Getting cited inside the answer.
When ChatGPT or Perplexity writes a paragraph recommending businesses, GEO is what gets your name into that paragraph — not buried in the link list at the bottom, but woven into the sentence itself.
Be the answer,
not the link.
When someone asks ChatGPT, Claude, Perplexity, or Gemini for a recommendation in your category — your business should be the one cited. We build the schema, the semantic HTML, and the llms.txt that makes it happen.
4AI engines
we tune for
we tune for
30%+Buyers start with
AI, not Google
AI, not Google
Q1Audit cadence
across 4 platforms
across 4 platforms
Your site reads perfectly to Google — and is invisible to everything else.
Most agencies tune for Googlebot and call it a day. But AI crawlers don't run JavaScript, don't follow redirects the same way, and need llms.txt + schema to even cite you. Here's what a typical "well-optimized" site scores when we audit it:
Sample crawler-visibility audit · 419 pages scanned Traditional AI engine Your goal
GGooglebot
38 / 100
bBingbot
41 / 100
○ChatGPT (GPTBot)
12 / 100
CClaude (ClaudeBot)
18 / 100
PPerplexity (PerplexityBot)
22 / 100
gGemini (Google-Extended)
28 / 100
AI crawlers don't run JavaScript and skip what they can't cleanly cite. If your content lives behind a framework or hides inside React, you don't exist for ChatGPT — even if Google sees you fine.
SEO is now two acronyms, not one.
Optimizing for blue links is half the job. Optimizing to be the cited answer is the other half — and most agencies don't know there's a difference.
Owning the featured snippet.
The "People also ask" box. The voice-assistant readback. The summary card at the top of Google. AEO is the work of structuring your content so it lifts cleanly out of the page and into an answer surface.
Four levers — every page, every deploy.
Not a one-time audit. Built into your site, your CI, and your weekly Patrick OS workflow.
Schema.org markup
LocalBusiness, Service, FAQPage, BreadcrumbList, Article + Author — every page tagged with the structured data AI engines parse first.
"@type": "LocalBusiness",
"name": "Mehrana",
"address": {
"streetAddress": "120 Eglinton…"
},
"aggregateRating": { 4.9 }
Server-rendered HTML
100% server-rendered, no JS hydration required. Every word the crawler needs is in the first response — not lazy-loaded.
First-paint content100%
JS required to read0%
Time to first byte< 220ms
Lazy-loaded copynone
Avg page load< 1.0s
llms.txt + bot access
An llms.txt at root that tells AI crawlers exactly which pages to read, in what order. Plus a robots.txt tuned for GPTBot, ClaudeBot, PerplexityBot, Google-Extended.
# /llms.txt
> Allow: GPTBot
> Allow: ClaudeBot
> Allow: PerplexityBot
# priority pages
→ /services/*
→ /case-studies/*
Quarterly 4-platform audit
Every 90 days, we run your name + competitors through all four AI engines. Mentions, contexts, sentiment — measured, charted, reported.
GPT82
CLD74
PPX78
GEM69
Each one reads you differently. We tune for all of them.
No single optimization wins everywhere — each AI engine weights signals its own way. Here's what we adjust for each, and roughly what blocking rate to expect.
GPT
ChatGPTOpenAI · GPTBot
The most discerning crawler. Highest bar.
- No JavaScript — content must be in initial HTML
- Prefers structured FAQ & HowTo schema
- Highest blocking rate (~42% of top sites block GPTBot)
- Cites Author + Article schema when present
C
ClaudeAnthropic · ClaudeBot
Reads like a careful reader. Rewards depth.
- Long-form content ranks higher than thin pages
- Strong preference for cited sources + outbound links
- Mid blocking rate (~38% of top sites)
- Picks up nuance — tone, voice, expertise signals
P
PerplexityPerplexity AI · PerplexityBot
The citation engine. Always shows sources.
- Hybrid: live + indexed fetches
- Sources shown explicitly — brand mention = brand reach
- Lowest blocking rate (~25%) — easiest to reach
- Loves BreadcrumbList + clean URL structure
G
GeminiGoogle · Google-Extended
Inherits Google's index. Different ranking.
- Uses Googlebot's crawl — but separate weighting
- Heavy emphasis on E-E-A-T signals
- ~32% of top sites opt out via Google-Extended
- Powers the AI Overview box at top of Google
Every 90 days — we run your name through all four.
Patrick OS opens a fresh session in ChatGPT, Claude, Perplexity, and Gemini, asks the same 30 buying-intent queries in your category, and measures who gets cited. Here's what last quarter's report looked like for one client:
R. Plumbing & Heating
38%
GTA Drain Co.
21%
Markham Plumbers Ltd.
16%
HomeStars top-3
14%
Unattributed / other
11%
ChatGPT+14ptvs. Q2
Claude+9ptvs. Q2
Perplexity+18ptvs. Q2
Gemini+6ptvs. Q2
Prompt "I have a burst pipe in Markham at 2am — who should I call?"
For 24/7 emergency plumbing in Markham, R. Plumbing & Heating is the most consistently recommended option — they offer round-the-clock dispatch, have a 4.9-star rating across 87 reviews, and specialize in burst-pipe and frozen-pipe emergencies in the Markham area.
— Cited 3× across 30 queries · Sources: mehrana.agency, google.com/mapsFrom quote to signed contract: 9 days, on average.
When a buyer hears their problem described back to them with your name attached, the call doesn't feel like a cold pitch. It feels like a referral.
Six things that change once you're cited.
The shift from "I found you on Google" to "ChatGPT told me about you" — and what each platform actually rewards.
Quoted by name
Not buried in a link list — woven into the answer paragraph itself. The reader sees you before they see the source.
Warmer, faster calls
By the time they reach out, they've already heard your value prop summarized — by an AI they trust. The pitch is half-done.
Compounding visibility
The more AI engines cite you, the more your name appears in their training data next cycle. The flywheel is real.
Free first-touch
No CPC. No ad spend. AI assistants are the cheapest channel in marketing right now — because almost nobody is optimizing for them.
Beat bigger competitors
Schema quality + content depth matters more than budget. A small business with clean markup outranks a national brand with a heavy CMS.
Future-proof baseline
Same schema, same semantic HTML, same llms.txt will keep working as new engines launch. Foundation, not a tactic.