- Google AI Overviews now appear for an estimated 30–40% of English searches. A BrightEdge study found 42% of AI citations come from pages outside the organic top 10.
- GEO (Generative Engine Optimization) is about getting AI tools to cite your content. Adding data with sources increases AI citation rates by up to 52% (Princeton, 2024).
- AEO (Answer Engine Optimization) targets Featured Snippets, People Also Ask, and AI Overviews with clear 40–60 word definitive answers.
- LLMs.txt is a free plain-text file you put at yourdomain.com/llms.txt. It tells AI crawlers what your site is and what to cite. Takes 10 minutes. Perplexity already respects it.
- 7 free steps at the bottom of this guide require no paid tools — just Google Search Console, a text editor, and schema markup.
- What Changed: The SERP in 2026
- SEO vs GEO vs AEO: Clear Definitions
- Google's Key Updates in 2024–2025
- GEO: How to Get AI to Cite You
- AEO: Winning the Answer Box
- LLMs.txt: The New Standard for AI Crawlers
- XML Sitemap + AI — What Changed
- How to Get Cited by Each AI Tool
- 7 Free Steps You Can Do Today
- Why Next.js Wins for SEO & AI Crawlers
- Frequently Asked Questions
What Changed: The SERP in 2026
Open Google and search "how to hire a web developer." At the top, before any blue links, you will see a paragraph written by Google's AI — with three or four source links cited inline. The user reads the AI answer. Many never scroll to the organic results below.
This is not a future scenario. Google launched AI Overviews in the United States in May 2024 and expanded globally throughout 2024–2025. By 2026, they appear consistently for factual, how-to, comparison, and definitional queries across most English-language searches.
Meanwhile, external AI tools — ChatGPT with Browse, Perplexity, Claude, and others — are now answering search-like queries directly, pulling content from the web in real time. The question is no longer "how do I rank #1?" The question is "how do I become the source AI uses?"
SEO vs GEO vs AEO: Clear Definitions
These three terms overlap but are not the same. Understanding the distinction determines which tactics you prioritize.
The relationship: SEO is the foundation. AEO is a focused tactic within SEO. GEO is a broader discipline that extends optimization beyond Google to all AI platforms. In 2026, a complete strategy requires all three layers — shown in the pyramid further below.
Google's Key Updates in 2024–2025
March 2024 Core Update: The End of AI Content Spam
Google's March 2024 Core Update, combined with its Spam Update, integrated the Helpful Content System directly into the core ranking algorithm. The target was clear: sites using mass AI-generated content with no editorial oversight saw drops of 30–70% in organic traffic. Sites with genuine first-hand expertise and experience gained.
Key shift: Google no longer penalizes AI-generated content as a category. It penalizes content that exists primarily to rank rather than to help. A 3,000-word article written by AI but reviewed, supplemented, and published by a named expert with real-world experience can rank. The same article mass-published across 200 sites without attribution or editing gets deindexed.
AI Overviews: What Google Actually Said
Google has stated that AI Overviews use the same signals as their core ranking algorithm — but apply additional filtering for answer quality. The AI does not just take the #1 result. It synthesizes across multiple high-quality sources and cites the ones that provide the clearest, most specific, most verifiable answers for each part of the query.
Google also confirmed that sitemaps are a priority signal for AI Overview sourcing. Pages that are not in your sitemap, not internally linked, or not recently crawled are less likely to appear in AI Overviews — regardless of their keyword ranking.
How AI Overviews Decide What to Cite
GEO: How to Get AI to Cite You
GEO is grounded in real research. A 2024 study by researchers at Princeton University, Georgia Tech, IIT Delhi, and others tested specific content modifications against generative AI models to measure citation rate changes. The findings were published as "GEO: Generative Engine Optimization."
| Optimization Change | What It Means in Practice | Avg. Lift in AI Citation Rate |
|---|---|---|
| Add authoritative citations | Link claims to named studies, reports, or data sources | |
| Add statistics with sources | "72% of Nepal's web traffic is mobile (Statcounter, 2024)" | |
| Include expert quotations | Named quotes from industry figures or your own expertise | |
| Improve writing fluency | Clear prose, short sentences, no filler. AI models penalize hard-to-parse content. | |
| Keyword relevance optimization | Standard SEO keyword targeting — still matters but less than the above |
The pattern is clear: AI models prefer content that is verifiable, attributed, and specific. The biggest gains come from adding citations and data — not from keyword density or content length. This is a fundamental shift from how most SEOs have been writing content for the past decade.
5 Core GEO Tactics
- Replace vague claims with specific, sourced data. "Most businesses" becomes "62% of Nepal SMEs (NRB Survey, 2025)". AI models cite the specific version. They ignore the vague version.
- Write definitive, quotable sentences. Structure key claims as standalone declarative sentences. "A Nepal business website typically costs NPR 20,000–80,000 depending on complexity." This is quotable. A five-sentence paragraph discussing cost considerations is not.
- Establish entity recognition across all pages. Mention your name, your company, your location, and your expertise consistently. Google's Knowledge Graph links entities. When an AI model processes your content, named entities with consistent presence get weighted more than anonymous sources.
- Add structured data for every content type. FAQPage schema for Q&A sections. HowTo schema for step-by-step content. Article schema with author and datePublished on every post. Schema markup is machine-readable E-E-A-T — it directly communicates to AI what your content is and who wrote it.
- Earn citations from other sites. Traditional backlinks still matter for GEO. If other websites cite you as a source ("according to Rahul Ranjan's guide at rahulranjan.com.np"), AI models inherit that attribution signal. Write content good enough to be cited, then promote it to earn those citations.
AEO: Winning the Answer Box
AEO is the tactical implementation of GEO specifically for Google's SERP answer features. The mechanics are consistent: write a clear question as a heading, follow it immediately with a direct 40–60 word answer, and mark the entire section with FAQPage or HowTo schema.
The 40–60 Word Answer Rule
Featured Snippets and AI Overviews consistently pull answers between 40 and 60 words. Google's AI does not paraphrase — it extracts. Put the answer before the explanation. If your explanation comes first, the AI cannot find the extractable answer block and moves to the next source.
"When considering the question of how much a website costs in Nepal, there are many factors to take into account including the type of site, the developer's experience level, the technology stack chosen, and the required features. Based on all of these variables, the price can vary significantly..."
AI cannot extract a clean answer. Skipped.
A Nepal business website typically costs NPR 20,000–80,000 for a standard informational site, NPR 50,000–2,00,000 for e-commerce, and NPR 2,00,000+ for custom SaaS or web applications. The final price depends on the technology stack, number of features, and whether you hire a freelancer or agency.
Clear extractable answer. AI cites this.
Format Matches Search Intent
| Query Type | Best Format | Schema to Use |
|---|---|---|
| "What is X" | Paragraph definition (40–60 words) | FAQPage |
| "How to X" | Numbered list of steps | HowTo |
| "Best X" / "Top X" | Bulleted list with brief descriptions | ItemList |
| "X vs Y" | Comparison table | FAQPage + Table |
| "Cost of X" | Table with price ranges + context | FAQPage |
LLMs.txt: The New Standard for AI Crawlers
In September 2024, Jeremy Howard — founder of fast.ai and co-creator of the fastai library — proposed a standard called llms.txt. The concept mirrors robots.txt, but instead of allowing or blocking crawlers, llms.txt helps AI language models understand and correctly attribute your content.
Where it lives: yourdomain.com/llms.txt — a plain text file at your domain root. No server configuration needed. No CMS plugin needed. Write it in a text editor, upload via FTP or your hosting panel.
Which AI Tools Respect LLMs.txt?
| AI Tool | LLMs.txt Status | Action Required |
|---|---|---|
| Perplexity | Confirmed support | Create llms.txt — Perplexity respects it now |
| Anthropic (Claude) | In progress (2025–26) | Create it now — support being rolled out |
| OpenAI (ChatGPT) | No official stance | Ensure Bing index access via robots.txt |
| Google (AI Overviews) | No endorsement yet | Focus on schema + sitemap + E-E-A-T |
Some sites also publish
llms-full.txt — a single file containing the full text of all key pages, formatted for AI consumption. This portfolio has both. It allows AI tools that train on web content to understand your site without crawling every URL. It's more maintenance work but increases comprehensiveness of AI understanding.
XML Sitemap + AI — What Changed
Your XML sitemap at yourdomain.com/sitemap.xml now serves two audiences: Google's traditional crawler and Google's AI indexing system. Both use the sitemap to determine what content exists and how recently it was updated.
Sitemap Best Practices for 2026
- Keep
<lastmod>honest. Google has publicly stated it ignores lastmod values it considers inaccurate. If you update a page, update the date. If you haven't touched a page, leave the tag as-is. A false lastmod date trains Google to distrust all your lastmod values. - Include all canonical pages you want in AI Overviews. Orphaned pages — not in your sitemap and not linked internally — rarely get cited by AI Overviews. If a page isn't in your sitemap, assume AI doesn't know it exists.
- Set priority values correctly. Your most important pages (homepage, service pages, flagship blog posts) should carry
<priority>0.9</priority>. Utility pages get 0.3–0.5. This signals to AI which content matters most on your site. - Add image sitemaps for pages with diagrams or screenshots. Google uses image sitemaps for AI-augmented visual answers. If your service pages have screenshots, architecture diagrams, or UI mockups, add them to an image sitemap.
- Submit to Bing Webmaster Tools. Bing powers ChatGPT's browsing. Submitting your sitemap to Bing Webmaster Tools (free, takes 20 minutes) is the single most direct action to improve ChatGPT citation rates.
How to Get Cited by Each AI Tool
Each AI tool has a different sourcing mechanism. What works for Google AI Overviews is not identical to what gets you cited by Perplexity or ChatGPT. Here is the breakdown:
- Rank on page 1 for target keywords (but not required — 42% of citations are page 2+)
- Pass Core Web Vitals on mobile
- Add FAQPage and Article schema with author
- Write definitive 40–60 word answers per section
- Keep sitemap updated with accurate lastmod
- Create an llms.txt file — Perplexity officially respects it
- Ensure PerplexityBot is not blocked in robots.txt
- Write content with clear, citable statements and source links
- Author bio with credentials increases citation probability
- Fast page load = higher crawl priority
- Submit site and sitemap to Bing Webmaster Tools (free)
- Ensure BingBot access in robots.txt
- Bing values the same E-E-A-T signals as Google
- Microsoft Clarity (free) analytics can improve Bing signals
- Use Bing URL Submission API for faster indexing
- llms.txt support being actively integrated (2025–26)
- High-quality, frequently-cited content enters training data
- Real-time search uses Bing index (same as ChatGPT)
- Anthropic's crawler (ClaudeBot) respects robots.txt
- Content with clear authorship and domain reputation ranks higher
7 Free Steps You Can Do Today
Every step below requires zero paid tools. Total time: under 3 hours for a complete implementation on an existing site.
Why I Use Next.js Over Other Frameworks for SEO
I get asked this constantly: "Why not just React? Why not WordPress?" The answer is rooted in how Google and AI crawlers actually index content — and the specific bottlenecks I've seen cause ranking failures across dozens of Nepal sites.
The Core Problem with React SPAs (Single Page Apps)
A plain React app renders in the browser using JavaScript. When Googlebot (or any AI crawler) hits a React SPA, it receives an almost-empty HTML file with a single <div id="root"></div>. The actual content only appears after JavaScript executes — which happens in a second crawl pass Google calls "rendering."
The problem: Google may never complete that second render pass for low-priority pages. Perplexity and ChatGPT's browsing crawlers do not run JavaScript at all — they read raw HTML only. This means a React SPA's content is invisible to most AI crawlers, regardless of how well-optimized the JavaScript is.
- ❌ Empty HTML on first load
- ❌ AI crawlers see blank page
- ❌ Meta tags not readable without JS
- ❌ Googlebot may skip rendering
- ✅ Great for dashboards/apps
- ✅ Server-rendered HTML
- ❌ Slow default performance (LCP 3–6s)
- ❌ Plugin bloat kills Core Web Vitals
- ❌ Difficult to optimize to LCP <2.5s
- ✅ Easy content management
- ✅ Full HTML on first load (SSR/SSG)
- ✅ AI crawlers read complete content
- ✅ Per-page meta tags built in
- ✅ Consistently achieves LCP <2s
- ✅ Automatic image optimization
Next.js Image Handling: The SEO Advantage Nobody Talks About
The next/image component is one of the most underrated SEO tools in any framework. Here is what it does automatically:
| Feature | What next/image Does | SEO Impact |
|---|---|---|
| Format conversion | Serves WebP/AVIF automatically to browsers that support it | 20–40% smaller files → faster LCP |
| Responsive sizes | Generates multiple sizes, serves the right one per device | No oversized images on mobile → better CLS, LCP |
| Lazy loading | Below-fold images load only when scrolled into view | Reduces initial page weight → faster FCP |
| Priority preload | Hero images get priority prop → browser preloads them | Direct LCP improvement — Google's #1 ranking signal |
| Size reservation | Requires width/height — reserves space before load | Eliminates layout shift → CLS = 0 |
| blur placeholder | placeholder="blur" shows low-quality preview | Better perceived performance |
next/image configuration and SSG, the mobile LCP dropped to 1.7s. Google Search Console showed a 34% increase in organic impressions within 8 weeks — driven entirely by the Core Web Vitals improvement.
Next.js for GEO and AI Crawlers
SSR and SSG mean that every page's complete HTML — including all the schema markup, meta tags, Open Graph tags, and article content — is readable by Googlebot and all AI crawlers on the very first HTTP request. There is no JavaScript dependency, no second rendering pass, no "maybe it got indexed correctly."
This matters enormously for GEO. Perplexity's PerplexityBot and ChatGPT's OAI-SearchBot do not execute JavaScript. They read raw HTML responses. A React SPA is invisible to them. A Next.js SSG-built page is completely readable — every heading, every FAQ section, every schema block in the <head>.
The bottom line: If you are building a site where SEO and AI citation are goals — a business site, a portfolio, a blog, a service page — Next.js with SSG is the correct framework choice. React SPA for dashboards and apps. Next.js for anything public-facing that needs to rank.
Why This Matters More in Nepal Than Anywhere Else
Nepal's digital competitive landscape has a structural advantage that most sites globally cannot replicate: almost no competing site is doing any of this. While global SEOs are scrambling to adapt to GEO and AEO requirements, Nepal business websites are still missing basic schema markup and author bios.
This means the bar for AI citation in Nepal-relevant queries is extremely low. A site that implements the 7 steps above is, by a significant margin, the most AI-optimized Nepali content on most topics. Google and Perplexity will cite you not because you outrank global competitors — but because you are the only Nepali source they can extract a clean, attributed, schema-marked answer from.
I implement all of these on every site I build or audit
LLMs.txt, schema markup, sitemap accuracy, author bio, Core Web Vitals — these are standard in every web project I take on. If your site is missing these, let's fix it together.
Get a free SEO auditFrequently Asked Questions
What is GEO (Generative Engine Optimization)?
What is AEO (Answer Engine Optimization)?
Does LLMs.txt actually help SEO?
Will AI Overviews kill SEO?
How do I check if Google AI Overviews is showing my content?
Is this relevant for Nepal-based websites?
Sources and Further Reading
- "GEO: Generative Engine Optimization" — Aggarwal et al., Princeton/Georgia Tech/IIT Delhi (2024)
- BrightEdge AI Overviews Generative Parser Research (2024) — 42% non-top-10 citation finding
- Google Search Central — Helpful Content System and E-E-A-T documentation (2024–2025)
- LLMs.txt proposal — Jeremy Howard, answer.ai (September 2024)
- Perplexity blog — LLMs.txt support and PerplexityBot documentation (2024)
- SE Ranking — AI Overview appearance rate across query types study (2025)
- Google Search Console Help — AI Overview Performance reporting (2025)
- Bing Webmaster Tools documentation — ChatGPT Browsing index relationship
Let me implement all of this on your site
LLMs.txt, FAQPage schema, author E-E-A-T, sitemap accuracy, Core Web Vitals — I audit AND implement. You get results, not a spreadsheet of problems.
