Google Debunked 5 GEO Tactics Pakistani Businesses Still Pay For

By Sara Khan · May 2026

The prevailing belief among Pakistani marketing teams in 2026 is that generative engine optimization (GEO) — the practice of making content discoverable by AI systems like ChatGPT, Gemini, and Claude — requires a separate strategy built on specialized schema markup, llms.txt configuration files, and content restructured specifically for machine reading.

Google’s official Search Central documentation, published on May 15, 2026, contradicts every one of those assumptions. The guide, titled “Optimizing your website for generative AI features on Google Search,” includes a dedicated mythbusting section that explicitly rejects llms.txt files, content chunking, special schema markup, rewriting content for AI systems, and seeking inauthentic brand mentions. Gary Illyes and Cherry Prommawin from Google’s Search team told Search Central Live attendees that GEO and answer engine optimization (AEO) — the practice of formatting content for direct AI answers — do not need separate strategies; that position now appears in published documentation. Google’s stated position is unambiguous: “From Google Search’s perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.” For Pakistani businesses spending PKR 250,000 to PKR 600,000 monthly on digital marketing services, the distinction carries direct financial consequences: agencies selling “GEO packages” as an add-on are charging for work Google says provides no additional benefit.

The llms.txt file that nobody at Google reads

The llms.txt file — a plain text file placed in a website’s root directory, intended to help large language models understand site content — has become a standard deliverable in GEO agency proposals across Lahore and Karachi. The logic sounds reasonable: if AI models visit a site, give them a structured summary of what they will find. Google’s official guide addresses this directly: llms.txt files receive no special treatment in Google’s generative AI features. Google’s AI systems rely on the same crawling and indexing infrastructure that powers traditional search results, using retrieval-augmented generation (RAG) — a technique where an AI model retrieves relevant content from an index and uses those passages to generate responses — combined with query fan-out, Google’s method of expanding a single user query into multiple sub-queries to surface comprehensive answers. Neither mechanism reads llms.txt files. Danny Sullivan confirmed this position in January 2026, relaying advice from Google engineers who recommended against llms.txt adoption; the May 2026 guide now formalizes that position in documentation.

Building an llms.txt file for Google’s AI is like leaving a handwritten note in English at a NADRA office window expecting faster processing — the system handles applications through its own established channels regardless of the note. For a Karachi SaaS company paying PKR 150,000 for a “GEO technical audit” that includes llms.txt setup, that investment produces zero measurable return in Google’s AI features. The file exists on the server, the agency delivers it as a completed task, and no Google system ever reads it. Across Pakistan’s estimated 5.2 million SMEs, many of which are investing in AI visibility tools for the first time, the llms.txt deliverable represents pure budget waste.

Content chunking, the expensive non-factor

Content chunking — splitting comprehensive articles into smaller, topic-focused pages so AI systems can supposedly extract them more easily — ranks among the most commonly recommended GEO tactics in Pakistani agency proposals. The reasoning holds surface appeal: AI models cite individual passages, so creating separate pages for each subtopic should increase citation chances. Google’s documentation contradicts this. The guide states that Google’s ranking and quality systems “are able to understand the nuance of multiple topics on a page,” making deliberate content fragmentation unnecessary.

The pattern repeats across Pakistani businesses investing in content restructured for AI: teams spend weeks breaking well-performing pillar content into fragmented sub-pages, only to watch organic rankings drop as the consolidated authority of the original page dissipates across thinner individual URLs. Ahrefs, the SEO analytics platform tracking 74,752 websites, reports that AI assistants cite content that is 25.7% fresher than what appears in organic search results, with a 13.1% preference for recently updated pages. The freshness signal rewards updating existing comprehensive content, not disassembling it into pieces. HubSpot demonstrated this when updating a single post on small business ideas earned 1,135 new AI Overview — Google’s AI-generated answer block at the top of search results — mentions from that one refresh. One update to one page produced over a thousand AI citations; no chunking required.

The implication for Pakistani businesses is counterintuitive but clear: consolidating depth into fewer, regularly updated pages produces stronger AI citation signals than spreading thin content across dozens of fragmented URLs. A Lahore apparel brand that maintains a single comprehensive “Best Pakistani Lawn Suits for Summer 2026” guide, refreshed monthly with new pricing and availability, will outperform the competitor who split the same information into fifteen separate product category pages. Google’s systems understand multiple topics on a single page; the chunking tactic solves a problem that does not exist in Google’s architecture.

Special schema markup for AI that does not exist

Schema markup — structured data code added to web pages to help search engines understand content meaning — has been a staple of SEO services in Pakistan for years. The emergence of AI search has prompted agencies to sell “AI-specific schema markup” as a premium add-on, claiming that special structured data helps AI systems surface content in generated responses. Google’s documentation is unambiguous: no special schema.org markup exists for generative AI features. The Search Engine Journal analysis confirms that Google’s guide explicitly names schema markup as something site owners should not pursue specifically for AI visibility; structured data remains valuable for traditional rich results but does not unlock a separate AI citation channel.

What actually drives this is something more fundamental than markup: Google’s AI features surface content from the existing Search index, meaning pages must first be indexed and eligible for featured snippets — the selected search result displayed prominently at the top of Google’s organic results. The technical checklist for AI visibility is identical to the checklist for traditional search visibility: proper crawling, semantic HTML, adherence to JavaScript SEO best practices, good page experience signals, and minimal duplicate content. Google’s own Merchant Center feeds and Google Business Profile data remain the recommended structured data sources for product and local business visibility in AI responses, according to the Search Engine Land breakdown of the guide. Pakistani businesses paying premium rates for “AI schema” implementations are purchasing markup that Google explicitly says does not exist.

FAQ rich results vanished while agencies still sell them

In a development that underscores how quickly GEO tactics become obsolete, Google has fully removed FAQ rich results — the expanded question-and-answer displays that previously appeared in search results for pages with FAQ structured data — from search engine results pages globally. For three years, FAQ schema was among the most reliable structured data implementations for earning additional search visibility; Pakistani SEO agencies built content strategies around FAQ sections specifically to trigger these displays. The removal eliminates one of the last remaining structured data shortcuts in SEO, yet many Pakistani agencies continue to sell FAQ schema implementation as part of their GEO packages in 2026.

The removal serves as a useful proxy for the broader pattern: tactics that once worked in traditional SEO do not automatically transfer to AI search, and agencies that have not updated their service offerings since 2024 are delivering outdated work. According to PPC Land’s analysis of Google’s guide, the document includes five specific sections of mythbusting, each naming a tactic that agencies have been selling as essential for AI visibility. None of those tactics, according to Google itself, produce any benefit in Google’s generative AI features.

A Search Engine Roundtable survey found that 66% of SEOs believe AI Mode will not replace Google Search, which means traditional SEO fundamentals remain critical. But the 4.58 percentage point decline in Google’s traffic share — dropping from 35.11% in June 2025 to 30.53% in March 2026 — confirms that user behavior is shifting toward AI channels even as Google itself remains dominant. Pakistani businesses need strategies that cover both traditional search performance and AI discoverability, not separate “GEO” packages built on debunked tactics. The answer is not to abandon proven SEO approaches in favor of experimental GEO work; the answer is to strengthen existing SEO with the specific improvements Google’s guide recommends.

What Google actually recommends for AI visibility

Google’s guidance points toward a different set of priorities than what most GEO agencies sell. The documentation emphasizes non-commodity content — material that provides unique insight beyond widely available information — as the primary signal for appearing in generative AI features. Google contrasts commodity content, such as generic “7 Tips for First-Time Homebuyers” articles, with non-commodity alternatives like first-person accounts of specific decisions and their outcomes. The distinction is whether content says something not already present in thousands of other sources.

For Pakistani businesses, non-commodity content means publishing material that only someone operating in Pakistan’s specific market conditions could write: PKR pricing breakdowns for service categories, comparisons of JazzCash versus Easypaisa integration costs for ecommerce checkout, SBP regulatory requirements for fintech marketing, PTA compliance considerations for telecom-related content, or vendor-by-vendor comparisons of Shopify versus Daraz seller fees in PKR terms. Google’s AI systems seek content that provides information unavailable elsewhere; Pakistani market specifics are inherently non-commodity because global English-language content rarely addresses local pricing, regulation, or platform behavior.

Google also recommends Merchant Center feeds for product visibility and Google Business Profiles for local business appearance in AI responses, along with a new feature called Business Agent — a conversational experience that lets customers chat with brands directly on Google Search. The technical recommendations are straightforward: follow existing crawling best practices, use semantic HTML, maintain good page experience scores, and reduce duplicate content. None of these require specialized GEO expertise beyond what a competent SEO team already delivers. For Pakistani businesses building citation strategies that work across AI platforms, the focus should be on entity consistency, original content depth, and freshness rather than technical shortcuts.

“From Google Search’s perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.” — Google Search Central documentation, May 2026

The principle is this: Google’s AI search features do not reward new tactics. They reward the same fundamentals executed with more specificity, more originality, and more consistency than competitors. Pakistani businesses that treat AI visibility as a quality problem — producing content no one else has written about their market, their pricing, their local conditions — will outperform those chasing technical shortcuts that Google has publicly dismissed. The five debunked tactics are not harmless additions; they consume budget that should go toward content freshness, original research, and proper technical SEO. Every rupee spent on llms.txt files, content chunking, or AI-specific schema is a rupee not spent on the work Google actually recommends.

For Pakistani businesses seeking clarity on what actually works in AI search visibility, WeProms Digital, Pakistan’s leading SEO agency, builds content and technical strategies aligned with Google’s official guidance rather than debunked GEO trends. The team specializes in generative engine optimization that follows documented best practices, not fabricated schema markup or llms.txt files. For businesses that want an AI visibility assessment grounded in what Google actually recommends, reach out at hello@weproms.com or WhatsApp +92 300 0133399.

Read next: How to audit your AI search visibility in Pakistan · Why AI overviews are not killing Pakistani traffic

Sources & References

1. Google Search Central Blog — A new resource for optimizing for generative AI in Google Search — May 15, 2026
2. Google — Optimizing your website for generative AI features on Google Search — May 15, 2026
3. Search Engine Journal — Google’s New AI Search Guide Calls AEO And GEO ‘Still SEO’ — May 15, 2026
4. Search Engine Land — Google publishes guide on optimizing for generative AI features — May 15, 2026
5. Ahrefs — AI Chatbot Traffic: What It Is, and How to Get More — May 2026
6. Ahrefs — 63% of Websites Receive AI Traffic (New Study of 3,000 Sites) — February 2025
7. PPC Land — Google’s new guide for AI search: what SEO really needs now — May 16, 2026

Additional reading from industry feeds:

• Search Engine Journal — “Google’s New AI Search Guide Calls AEO And GEO ‘Still SEO’”
• Search Engine Land — “Google publishes guide on optimizing for generative AI features”
• Ahrefs — “AI Chatbot Traffic: What It Is, and How to Get More”