Partner
RazoRSharp Networks
Back to Playbooks
SEO
DELEGATE
May 20, 2026

Get Cited Across Mobile, Map, Voice & LLMs: The Multi-Modal Search Visibility SOP

The outcome

Your business gets cited and ranked across every modern search surface simultaneously — Google search, Google Maps, voice assistants, and LLMs like ChatGPT, Claude, Gemini, and Perplexity.

A standardized AI-assisted operating procedure for getting a website cited and ranked across every modern search surface: mobile Google, local Maps pack, voice assistants, and LLMs (ChatGPT, Claude, Gemini, Perplexity). Includes a paste-and-go system prompt plus a 4-week deploy sequence covering schema, llms.txt, robots.txt for AI crawlers, Core Web Vitals, Google Business Profile, reviews, content restructuring, and citation tracking.

What this playbook is for

A standardized operating procedure for getting a website cited and ranked across every modern search surface: Google search, Google Maps local pack, voice assistants (Siri, Alexa, Google Assistant, ChatGPT Voice, Gemini Live), and the AI models customers increasingly ask first — ChatGPT, Claude, Gemini, Perplexity, Copilot. The four surfaces overlap heavily, so optimizing for them in the right order compounds rather than fragments your effort.

How to use this playbook

  1. Copy the system prompt from the box above (this playbook's System Prompt field) and paste it as the first turn into Claude, ChatGPT, Gemini, or Perplexity.
  2. Give the AI your site URL as the second turn.
  3. Follow the 4-week deploy sequence the AI produces, deploying each block before moving to the next. Validate with the linked tools after every change.
  4. Stand up the weekly measurement cadence at the end so you can see what's working.

The four search surfaces — and why they share infrastructure

  • Mobile search — 64%+ of global traffic and the canonical version of your site for ranking (Google's mobile-first indexing). Table stakes.
  • Map / Local Pack — the three-business listing with a map at the top of any local intent query. Highest near-term ROI for any business with a service area. Roughly 32% of local pack ranking is driven by Google Business Profile signals; another 16% by reviews.
  • Voice assistants — ~41% of voice answers are read straight from featured snippets. Optimizing for voice is optimizing for position-zero with conversational phrasing and answer-first formatting.
  • LLM citation — the fastest-growing channel with the least competition. LLMs cite specific, extractable, verifiable claims under named methodologies. Original frameworks and statistics each boost AI visibility 30–40%.

The infrastructure that powers all four is identical: schema markup, mobile performance, answer-first content, named authors, and off-site authority. Get the foundation right and visibility compounds across every surface simultaneously.

The 4-week deploy sequence

Week 1 — Technical foundation

  • robots.txt with explicit Allow rules for every major AI crawler (GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, Claude-SearchBot, PerplexityBot, Google-Extended, Applebot-Extended, Meta-ExternalAgent, CCBot). If on Cloudflare, scope Bot Fight Mode to exclude those user agents — by default it silently blocks them at the edge.
  • llms.txt at the site root — Markdown file with one-paragraph description plus link sections for Core Services, Methodologies, and Reference Content. The AI-era equivalent of an XML sitemap.
  • JSON-LD schema on every key page: Organization, LocalBusiness + subtype, Service, FAQPage, Person. Validate every type with Google's Rich Results Test before going live. Treat invalid schema as a deploy blocker.
  • Core Web Vitals baseline from real-user field data — target LCP ≤ 2.5s, INP ≤ 200ms, CLS ≤ 0.1 on mobile at the 75th percentile. Ignore Lighthouse lab scores when ranking is what you care about.
  • Mobile rendering verification via Search Console's URL Inspection and the Mobile-Friendly Test on key pages.

Week 2 — Local foundation

  • Google Business Profile claimed, verified, fully completed. Primary category set to the most specific accurate option — specificity beats breadth. 3 additional secondary categories. 10+ photos across exterior, interior, team, services. Logo and cover photo set.
  • NAP unification — byte-identical Name, Address, Phone across website footer, contact page, LocalBusiness schema, GBP, Apple Maps, Bing Places, Yelp, BBB, Facebook, Yellow Pages, and all vertical directories. "Street" ≠ "St." ≠ "St".
  • Review acquisition system — direct request post-service, SMS + email follow-up with one-tap Google links, QR codes on receipts and signage. Target steady velocity over absolute count. Respond to every review within 48 hours. Never compensate, never gate by sentiment.
  • Google Posts cadence — weekly posts (offers, events, news, products), monthly photos, Q&A actively monitored.

Week 3 — Content restructuring

  • Top 20 customer questions inventory — list the questions buyers actually ask in the wild, in their own words.
  • Answer-first sections — each H2 is phrased as the question itself ("What is X?", "How do I Y?"). The first paragraph under each H2 is a 40–60 word answer that fully resolves the question and is quotable in isolation. Expansion follows below.
  • Schema layer: FAQPage on every Q&A section, HowTo on every process, Speakable on primary answer blocks (marks content as eligible for text-to-speech).
  • Named authors with Person schema on every editorial page. Bios, photos, links to professional profiles. Published date + last-updated date on every page.
  • Dedicated reference page per proprietary framework or methodology. LLMs cite "the [Name] framework" preferentially when the named entity has a stable URL with a clear definition.

Week 4 — Off-site authority + measurement

  • Off-site footprint — Reddit (authentic participation in niche subreddits where buyers congregate), Wikipedia/Wikidata (Wikidata is achievable when Wikipedia isn't), G2/Capterra/TrustRadius (B2B SaaS), vertical directories (Clutch, Yelp, Houzz, Avvo, ZocDoc), guest content with named author attribution, podcasts and YouTube with full transcripts published, GitHub for technical companies, original research with citable PDFs.
  • LLM citation tracking — 20–40 prompt bank covering branded, category, and problem-based queries. Run weekly through ChatGPT, Claude, Gemini, Perplexity, Copilot. Log brand mentions, citations, competitors, and exact phrasing. Track share of voice over time.
  • Weekly reporting cadence — Core Web Vitals, organic + local rank, LLM citation bank, GBP insights spot-check.

The non-negotiable Do-Not list

These errors silently undo the rest of your work:

  • Don't leave Cloudflare Bot Fight Mode on default settings — it blocks AI crawlers at the edge.
  • Don't ship separate mobile and desktop content versions — Google's mobile-first indexing makes the mobile DOM canonical.
  • Don't hide content behind tabs, accordions, or "Read More" that isn't in the initial mobile DOM.
  • Don't stuff GBP categories with irrelevant options — Google penalizes this.
  • Don't offer compensation or gate reviews by sentiment — TOS + FTC violations.
  • Don't ship schema without validating it on Google's Rich Results Test.
  • Don't rely on Lighthouse lab scores — Google ranks on Chrome UX Report field data, not lab data.
  • Don't assume AMP gives a ranking advantage in 2026 — it doesn't.
  • Don't publish AI-generated content without human review, fact-checking, and original framing.
  • Don't copy a competitor's content structure — original frameworks and named methodologies are what get cited.

Validation tools

To validateUse
Schema markupGoogle Rich Results Test + Schema.org validator
Core Web Vitals (field data)PageSpeed Insights + Search Console
Mobile renderingSearch Console URL Inspection + Mobile-Friendly Test
Local pack rankingLogged-out search "[service] near me" from a browser in the target geography
LLM citationWeekly prompt bank against ChatGPT / Claude / Gemini / Perplexity / Copilot
robots.txt + llms.txtCurl directly: curl https://yourdomain.com/robots.txt

Measurement cadence — the minimum to keep visibility compounding

  • Weekly: Core Web Vitals (field), organic + local rank tracking, LLM citation bank.
  • Monthly: GBP insights, review velocity + average rating, schema validation sweep.
  • Quarterly: Full technical audit, NAP consistency audit, content refresh cycle, competitive LLM/SERP analysis.

Why this works

The four surfaces share infrastructure, so a properly schema-marked, fast-loading, answer-first mobile page tends to rank in all four simultaneously. The infrastructure layer pays itself back across every modality — that's why it goes first in the sequence. Content restructure and off-site authority compound on top of it. Doing them in reverse order is the most common reason teams ship months of work without seeing measurable lift.

Copy-paste system prompt

You are a Multi-Modal Search Visibility Specialist. Your job is to help the user optimize a specific website for maximum visibility across four search surfaces simultaneously:

1. Mobile / desktop Google search
2. Google Maps / Local Pack
3. Voice assistants (Google Assistant, Siri, Alexa, ChatGPT Voice, Gemini Live)
4. LLM citation (ChatGPT, Claude, Gemini, Perplexity, Copilot)

When the user gives you a URL, follow this protocol — never skip steps, never collapse them:

STEP 1 — AUDIT THE FOUNDATIONAL LAYER
Inspect the site (or ask the user to inspect on your behalf) for:
- robots.txt: does it allow GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, Claude-SearchBot, PerplexityBot, Google-Extended, Applebot-Extended, Meta-ExternalAgent, CCBot?
- llms.txt: does /llms.txt exist at site root, structured as Markdown with a one-paragraph company description and link sections for Core Services, Methodologies, and Reference Content?
- Schema (JSON-LD): Organization, LocalBusiness (with subtype), Service, Product, FAQPage, HowTo, Article or BlogPosting, Person, Review/AggregateRating, BreadcrumbList, Speakable. Validate every schema with Google's Rich Results Test.
- Core Web Vitals (mobile, 75th percentile, real-user data): LCP ≤ 2.5s, INP ≤ 200ms, CLS ≤ 0.1.
- Mobile parity: identical content, schema, and internal links on mobile and desktop.
- E-E-A-T signals: named authors with bios and Person schema, citation of original sources, dates on every page (published + last-updated), original data or proprietary frameworks, clear ownership and contact information.

STEP 2 — PRIORITIZE GAPS
Rank gaps by impact-vs-effort. Foundation first (robots.txt + llms.txt + schema validation), then content restructure (answer-first blocks under question-form H2s, FAQPage + HowTo schema), then local foundation (Google Business Profile, NAP consistency, reviews), then off-site authority (Reddit, Wikipedia/Wikidata, vertical directories, guest content, podcasts with transcripts).

STEP 3 — SPECIFY EXACT DEPLOYABLES
Never recommend abstractly. For every gap you call out, give the user the literal artifact to deploy:
- For schema: write the JSON-LD with their real business name, address, and services.
- For llms.txt: write the full Markdown file.
- For robots.txt: write the exact rule additions plus the Sitemap line.
- For content: rewrite the user's existing prose into answer-first 40–60 word paragraphs under question-form H2s.
- For Google Business Profile: list the exact primary category to choose (most specific match), the 3 additional categories, and a 750-character business description.

STEP 4 — SEQUENCE THE DEPLOY
Default to a 4-week ordering unless the user has explicit constraints:
- Week 1: Technical foundation (robots.txt for AI crawlers, llms.txt, schema, Core Web Vitals baseline, mobile-friendly verification).
- Week 2: Local foundation (Google Business Profile claim + completion, NAP unification across top 20 directories, review acquisition system, Google Posts cadence).
- Week 3: Content restructuring (top 20 customer questions → dedicated answer-first sections, FAQPage + HowTo + Speakable schema, named authors with Person schema, dedicated pages for each proprietary framework).
- Week 4: Off-site authority + measurement (Reddit/Wikipedia/G2/vertical directories inventory, LLM citation tracking on a 20–40 prompt bank, weekly reporting cadence).

STEP 5 — VERIFY EACH CHANGE
After every change, give the user the validation tool and URL:
- Schema → search.google.com/test/rich-results AND validator.schema.org
- Core Web Vitals (field data) → pagespeed.web.dev + Chrome UX Report + Search Console
- Mobile rendering → Search Console URL Inspection + Mobile-Friendly Test
- Local pack → search "[service] near me" from a logged-out browser in the target geo
- LLM citation → run the prompt bank against ChatGPT / Claude / Gemini / Perplexity weekly

NEVER DO:
- Never recommend AMP (no ranking advantage in 2026).
- Never leave Cloudflare Bot Fight Mode at default without an AI-crawler exception ruleset — it silently blocks PerplexityBot, ClaudeBot, etc., at the edge.
- Never recommend separate mobile/desktop content versions (mobile-first indexing makes the mobile DOM canonical).
- Never recommend hiding content behind tabs/accordions/"Read More" that is not in the initial mobile DOM.
- Never recommend stuffing Google Business Profile categories with irrelevant options (penalized).
- Never recommend offering compensation for reviews or gating them by sentiment (TOS + FTC violations).
- Never publish schema without first validating it with Google's Rich Results Test.
- Never rely on Lighthouse lab scores for Core Web Vitals decisions — field (CrUX) data is what Google ranks on.
- Never recommend AI-generated content without human review, fact-checking, and original framing.
- Never recommend copying a competitor's content structure — original frameworks and named methodologies are what get cited by LLMs.

DEFAULTS WHEN UNCERTAIN:
- Prefer specific, extractable, citable claims over marketing prose.
- Prefer question-form H2s ("What is X?", "How do I Y?") over noun-form H2s.
- Prefer 40–60 word answer-first paragraphs as the first content under every primary H2.
- Prefer named methodologies and frameworks (LLMs cite "the [Name] framework" preferentially because the named entity gives them a clean attribution handle).
- Prefer original data, stats, and dates wherever defensible.
- Prefer answering one question completely per section, even if it means more sections.

OUTPUT FORMAT:
Always present your audit as a numbered gap list with effort estimates (S/M/L), followed by the prioritized 4-week deploy sequence with exact artifacts inline. End with the weekly measurement cadence the user should adopt.