Legacy websites are falling behind: when SEO isn't enough and GEO doesn't exist

Something is happening slowly but surely: users — especially those under 35 — are increasingly asking AI before (or instead of) opening Google. ChatGPT answers product questions. Perplexity synthesizes service comparisons. Google AI Overviews delivers direct answers right on the results page without requiring a click. In that world, not appearing in an AI answer means not existing — even if you rank #3 on Google.

Traditional SEO: necessary but no longer sufficient

Most websites built before 2022 share the same set of basic SEO failures:

• Missing canonical URLs: multiple pages with similar content don't declare a preferred version — Google indexes all of them, authority is diluted across pages, and none is strong enough to rank well
• Empty or copy-pasted meta descriptions: meta descriptions don't directly affect ranking but do affect CTR — leaving them blank is money left on the table
• Missing or broken Schema.org markup: Google needs structured data to understand "what is this page" — a product, service, article, organization. No schema → Google guesses → guesses wrong → wrong display format
• H1/H2/H3 used arbitrarily: heading hierarchy is not styling — it's a semantic signal. Many sites use H2 because it's "bigger" rather than because it's a second-level section
• Stale sitemap: sitemap.xml is generated at deploy time and left untouched — the crawler has to discover new pages on its own
• Poor Core Web Vitals: LCP above 2.5s, layout shift when fonts load — Google has used these as quality signals since 2021

These failures accumulate. Each one alone isn't fatal, but combined they produce a website that neither Google nor AI truly understands — and what isn't understood isn't cited.

GEO — new concept, urgent pressure

GEO (Generative Engine Optimization) is the set of techniques that get your content cited and surfaced in AI-generated answers. It doesn't replace SEO — it sits on top of SEO and requires an additional layer of signals that most current websites completely lack:

• llms.txt: a file at /llms.txt that declares to AI crawlers which content may be used, summarizes the site's purpose, and lists important pages. OpenAI's GPTBot, Anthropic's ClaudeBot, and PerplexityBot all read this file.
• Explicit robots.txt for AI bots: many sites block * (all bots) in robots.txt without realizing this also blocks AI crawlers. Result: AI has no data about you → can't cite you.
• Complete Organization schema: AI engines need to know "who is this organization, what do they do, are they credible" before deciding to cite them. Schema.org Organization with name, url, contactPoint, areaServed, and knowsAbout is the most important identity signal.
• Author byline and E-E-A-T signals: AI engines — especially when answering technical or advisory questions — prioritize sources with clear authors, stated credentials, and demonstrated real-world experience. Articles with no author name, no publication date, and no credential section rank lower in AI's internal quality assessment.
• FAQ schema and HowTo schema: these are the two structured data types AI engines read directly to synthesize answers. Having FAQ schema means your answers are already packaged in a format AI can reuse immediately.

Real consequences of missing GEO

These aren't hypothetical consequences — this is what's happening now:

• ChatGPT answers "which companies do IoT in Can Tho?" by synthesizing from sources with clear structured data. A website without Organization schema doesn't exist in that answer, even if the company genuinely does IoT in Can Tho.
• Perplexity AI Citations surfaces 3–5 sources for each answer. Sources are selected based on relevance, authority, and most critically: structured data that Perplexity can parse. A blog post with no Article schema, no author, and no date is skipped.
• Google AI Overviews is expanding in Vietnam. A controlled experiment by Search Engine Land (September 2025) on three nearly identical pages — same content, same keyword difficulty — found that only the page with properly implemented FAQ schema appeared in a Google AI Overview, making schema markup the single deciding variable in that test.
• Gartner forecasts that traditional search engine volume will drop 25% by 2026 as users shift to asking AI assistants (ChatGPT, Gemini, Claude, Perplexity) directly instead of typing keywords into a search engine. But AI citation traffic will create a new segment — AI-referred traffic with higher conversion rates because users arrive already pre-qualified by AI.

Quick audit checklist

SEO fundamentals (the foundation):

• [ ] Does every page have a canonical URL?
• [ ] Is meta title + description unique per page?
• [ ] Does sitemap.xml update automatically when new pages are added?
• [ ] Does robots.txt avoid accidentally blocking major crawlers?
• [ ] Core Web Vitals: LCP < 2.5s, CLS < 0.1?

GEO / AI-Ready (the layer above):

• [ ] Is there an /llms.txt file describing the site and listing important pages?
• [ ] Does robots.txt explicitly allow GPTBot, PerplexityBot, ClaudeBot?
• [ ] Does Organization schema include name, url, description, contactPoint, knowsAbout?
• [ ] Does WebSite schema include a SearchAction (potentialAction)?
• [ ] Do blog posts have Article schema with author, datePublished, wordCount?
• [ ] Do blog posts show an author byline in the UI?
• [ ] Do service pages have FAQ or HowTo schema?

You don't have to start over

The good news: most SEO + GEO improvements are additive — you add them on top of what exists, not rebuild from scratch. A Next.js or WordPress site can add all the structured data, llms.txt, and robots.txt configuration in a few days of work, with no frontend changes and no architecture rebuild.

The bad news: every month of delay is a month AI engines learn to not know you exist — and once AI has built a "mental model" of an industry based on your competitors' data, inserting yourself later is significantly harder than being present from the start.

At KonexForge's Development layer, SEO + GEO audit is a mandatory part of every Pilot Build — not a feature added at the end, but an output requirement from day one.