GEO Guide: How to Get ChatGPT and Perplexity to Cite Your Content

Your SEO is decent, your Google rankings are fine, but when you ask a friend to "search this topic on ChatGPT," your content is nowhere in the answer. You're not alone. According to Backlinko's research on English-language websites, only 11% of sites are cited by both ChatGPT and Perplexity simultaneously (the study's sample skews English; overlap rates for other languages may differ). The logic AI search engines use to select citation sources is fundamentally different from Google's ranking algorithm.

This guide is based on Princeton University and IIT Delhi's GEO research published at KDD 2024, combined with hands-on deployment experience from our own platform, Shareuhack. AI-referred sessions grew 527% year-over-year in the first 5 months of 2025, and there's still a relative first-mover advantage — especially in markets where deep, practical GEO content doesn't yet exist.

TL;DR

• GEO isn't a new technology — it's about making your content directly "extractable" by AI. Answer Capsules (40-60 word direct answers below H2 headings) are the fastest entry point
• ChatGPT and Perplexity are two almost entirely different citation ecosystems — you can't use the same strategy for both
• Removing promotional tone matters more than any technical optimization. According to Semrush, promotional language reduces AI citation rates by 26.19%
• Personal blogs have bigger GEO opportunities than you'd think. Princeton's research shows low-ranking websites can boost visibility by up to 115% through GEO
• All quantitative data comes from English-language research. Applicability to other markets remains unverified. This article is itself a GEO experiment

Why AI Doesn't Cite Your Content: 5 Root Causes

When AI doesn't cite your content, it's usually not a quality problem — it's that AI literally can't "read" or "extract" your answers. Here are the five most common root causes:

1. GPTBot and PerplexityBot are blocked by your robots.txt

This is the most overlooked issue. Many websites' robots.txt files block all non-Google crawlers by default. According to Cloudflare's 2025 crawler report, GPTBot now crawls 8x more frequently than Googlebot with request volume growing 305%, while PerplexityBot has grown by 157,490%. If you haven't explicitly allowed these crawlers, AI search engines simply can't see your content.

2. Your answers are buried in prose — AI can't extract them

AI search engines need to pull out a standalone passage from your article that directly answers a user's question. If your answer is scattered across three paragraphs requiring readers to piece it together, AI will skip you for a clearer source. According to Search Engine Land, cited content contains 32% more explicit concept definitions than uncited content.

3. You're not in AI's trusted source network

ChatGPT draws 47.9% of its citations from Wikipedia, while Perplexity favors Reddit-style community discussions at 46.7%. This doesn't mean you need to be on Wikipedia, but AI has its own "trust circle." If your site isn't referenced by trusted sources or linked from high-authority external sources, your chances of entering the AI citation network decrease.

4. Missing verifiable factual signals

AI engines prefer content with concrete data, clear definitions, and traceable sources. "This tool is great" won't get cited. "According to our March 2026 tests, this tool reduced processing time from Y minutes to Z minutes in X scenario" will.

5. You're not in Bing's index (ChatGPT's blind spot)

ChatGPT uses Bing as its search starting point. According to Averi.ai, 87% of ChatGPT citations come from Bing's top-ranking results. If your website has poor Bing index coverage, ChatGPT won't even consider your content.

Self-diagnostic checklist: (1) Check if your robots.txt allows GPTBot and PerplexityBot (2) Verify your index status on Bing Webmaster Tools (3) Pick an article and see if you can find a standalone paragraph that answers a question within 10 seconds. If you can't find one, neither can AI.

Princeton's 9 GEO Techniques: Which 3 Are Worth Implementing Now

The GEO paper published at KDD 2024 by Princeton University and IIT Delhi is currently the most rigorous GEO effectiveness study, testing 9 content optimization techniques. No guessing needed — academic research already has the answers.

Technique	Effect	Best For	Recommendation
Statistics Addition	+37% on Perplexity	All content	Implement first
Quotation Addition	Up to +41% overall	All content	Implement first
Cite Sources	+115.1% for low-ranking sites	Small/medium sites	Implement first
Authoritative Tone	Positive	Legal, medical, history	Use when appropriate
Technical Terms	Positive	Technical content	Use when appropriate
Easy-to-Understand	Positive	Broad audience topics	Use when appropriate
Fluency Optimization	Slightly positive	All content	Lower priority
Unique Words	Slightly positive	All content	Lower priority
Keyword Stuffing	10% worse than baseline	None	Avoid completely

The three-in-one strategy: Statistics Addition + Quotation Addition + Cite Sources. Here's how:

1. Replace qualitative with quantitative: "This is growing fast" → "According to Cloudflare's 2025 report, GPTBot request volume grew 305% year-over-year"
2. Add at least one external citation per core paragraph: Reference academic papers, official documentation, or industry reports so AI can verify your claims
3. Cite sources to build AI trust: For small and medium sites, Cite Sources has the most dramatic effect. In Princeton's study, low-ranking sites saw visibility increase by 115.1% with this technique alone

Research limitations: This Princeton paper tested English content, and its benchmark platforms included the then-current BingChat. Whether these findings fully apply to non-English content in AI search engines hasn't been independently verified. Use these numbers as directional guidance, and track your own results.

Writing AI-Citable Answer Capsules: The 40-60 Word Template

An Answer Capsule is a 40-60 word self-contained direct answer placed immediately below an H2 heading. According to Norg.ai, 72.4% of blog posts cited by AI contain identifiable Answer Capsules, and ChatGPT draws 44% of its citations from the first third of articles.

The core principle is passing the "Information Island" test: the passage makes complete sense when read in isolation, without any surrounding context.

A note on word count: The 40-60 word guideline comes from English-language research. In practice, aim for 2-3 sentences that independently answer a question, rather than hitting a specific word count.

Answer Capsule formula:

[Topic] is [definition, 1-2 sentences]. According to [source], [quantified fact].

Before and after comparison:

Before (traditional blog style):

A lot of people are talking about this new concept called GEO. It's basically a way to help your content get seen by AI search engines. It's becoming increasingly important these days.

After (Answer Capsule format):

GEO (Generative Engine Optimization) is the practice of optimizing content visibility for AI search engines like ChatGPT and Perplexity. According to Princeton's KDD 2024 study, adding statistics and source citations can boost AI citation rates by 30-40%.

The second version can be directly extracted by AI as an answer. The first can't. The difference: BLUF (Bottom Line Up Front) structure, verifiable facts, and explicit source attribution.

Another key finding: According to Semrush, Q&A formatting boosts AI citation rates by 25.45%, while promotional tone reduces them by 26.19%. "Absolutely essential," "the best choice," and "act now" are GEO killers. If your articles are filled with marketing language, removing it is a higher priority than adding Answer Capsules.

What you can do right now: Open your best-performing article, add a 40-60 word Answer Capsule below each H2 heading. Fix structure first, then worry about technical details.

Platform Strategy Divergence: How to Approach ChatGPT, Perplexity, and Google AI Overview Differently

Optimizing for ChatGPT, Perplexity, and Google AI Overview as if they were one platform is like running the same strategy for LinkedIn and TikTok. According to ZipTie.dev, the three platforms have entirely different citation logic and require tailored strategies.

Dimension	ChatGPT	Perplexity	Google AI Overview
Search basis	Training data + Bing index	Real-time web search	Google index + E-E-A-T
Source preference	Wikipedia (47.9%)	Reddit (46.7%)	Strong E-E-A-T sources (96%)
Citation criteria	Bing top results (87%)	Recently updated content (76.4% within 30 days)	YouTube (18.8%) has extra advantage
Update reflection	6-12 weeks	2-4 weeks	2-4 weeks
Small site opportunity	Medium (needs Bing index)	High (values niche, real-time content)	Low (highest E-E-A-T bar)

Time is limited — focus on one platform first. Choose based on your situation:

• If you're a personal blogger or niche site → Start with Perplexity. It favors real-time, niche expert content, offers the fastest feedback loop (2-4 weeks), and is the friendliest to small sites
• If your site already has solid Bing index coverage and rankings → Start with ChatGPT. It has the largest market share (64.5% as of March 2026 per First Page Sage), and your Bing ranking advantage transfers directly
• If you already have strong E-E-A-T signals and YouTube content → Google AI Overview is your home turf

Don't spread your efforts across all three platforms simultaneously. Get results on one, then expand.

Technical Infrastructure: llms.txt + Schema + robots.txt GEO Configuration

Technical infrastructure is GEO's "entry ticket," not the "ride." First ensure AI can access your site, then optimize how it cites you.

robots.txt: Make Sure AI Crawlers Can Get In

Confirm these settings in your robots.txt:

User-agent: GPTBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Google-Extended
Allow: /

According to Perplexity's official documentation, PerplexityBot robots.txt updates are reflected within a maximum of 24 hours. Changes take effect the next day.

llms.txt: A Low-Cost AI-Friendly Signal

llms.txt was proposed by Jeremy Howard (founder of fast.ai) in September 2024. It's a Markdown file placed at your site's root directory that tells AI "what this website is about and which content matters most."

Its relationship with robots.txt is complementary, not a replacement: robots.txt controls "where crawlers can go," while llms.txt communicates "what this site focuses on."

An honest note: A 2025 August independent audit found that mainstream AI crawlers ignored most llms.txt files, and no major LLM provider has formally committed to following this specification. However, the deployment cost is minimal (under 30 minutes), and companies like Anthropic and Mintlify have adopted it. Worth deploying, but don't expect it alone to change your citation rates.

Deploying on Next.js: Simply place a static file at /public/llms.txt. Template:

# Your Site Name

> One-sentence description of your site's core value

## Main Content Categories

- [Category Name](/category-path): Brief description
- [Category Name](/category-path): Brief description

## Featured Articles

- [Article Title](/posts/slug): One-sentence description

Here's what Shareuhack's actual deployed llms.txt looks like:

# Shareuhack

> Helping people who want to live smarter do smarter things. Covering tech tools, remote work, digital nomad life, and side hustle monetization.

## Content Categories

- [Tech & Tools](/tech): AI tool reviews, developer productivity, open source projects
- [Work](/work): Remote work, career strategy, project management
- [Money](/money): Side hustle playbooks, financial planning
- [Life](/life): Digital nomad guides, expat living

## Popular Articles

- [AI Side Hustle Scam Guide](/posts/ai-side-hustle-scam-guide-2026): A practical framework for identifying AI side hustle scams

Copy this structure, replace with your own site info, drop it at /public/llms.txt, and you're done.

Schema Markup: FAQPage First, HowTo Is Deprecated

Research on Schema's GEO impact is contradictory. Some studies show pages with Schema have a 36% advantage in AI summaries, but Atlas's December 2024 study found no direct correlation between Schema coverage and AI citation rates.

The conclusion follows the 80/20 rule: spend 70% of your effort on content structure optimization, 30% on Schema.

Priority order:

1. FAQPage Schema — Best match for AI's Q&A response patterns, most consistent results
2. Article Schema + Organization Schema — Basic structured signals
3. HowTo Schema — Google deprecated it in January 2026; don't invest more time here

If your site already has FAQ functionality (e.g., frontmatter faqs field + UI auto-rendering accordion), AI can already extract FAQ content from the HTML structure. But for a stronger signal, adding JSON-LD format FAQPage Schema is more reliable. For example in Next.js:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "Your question?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Your answer."
      }
    }
  ]
}
</script>

Add this to your Next.js <Head> or page component, replacing the questions and answers with your FAQ content.

30-minute technical checklist: (1) Confirm robots.txt allows GPTBot and PerplexityBot (2) Deploy /public/llms.txt (3) Verify Bing Webmaster Tools index status. Advanced items (requires dev time): add FAQPage JSON-LD, set up Article Schema.

Measuring GEO Results: Free Methods to Know If You're Being Cited

You don't need paid tools to establish a GEO measurement baseline.

Minimum viable tracking path (start in 10 minutes):

1. HubSpot AEO Grader — A free tool. Enter your brand name and URL to instantly see your visibility baseline across ChatGPT, Perplexity, and Gemini. No prerequisites, just use it
2. Weekly manual checks — Pick 3 keywords you most want to be cited for. Search each weekly on ChatGPT, Perplexity, and Google AI Overview. Track in a Google Sheet: date, platform, keyword, cited or not, citation position. Takes 5 minutes
3. Evaluate paid tools after 3 months — If manual tracking shows you're getting cited, consider Profound (G2 Winter 2026 Leader, monitors 10+ AI engines automatically)

Realistic timeline expectations:

Platform	First appearance	Stable citations
Perplexity	2-4 weeks	1-3 months
ChatGPT	6-12 weeks	3-6 months
Google AI Overview	2-4 weeks	3-6 months

According to Cloudflare's report, GPTBot crawls 8x more frequently than Google. Technical changes (opening robots.txt, deploying llms.txt) are reflected faster than content optimization.

Start with Perplexity as your quick validation platform: the shortest feedback cycle, with initial results observable in 2-4 weeks.

Risk Disclosure: Black Hat GEO and Realistic Expectations for Small Sites

Black Hat GEO: Don't Do These

According to Search Engine Land, these are confirmed black hat GEO tactics:

• AI crawler cloaking: Serving different content to GPTBot than to real users
• Fake E-E-A-T: Using AI to generate fictional author personas and credentials
• Schema abuse: Injecting structured markup that doesn't match actual content
• Data poisoning: Injecting misleading competitor information into AI models (China's 2026 CCTV 3·15 exposé revealed an entire industry built around paid manipulation of AI rankings)

The consequences are severe: full site deindexing, manual penalties, and complete loss of AI citations.

The test for legitimate GEO is simple: Does this optimization make the content more useful for readers? If yes, it's legitimate GEO.

Realistic Expectations for Small Sites

Good news: GEO's opportunity structure favors small sites more than SEO does. Princeton's research shows websites ranked 5th in SERPs can boost visibility by up to 115% through GEO techniques, with the lowest-ranking sites benefiting the most. AI engines prioritize "topic depth" over "domain age."

But keep expectations realistic:

• Topic cluster strategies take time: Start by rewriting your 1-2 best existing articles rather than building a 5-10 article cluster from scratch. For side-project bloggers publishing 1-2 posts per month, building gradually over 6 months is more reasonable than trying to produce everything at once
• First-hand perspective is your structural advantage: Personal blogs can offer something large media sites can't replicate. "I actually tested X, and the result was Y" — this type of first-hand experience is an E-E-A-T signal that AI increasingly values
• All data comes from English-language research: Whether AI citation behavior for non-English content is identical remains unverified by independent studies. Use these numbers as directional guidance; ultimately, trust your own tracking data

Conclusion

GEO isn't replacing SEO — it's making your existing good content readable by AI. The entry ticket is structure (Answer Capsules, clear H2/H3 headings), the passport is factual density (statistics, source citations), and the visa is first-hand perspective (your tests, your experience).

This article is itself Shareuhack's GEO experiment. We've deployed llms.txt and FAQPage Schema, and will continue tracking how this article gets cited across AI platforms, updating with real data as it comes in.

The one thing you can do today: Open your best article and add a 40-60 word Answer Capsule below the first H2. It's the lowest-cost, highest-potential single change you can make.