organic and local layers produced.
6. The compounding curve favors early movers
The reason to run this retrofit now rather than in 12 or 18 months is that answer engines are still building their preference signals. When Google's AI Overview parses a jurisdictional legal question, it decides which local sources to cite based on entity confidence, content structure, schema clarity, and cross-web signal density. Those decisions are not resampled from scratch on every query. Once the engine has decided that a particular firm is the credentialed local source for a particular query set, that preference persists and reinforces itself as the citation loop produces more incoming links, more entity mentions, and more confirmation that the engine's initial choice was correct. This is the shape of an early-mover asymmetry across every discovery platform in the history of discovery platforms, and the same mechanic is running now on the answer engines.
Firms that seed authority now become the default local citation. Firms that arrive 18 months later will need to do more work, spend more budget, and wait longer to displace the incumbent citation. This is a first-mover asymmetry, not a permanent moat, and firms already inside the citation share will need to keep the content fresh and the entity signals healthy to hold their position. But the effort required to hold a position is a fraction of the effort required to take one from the current holder, and the ambient inquiry volume the incumbent captures during the holding period funds the maintenance work several times over.
E-E-A-T signal accumulation compounds because the answer engines weight it as a decay-resistant confidence input. Bar admissions, notable representations with proper disclaimers, published bylines, third-party citations, editorial mentions in credentialed publications, and consistent Person-schema hasCredential exposure all accumulate on the entity graph over time. The cumulative effect over 18 to 36 months is the difference between an entity the answer engines treat as a credible source and one they treat as a plausible source. Credible sources get cited. Plausible sources sometimes get cited when the credible sources are not available. That distinction is the whole game.
Citation-graph reinforcement is the mechanism by which early citations produce more citations. Once an answer engine has cited a firm as the source on a specific query, that citation appears in the wild: someone screenshots it, someone quotes it in a follow-up piece, someone links to the source page in their own content because the citation validated the source's credibility. Each subsequent citation and link is fresh evidence to the answer engine that its original choice was correct. Firms that spend six months trying to break into a citation set another firm has already secured are working against a loop that keeps producing evidence they are the less-credible source.
AI training data lag is a separate mechanic that also favors early movers. Foundation models are trained on snapshots of the web that lag the live web by months to years. A firm that establishes strong entity signals and published thought leadership now will be represented in the training data of models that ship in 2027 and 2028. A firm that arrives 18 months from now will be represented in models that ship in 2029 and later. Model providers can and will fine-tune and retrieve-augment newer sources, but the base-layer training data preferences carry forward and inform behavior in ways that retrieval augmentation cannot fully override.
What this means for the firm we worked with: the retrofit is not a completed project. It is the foundation for a compounding position. The measurement dashboard runs weekly, the content backfill continues at two to three pieces a month, the thought-leadership pipeline stays in cadence, the citation index gets tracked against a growing prompt set, and the local and reputation stack runs on ongoing hygiene. Year two shifts the workload roughly 40 percent lower than year one because the foundational IA and schema work is one-time capital rather than recurring operating cost. Year three shifts another 20 percent lower as the citation graph reinforces itself.
Ranking Surfaces Playbook applied to the retrofit
7. What carries over to other practice areas
The engagement described here was a personal injury retrofit, but the framework is not personal-injury-specific. Roughly 90 percent of the mechanics carry over to every other consumer-facing practice area where a local firm competes for individual buyers on the discovery surfaces. Discovery audit, IA rewrite, schema stack, editorial pipeline, local and reputation stack, and the measurement dashboard are structurally identical. The 10 percent that changes is practice-area language, sub-vertical topic map, and the specific compliance surface each state bar and each practice-area regulator imposes. Below is how the retrofit adapts across the adjacent practice-area playbooks documented on this site.
Family law
Adjacent shape, tighter emotional register, longer research cycle. Family buyers spend 30 to 90 days quietly researching divorce, custody, and mediation options before they contact a firm, and the answer-engine surfaces are heavily weighted in that private research phase. Content backfill topics shift toward jurisdiction-specific process questions (how equitable-distribution states differ from community-property states, how the local court schedules temporary hearings, how child-support calculations work under the state's guidelines). The attorney thought-leadership pipeline gets more sensitive because the audience is in a personal crisis and tone matters more than it does on a PI page. The compliance surface is slightly different: solicitation rules on family cases have specific quirks in some states, and testimonial disclosure is more strictly scrutinized. See the family law firms playbook for the full adaptation.
Immigration
Multilingual demand changes the whole content surface. USCIS-process content in Spanish, Portuguese, Mandarin, Haitian Creole, and Arabic (in metros where the demand is present) is not optional. Schema stacks need language-scoped canonicals and hreflang for each translated page. The compliance surface is federal (immigration is federal law) rather than state, which changes which words are regulated and which advertising claims are safe. Content topics shift toward specific visa categories, USCIS processing-time realities, RFE response strategy, and asylum-process education. See the immigration law firms playbook for the full adaptation.
Workers compensation
Injured-worker buyer, denied-claim query intent, bilingual demand in most metros. The AEO layer is particularly strong for workers comp because the queries tend to be highly specific ("denied workers comp claim in [state]" or "workers comp settlement value for [injury type]") and the state-by-state variation in benefit calculation makes generic national content inadequate. Firms that build state-specific answer content dominate the citations. See the workers compensation law firms playbook.
Social Security Disability
Disabled buyer, fee-cap economics that constrain what the firm can spend on acquisition, appeals-process content strategy. SSD fee structure is federally capped at 25 percent of past-due benefits with a hard dollar cap, which makes CAC discipline tighter than in PI. AEO is highly effective for SSD because the queries around initial application, reconsideration, hearing preparation, and appeals council review are all highly structured and reward comprehensive answer content. See the Social Security Disability law firms playbook.
Criminal defense and estate planning
The adaptations to criminal defense (urgency-driven buyer, jurisdiction-specific process content, tighter compliance around anything that looks like a case-outcome prediction) and to estate planning (referral-heavy channel mix, longer sales cycle, higher-net-worth buyer, entirely different emotional register) run the same mechanical framework with different content and compliance overlays. For any of these adjacent practice areas, the retrofit runs on the same seven workstreams in the same sequence with the topic map, buyer language, and compliance checklist swapped for the practice area's specifics.
8. Method appendix
The single most useful framing for a firm considering this shape of engagement is which parts of the retrofit belong inside the firm forever and which parts are best run through an outside partner during the build and possibly beyond. What follows is the appendix material we handed to the managing partner at closeout, expanded so it can serve as a working reference for other firms.
Tool stack that supported the engagement
Deliberately conservative on tool choice. WordPress as the site platform with Yoast SEO Premium retained (with its schema output scoped down on the pages our custom layer covered) and WP Rocket handling caching and performance. A custom-built schema layer for the answer-engine-specific markup, hand-written rather than pulled from a plugin because plugin schema layers tend to over-abstract and under-cover the specific answer-engine cases. Google Search Console and GA4 as the base analytics, both instrumented against the same UTM taxonomy the CRM used. CallRail with dynamic number insertion for phone attribution, unique number pool per channel including AI-source referrals. A weekly AI citation tracking rig run against the fixed 92-prompt set across the five answer surfaces, results piped into the managing-partner dashboard through a Sheets-to-Looker Studio bridge. Ahrefs and Semrush for keyword and rank monitoring at the head-term level. Whitespark for citation audit and BrightLocal for ongoing local rank tracking and citation submission management. Screaming Frog for the quarterly technical crawl. CASEpeer as the case-management system, extended with a first-touch source field and an AI-assistant tag, and reconciled weekly against the marketing dashboard.
Substitutions worth naming. Firms on Clio Grow or Lawmatics can achieve the same source-tagging discipline with a small custom-field configuration. Firms on Litify get more powerful reporting out of the box but pay in setup complexity. Firms without CallRail budget can approximate phone attribution with a small pool of Twilio numbers, at roughly 40 percent of CallRail's monthly cost but materially higher setup and maintenance overhead.
In-house versus outside partner allocation
Keep in-house: the intake team, the case-management system, the attorney interview cadence for thought leadership, the state-bar compliance review, the signed-case attribution field at intake, and the weekly review of the managing-partner dashboard. These are the parts of the operation where the firm's own judgment is the point. An outside partner cannot answer the intake phone at 11pm on a Tuesday, cannot decide whether a specific case is worth signing, cannot know how the local judges rule on comparative negligence disputes, and cannot substitute for ethics counsel on a bar-rule question. Trying to outsource any of them creates fragile dependency and expensive rework.
Keep with an outside partner: the schema and structured-data implementation, the answer-engine citation tracking rig, the technical SEO audit cadence, the editorial layer that turns attorney interviews into publishable drafts, the ethics-reviewed content production, and the quarterly strategic review. These are specialist skills. Hiring a full-time in-house team to cover them at the mid-market firm scale is inefficient. A retained relationship with an outside partner who runs the same workstreams across multiple firms produces better craft and cheaper compounding. Above roughly $30M in fee revenue, some of these functions start making sense to bring in-house partially, with an outside partner still holding the AEO and GEO specialty.
Cost breakdown
For a small-to-mid regional PI firm the 120-day retrofit typically lands between $75,000 and $180,000 in blended outside spend, plus roughly 6 to 10 hours per week of internal time from the managing partner, the intake lead, and each participating attorney. Cost splits roughly as follows: 30 percent content production (the twelve practice-area pages, three refreshed pillar guides, and fifteen bylined pieces), 25 percent SEO and schema work (the audit, IA rewrite, schema stack, and ongoing technical hygiene), 20 percent attorney thought-leadership pipeline (interview scheduling, transcription, drafting, third-party placement), 15 percent reputation and GBP (category correction, review-flow deployment, response management, citation cleanup), and 10 percent measurement infrastructure (dashboard build, source-tagging discipline, weekly reconciliation). Ongoing maintenance from month five onward drops to a monthly retainer sized to the firm's fee revenue, typically 40 to 60 percent of the average monthly build spend.
Timeline template, weeks 1 through 16
Weeks 1 to 2: discovery audit and citation baseline. Weeks 3 to 4: IA rewrite planning and first schema deployment on top 5 pages. Weeks 5 to 6: IA rewrite across top 15 pages, schema expansion, GBP category correction and service-area redraw (spaced across the two weeks). Weeks 7 to 8: content backfill begins with two pages, first attorney interviews recorded, review flow deployed. Weeks 9 to 10: content backfill continues, first three bylined pieces publish, first monthly citation check-in. Weeks 11 to 12: mid-engagement compliance review, third-party placement pitches sent, dashboard hits its first fully reconciled monthly view. Weeks 13 to 14: paid media reallocation with vendor, streaming TV test in highest-value county. Weeks 15 to 16: final content backfill, closeout deck, transition to ongoing maintenance retainer. Two weeks of buffer handles normal slippage; add more if the firm has active trial dates or a thin attorney bench.
Vendor evaluation checklist for firms considering an outside partner
Six questions to ask any outside partner pitching an AEO retrofit. Ask to see the specific prompt-set methodology and the actual tracker they will use, not a screenshot of a marketing deck. Ask which schema types they deploy on which page templates, and ask them to explain the LegalService versus Organization overlap without notes. Ask for a written policy on state-bar compliance review and who owns the final publish authority. Ask how they handle a schema plugin conflict or a GBP suspension mid-engagement, and whether that response is documented in advance or handled ad-hoc. Ask which case-management systems they have integrated attribution with and which they have not. Ask for two references from firms in adjacent practice areas that are willing to speak to the actual outcome rather than the pitch. Any partner that hesitates on any of the six is not the right partner. The specialty is real and the partner should be able to articulate it in operational terms without a marketing filter.
KPI framework: leading, lagging, canary
Leading indicators (what to watch weekly, predicts pipeline in 4 to 8 weeks): AI citation count across the prompt set, review count growth, GBP call volume, organic sessions on the pages actively being rewritten, third-party placement acceptance rate. Lagging indicators (what to watch monthly, reflects business outcome 60 to 120 days after the leading indicator moved): qualified consult volume, cost per qualified consult, signed matter count, average expected fee revenue per matter, cost per signed case. Canary metrics (what breaks first when something is wrong): schema validation error count in Search Console, GBP profile status, LCP on mobile for the top three landing pages, intake answer-time distribution. A canary that trips does not mean the retrofit is failing. It means something upstream broke and the response window is short. Canaries get checked daily.
Compliance checklist quick-reference
Every state bar has its own advertising rules and every firm should have ethics counsel who owns the definitive read for its jurisdiction. Common items every pre-publish checklist covers: no unqualified guarantees of outcome, past-result disclaimers on any specific case reference with the standard "past results do not predict future outcomes" language, contingency-fee disclosure per the state's rule, testimonial disclosures per the state's rule, restrictions on comparative superiority claims (no "best" or "most successful" absent objectively verifiable substantiation), restrictions on the terms "expert" and "specialist" (regulated in some jurisdictions), solicitation-rule compliance especially on any direct-response format, and required attorney-of-record identification on ads. The ethical AI content workflow for law firms piece expands the discipline for AI-assisted drafting specifically.
Ongoing maintenance plan for months 4 through 12
Monthly cadence from month four onward: two to three new content pieces (mix of practice-area pages and bylined thought leadership), full prompt-set citation run with delta reporting to the managing partner, GBP post cadence held at weekly, review response rate held over 95 percent within 48 hours, entity audit refresh quarterly, technical crawl refresh quarterly. Third-party placement pitching runs at two active pitches at any time. Attorney bios get a light refresh quarterly for new bar admissions, published pieces, and notable representations. The prompt set grows at three to five new prompts per quarter as the firm's sub-vertical map evolves. Managing-partner review is monthly rather than weekly from month six onward, moving to quarterly strategic reviews with monthly dashboard emails from month nine onward. Total ongoing retainer at a well-managed maintenance rhythm typically lands at 40 to 60 percent of the peak monthly build spend, which is roughly what the paid media budget alone was costing before the retrofit reallocated the mix.
Who I am in this seat
Fifteen years in marketing, ten as CMO and Creative Director at Inkgility, working across professional-services accounts including legal, healthcare, and financial-services practices. The Ranking Surfaces framework this retrofit was organized against is documented at the 19 ranking surfaces and AEO, GEO, and the shape of modern search, with a related piece on the ethical AI content workflow for law firms that expands the compliance discipline described in section four. The category playbook this specific engagement was executed against sits at personal injury law firms, and the adjacent practice-area playbooks referenced in section seven are linked in the related section below.
9. Frequently asked questions
How long does an AEO retrofit take before a law firm sees AI Overview citations?
For a regional practice with respectable existing SEO, first citations in Google AI Overviews for long-tail jurisdiction-plus-practice-area queries typically appear in 45 to 75 days after the schema stack, direct-answer summaries, and attorney author markup ship. Consistent citation across ChatGPT, Perplexity, and Claude for the same query set follows in 90 to 120 days once the entity signals stabilize across the web. Firms that start earlier build a preference moat that late movers spend twice the budget to close.
What is the single largest lever in an answer-engine retrofit?
Rewriting the top 20 to 30 pages so each one opens with a direct-answer summary and a subhead structure phrased as the questions consumers actually ask. Schema and entity work matter, but the content shape is what makes a page eligible to be cited in the first place. A perfectly marked-up page that buries the answer under three paragraphs of throat-clearing will lose to a plainly written page with weaker schema every time.
Does the retrofit hurt existing Google organic traffic?
Not when the rewrite is done carefully. In this engagement organic click-through rate to the primary practice-area pages lifted 60 to 80 percent after the answer-first rewrite, because the same pages ranked more prominently for the long-tail question queries that the old page structure ignored. The temporary dip during the schema plugin conflict was recovered inside a two-week window with rollback and controlled redeploy.
How is answer-engine performance actually measured?
Track a fixed prompt set of 60 to 120 jurisdiction-plus-practice-area queries across Google AI Overviews, ChatGPT, Perplexity, Claude, and Gemini on a weekly cadence. Record citation presence, citation position, and quoted-passage attribution. Pair the citation index with a signed-case attribution layer in the case-management system that tags first-touch source at intake. The two together let the managing partner see AI-source referrals arriving as booked consults and signed cases, not just impressions.
What breaks during an AEO retrofit and how do you handle it?
Six real friction points across engagements of this shape. Attorney thought leadership stalls when senior lawyers cannot carve out writing time, fixed with an interview-to-draft workflow. Schema plugin conflicts on WordPress produce malformed JSON-LD, fixed by scoping the existing plugin's schema output and letting the custom layer handle the affected page templates. State-bar compliance flags outcome-guarantee wording, fixed with a pre-publish compliance checklist wired into the editorial workflow. Attorney turnover mid-pipeline breaks author schema, fixed with a documented departure protocol. GBP suspension recovery pulls the operational lead for six days, fixed by staging category and service-area edits over multiple weeks. Paid-media reallocation mid-engagement requires stakeholder alignment and vendor-fee restructuring.
What should a firm keep in-house versus outsource?
Keep in-house: the intake team, the case-management system, the attorney interview cadence for thought leadership, the state-bar compliance review, and the signed-case attribution field at intake. Keep with an outside partner: the schema and structured-data implementation, the AI-citation tracking rig, the technical SEO audit cadence, the content editorial layer that turns attorney interviews into publishable drafts, and the quarterly strategic review. The rule is simple: outsource the specialist skill, keep the client relationship and the legal judgment.
What does the retrofit cost and how is it phased?
For a small-to-mid regional PI firm the 120-day retrofit lands between $75,000 and $180,000 in blended outside spend, plus roughly 6 to 10 hours per week of internal time from the managing partner, the intake lead, and each participating attorney. Cost splits roughly 30 percent content production, 25 percent SEO and schema work, 20 percent attorney thought-leadership pipeline, 15 percent reputation and GBP, 10 percent measurement infrastructure. Ongoing maintenance from month five drops to a monthly retainer sized to the firm, typically 40 to 60 percent of the peak monthly build spend.
Does the retrofit carry over to other law-firm practice areas?
About 90 percent of the framework carries. Discovery audit, IA rewrite, schema stack, editorial pipeline, local and reputation stack, and the measurement dashboard are structurally identical across personal injury, family, immigration, workers compensation, Social Security Disability, estate planning, and criminal defense. Practice-area specifics change the buyer language, the compliance surface, and the sub-vertical topic map. The mechanics of the retrofit do not change. Adjacent playbooks on this site cover the practice-area-specific adaptations in detail.
What tools are actually required to run this properly?
WordPress or a comparable CMS the schema layer can attach to, Google Search Console and GA4 for baseline analytics, a call-tracking platform with dynamic number insertion (CallRail is the category standard, Twilio-based rigs work at higher setup overhead), a case-management system with a customizable intake source field (CASEpeer, Lawmatics, Clio Grow, Litify all support this), Ahrefs or Semrush for keyword monitoring, a local rank and citation tracker (BrightLocal or Whitespark), Screaming Frog for the technical crawl, and a lightweight prompt-set tracker for AI citations. The venture-funded AEO monitoring platforms are optional. A well-run custom sheet with structured prompt logs beats most of them for the first 12 months.
How does attribution work when a caller says they got the firm from ChatGPT?
Two-layer attribution. Layer one is the browser-referrer chain when the buyer clicks a linked source in the AI response; that pulls into GA4 and CallRail with a recognizable referring URL pattern. Layer two is the intake-team probe when the caller instead types the firm name into Google after seeing it in an AI answer. The intake script includes a direct question ("Do you remember where you first heard about our firm?") and a scripted probe when the answer is vague. The probe is what closes the attribution gap and moves the AI-assistant tag into a top-five source category.
Do we still need paid search if the retrofit is working?
Yes. Paid search covers transactional high-intent queries where the buyer is ready to call in the next hour. AEO and organic cover research-phase and shortlist-phase queries that seed the eventual transactional query. The retrofit rarely reduces paid budget below 40 to 55 percent of the pre-engagement level in year one. What changes is the mix inside the paid budget (more LSA and less broad-match Google Ads) and the efficiency of the paid dollars because the site converts intent traffic at a higher rate.
How does this differ for a firm that already has an in-house marketing team?
The workstream shape is identical. The build partner scope narrows to specialist workstreams (schema, AEO citation tracking, editorial layer, compliance-reviewed content production) while the in-house team owns everything client-facing. The engagement is typically 40 to 60 percent shorter in outside-time terms and lands at 50 to 70 percent of the cost. The strategic risk is under-investment in specialist workstreams the team is not experienced enough to run alone; the mitigation is a clear scope-of-work with the outside partner.
Can a solo or two-attorney firm run a scaled-down version?
Yes, with adjusted expectations. Drop the attorney thought-leadership pipeline, simplify the content backfill to five or six pages, keep the GBP and review workstream in full, run a manual monthly prompt-set check. Budget lands closer to $18,000 to $40,000 for the initial build. Time to first AI citations is 90 to 150 days rather than 45 to 75. The mechanics work; the ceiling is lower and the timeline is longer.
What happens if a firm stops the retrofit at month four instead of continuing maintenance?
Citations captured during the build hold for roughly two to three quarters before erosion becomes visible, as long as the content stays live and the schema stack does not break. GBP and review lift persist at slightly degraded levels. The organic CTR lift persists because on-page rewrites are permanent capital. What disappears is the compounding: no new citations get seeded, and the share the firm captured begins to leak to whichever competitors are still running maintenance. Firms that stop generally regret it inside six months and re-engage at higher cost.
Is answer-engine citation share worth pursuing in low-competition metros?
Yes, and often at higher ROI. Low-competition metros have less established citation share for the answer engines to default to, so a well-executed retrofit can capture near-total citation share on jurisdiction-specific queries with materially less investment. Absolute inquiry volume is smaller, but cost per acquired inquiry is often the lowest of any channel in the firm's mix. Do not defer the retrofit because the market feels small; the opposite pattern usually applies.
If your firm is trying to move any of these levers and the pipeline is not converting the way it used to, tell me what you are working on.
Start a conversation