01
Schema markup gap
Every cited competitor had Article or BlogPosting schema on their long-form content. Her posts had no structured data — the engines had nothing to attach the content to.
Case Study / 03 — Fershteyn Law
How we made a Brooklyn estate planning firm citable by AI search engines — in 8 weeks.
Law Office of Inna Fershteyn, P.C. is one of Brooklyn’s established estate planning, asset protection, and elder law practices. Before our engagement, the firm appeared in AI search responses on Gemini, Perplexity, and ChatGPT roughly 16% of the time when users asked about estate planning topics relevant to their practice. The first two weeks of the AIO/GEO engagement pushed that number to 27.8%. By week eight it had reached 62% — a 3.9× lift, with the largest gains on the queries that drive consultations.
16.1 → 62%
AI citation rate
600+
Posts schema-marked
3
AI engines monitored
8 wks
Engagement length
The system in motion
Twelve AI engines. One firm, cited.
Citations from Gemini, Perplexity, ChatGPT and nine more — each LLM beams into the firm entity once the schema, author chain, and crawler allowlist are in place.
Before
A 30-year practice that AI engines barely knew existed.
Inna Fershteyn has practiced estate planning in Brooklyn for over three decades. Her firm has clients across Brooklyn, Manhattan, Queens, and the broader Russian-speaking community in NYC. Her website was traditional-SEO-optimized — ranking competitively for “Brooklyn estate planning attorney” and similar Google queries — but largely invisible to AI search engines. The structured data wasn’t there. The author signals weren’t there. The crawler whitelist wasn’t there.
When we tested 30 estate-planning queries on Gemini, Perplexity, and ChatGPT in mid-engagement preparation, her firm was cited in 16.1% of them. Her direct competitors — smaller firms, less seasoned attorneys — were cited in 40 to 60%.
The methodology
We tested 30 query patterns across 3 engines, and looked at every citation.
The first step was a baseline AI-visibility audit. We selected 30 query patterns covering her practice areas: estate planning, asset protection, Medicaid planning, elder law, and trusts. For each query, we logged whether her firm was cited, which competitors were cited instead, and what those competitors had structurally that she didn’t.
The audit produced three gaps — consistent across every cited competitor.
02
Author signal gap
Cited content consistently had author Person schema with credentials linked to a LegalService entity. Her content was anonymous from a structured-data perspective.
03
Crawler access gap
Her robots.txt actively disallowed several AI crawlers that her competitors explicitly allowed — the bots couldn't even read the content if they wanted to.
The insight
English queries and Russian queries surfaced completely different competitor sets.
The most surprising finding wasn’t about her — it was about the market. When we tested the same 30 query patterns translated into Russian, the competitor set changed almost entirely. Three or four bilingual practices appeared in both Russian and English citations. The remaining 80% of the citation surface was monolingual.
For a firm that already serves a Russian-speaking client base, this is a strategic opening: significantly less competition for AI citations in Russian-language estate planning queries, and any firm that can credibly serve both languages can dominate a niche the English-only competitors can’t touch.
Citation surface · EN vs RU
Same 30 query patterns, tested in English and again in Russian. Each dot is one firm cited by at least one of the three AI engines. Same firms in both panels = bilingual practices. RU panel shows the open lane.
- Law Office of Inna Fershteynboth languages
- Other bilingual firms3 firms
- Monolingual competitorEN-only or RU-only
The asymmetry
The same query patterns translated into Russian return a much shorter citation list. The dominant English-language competitors don’t appear in the Russian results — only the 4 bilingual practices (including Fershteyn Law) bridge both languages. Russian-language estate planning content is a strategic gap her competitors fundamentally can’t close.
The schema
Roughly 600 posts received structured data. The site as a whole received entity definitions.
The single largest deployment was schema markup. Approximately 600 posts across the firm’s blog and resource library received BlogPosting and Article schema mapping the author, publish date, modified date, headline, and primary entity (the LegalService for the firm). The homepage received LegalService schema linked to a Person schema for Inna Fershteyn. Three FAQPage schemas were added to the highest-traffic landing pages.
Schema entity graph
Article posts link to a single Person; the Person and the Article both link to one LegalService. Three @types, one canonical graph.
Every post links here
~600 blog posts each carry author=Person and publisher=LegalService, so an engine reading any single article can trace back to a credentialed expert at a real firm.
Single Person entity
One canonical Inna Fershteyn entity with jobTitle, knowsAbout, and worksFor links — deduplicates the author across 600 posts.
Single LegalService
The firm anchors the graph. AreaServed and serviceType give engines explicit jurisdictional + practice-area signal.
sampleView deployed Article schema (anonymized URLs)
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Medicaid Asset Protection Trusts in New York",
"datePublished": "2026-02-14",
"dateModified": "2026-04-03",
"inLanguage": "en",
"author": {
"@type": "Person",
"name": "Inna Fershteyn",
"jobTitle": "Estate Planning Attorney",
"url": "https://brooklyntrustandwill.com/about",
"worksFor": { "@id": "#legal-service" }
},
"publisher": {
"@type": "LegalService",
"@id": "#legal-service",
"name": "Law Office of Inna Fershteyn, P.C.",
"url": "https://brooklyntrustandwill.com/",
"areaServed": "New York"
},
"mainEntityOfPage": "https://brooklyntrustandwill.com/medicaid-asset-protection"
}The signals
Author identity, attached to the firm, attached to thirty years of practice.
AI engines weight Experience, Expertise, Authoritativeness, and Trust (E-E-A-T) signals heavily — especially in YMYL (Your Money or Your Life) categories like estate planning. We deployed a unified author bio block via WPCode that appears on every long-form article, links to a centralized author page with credentials, and surfaces the Person → LegalService → publisher schema chain. AI engines reading the markup can now trace any article back to a credentialed individual at a real licensed legal practice.
The infrastructure
robots.txt and llms.txt — saying yes to the crawlers that matter.
Default WordPress installations frequently disallow useful crawlers via legacy robots.txt entries. We rewrote it to explicitly allow GPTBot, PerplexityBot, Google-Extended, and ClaudeBot, while preserving blocks for non-search crawlers (SEMRush, AhrefsBot, etc.) that the firm doesn’t benefit from feeding. We also deployed an llms.txt file — a community-driven standard that guides LLM crawlers to the most authoritative content on a site. Together, the crawlers responsible for AI citations now have a clear, well-marked path to her highest-quality content.
robots.txt · before vs after
The site shipped with default WordPress robots.txt actively blocking the bots responsible for AI citations. The rewrite flipped Disallow to Allow for the four engines we care about.
robots.txt · beforeblocking
# Default WordPress robots.txt
User-agent: GPTBot
Disallow: /
User-agent: PerplexityBot
Disallow: /
User-agent: ClaudeBot
Disallow: /
User-agent: Google-Extended
Disallow: /
robots.txt · afterallowing
# Welcome the bots that drive citations.
User-agent: GPTBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: Google-Extended
Allow: /
Before
Four Disallow directives meant the engines literally could not read the firm’s content even if they wanted to. Citation rate stuck below baseline regardless of how good the underlying copy was.
After
Four Allow directives + an llms.txt file pointing crawlers at the highest-authority pages. Within W1 of the engagement, GPTBot, PerplexityBot, ClaudeBot, and Google-Extended could index everything.
cfgView robots.txt (excerpt)
# Welcome the bots that drive citations. User-agent: GPTBot Allow: / User-agent: PerplexityBot Allow: / User-agent: ClaudeBot Allow: / User-agent: Google-Extended Allow: / # Block the crawlers that just scrape. User-agent: SemrushBot Disallow: / User-agent: AhrefsBot Disallow: / Sitemap: https://brooklyntrustandwill.com/sitemap.xml
The next layer
A Russian-language estate planning authority page, designed to dominate a niche her competitors can’t enter.
Building on the asymmetry insight, we recommended the firm publish a comprehensive Russian-language pillar page covering estate planning for the Russian-speaking NYC community. This is a content gap her competitors fundamentally cannot fill — most have no Russian-language capability at all, and the few that do haven’t built authority pages on the topic. This pillar is currently in draft and goes live in the next engagement phase.
The result
Citation rate nearly quadrupled. The queries that drive consultations now cite her first.
The first two weeks moved fast — crawler whitelist landed W1, the first wave of schema markup went out across W2, and citation rate climbed from 16.1% baseline to 27.8% almost immediately. Compounding signals (full schema deployment, E-E-A-T author signaling, the first authority pillar) kept lifting the curve through the rest of the engagement. By the W8 follow-up audit, average citation rate across Gemini, Perplexity, and ChatGPT had reached 62% — a 3.9× lift over the baseline. On the highest-converting query patterns, her firm appears in the top three cited sources on all three engines we monitored.
Baseline · W0
16.1%
Week 8 · follow-up audit
62.0%
Relative lift
3.9×
Citation rate · weekly
30 query patterns · 3 AI engines · 8-week engagement
- Gemini
- Perplexity
- ChatGPT
- Average
W8 endpoints
- Gemini68.1%
- Perplexity60.0%
- ChatGPT57.9%
Initial bump · W0–W2
Crawler whitelist (robots.txt + llms.txt) landed W1 and the first wave of schema markup went out W2 — the citation rate nearly doubled in two weeks.
Compound climb · W2–W8
Full schema deployment (~600 posts), E-E-A-T author signaling, and the first authority pillar rolled out across W3–W7. Compounding signals continued to lift citations through the end of the engagement.
| Query pattern | Before | After |
|---|---|---|
| estate planning attorney Brooklyn | Not cited | Top 3 on all 3 engines |
| Medicaid asset protection NY | Cited 1 of 3 | Top 3 on all 3 engines |
| elder law Russian-speaking attorney | Not cited | Top 1 on all 3 engines |
| irrevocable trust New York | Cited 1 of 3 | Cited 3 of 3 |
The infrastructure is in place. The bilingual pillar is next.
The pattern
Regulated professional services with deep expertise and weak structured signaling — this works for all of them.
The Fershteyn engagement is the case study, but the architecture transplants. Any professional service practice with deep authority, long history, real credentials, and weak structured-data deployment can run the same playbook. The playbook isn’t legal-specific — it’s a structured-data and authority-signaling layer that sits underneath any expertise-led business.
- Independent law practices in any specialty — immigration, tax, family, intellectual property
- Boutique financial advisory and tax practices
- Medical specialty practices and specialized clinics
- Architecture, engineering, and licensed design practices
- Any business where AI engines should be citing the expert, not a directory
Visit the firm
Law Office of Inna Fershteyn, P.C.
Brooklyn estate planning, asset protection, and elder law — thirty years of practice, now backed by an AI-citation infrastructure designed to surface that expertise where prospects actually ask the question.
Next step
Want this for your firm?
An AI-visibility audit across Gemini, Perplexity, and ChatGPT — and a concrete recommendation for what would move citation rate the most. 15-minute fit call.