SEO & GEO
What GEO Actually Means (and Why Most "AI SEO" Advice Is Wrong)
Pranjal Kukreja · May 12, 2026 · 7 min read
"GEO" has become shorthand for "do SEO, but mention ChatGPT." That's not what it is, and treating it that way is why most GEO advice doesn't move the needle. Generative Engine Optimization is a specific, mechanical discipline with its own technical requirements, separate from (though overlapping with) classic SEO. Here's what actually changes.
Search engines rank pages. Generative engines retrieve passages.
A traditional search engine's job is to rank a list of URLs against a query. The unit it operates on is the page. A generative engine, ChatGPT, Perplexity, Google AI Overviews, works differently: it retrieves relevant passages from a corpus, synthesizes them into an answer, and optionally cites a source. The unit it operates on is the passage, sometimes a single paragraph or even a sentence.
This matters because a page can rank well on Google while being nearly invisible to a generative engine, if the specific fact or answer a user is asking about is buried inside a wall of unstructured text, there's nothing for the retrieval system to cleanly pull out and cite. Content built for GEO gets restructured so individual passages can stand alone and answer a specific question, without requiring the reader to have scrolled through three paragraphs of preamble first.
llms.txt is a real thing, and most sites don't have one
llms.txt is an emerging convention, modeled loosely on robots.txt, that tells AI crawlers what a site is, what its most important content is, and where to find it in a clean, model-readable format. It's not universally adopted yet and it's not a guaranteed ranking factor, but it costs very little to implement and gives AI crawlers a direct signal instead of forcing them to infer structure from a rendered page. We treat it as table stakes, not an optional extra.
Entity consistency matters more than backlinks here
Classic SEO trust signals lean heavily on backlinks: other sites vouching for yours. Generative engines care more about entity consistency, whether the facts about your business (name, location, founding date, what you actually do, who runs it) are stated the same way across your own site and third-party sources like directories, press mentions, and structured profiles. Conflicting information across sources doesn't just confuse a human reader, it actively degrades a model's confidence in citing you at all.
Structured data isn't decoration, it's a parallel content layer
Schema markup (Organization, Product, Article, FAQ, and similar types) gives a model a machine-readable version of your content sitting alongside the human-readable version. Most sites either skip it entirely or implement it inconsistently across templates. For GEO purposes, we treat schema as a second, parallel content deliverable, not a technical checkbox to tick once and forget.
What we actually track
Measurement here is genuinely less mature than classic SEO, and we're upfront about that with clients. We track whether AI assistants reference or cite a business when asked relevant, realistic questions, alongside the classic Search Console metrics (clicks, impressions) that still matter because Google's own AI Overviews sit inside regular search results. On the Optima Bags case study, we saw a 5.6x increase in clicks and a 5.1x increase in impressions after this work started running, real Search Console numbers, not a GEO-specific metric, because the technical foundation (crawlability, structured data, content clarity) improves both channels at once.
If someone pitches you "AI SEO" without mentioning llms.txt, entity consistency, or passage-level content structure specifically, ask what they actually mean. There's a real, technical answer to that question, and "we'll mention AI in your content" isn't it.