{
  "schemaVersion": "1.0",
  "kind": "nilu.article",
  "id": "web-agent-ready-seo-vs-geo",
  "slug": "web-agent-ready-discovery-vs-citation",
  "locale": "en",
  "sourceLocale": "pt-br",
  "localizationStatus": "translated",
  "title": "Web agent-ready — discovery vs. citation: SEO, GEO, and evidence",
  "summary": "SEO is still the foundation; GEO has lab-grade method, but citation isn't a click or proof of return — what to actually measure.",
  "author": "NiLu",
  "publishedAt": "2026-08-07T00:00:00.000Z",
  "updatedAt": "2026-08-09T00:00:00.000Z",
  "canonicalUrl": "https://blog.nilu.pro/en/blog/web-agent-ready-discovery-vs-citation/",
  "representations": {
    "html": "https://blog.nilu.pro/en/blog/web-agent-ready-discovery-vs-citation/",
    "markdown": "https://blog.nilu.pro/en/blog/web-agent-ready-discovery-vs-citation.md",
    "json": "https://blog.nilu.pro/en/blog/web-agent-ready-discovery-vs-citation.json"
  },
  "bodyMarkdown": "# Web agent-ready — discovery vs. citation: SEO, GEO, and evidence\n\nSEO is still the foundation; GEO has lab-grade method, but citation isn't a click or proof of return — what to actually measure.\n\nAuthor: NiLu\nLocale: en\nPublished: 2026-08-07\nUpdated: 2026-08-09\nCanonical: https://blog.nilu.pro/en/blog/web-agent-ready-discovery-vs-citation/\nMarkdown: https://blog.nilu.pro/en/blog/web-agent-ready-discovery-vs-citation.md\nJSON: https://blog.nilu.pro/en/blog/web-agent-ready-discovery-vs-citation.json\n\n## In brief\n\n- SEO and GEO answer different questions: being found in a search engine is not the same job as being cited in a generative response.\n- For Google, optimizing for AI features in Search is still SEO — no special markup and no llms.txt requirement.\n- GEO with method exists in academic papers, but the reported gains (such as \"up to 40%\") come from the lab, not from a public ranking product.\n- Being cited is not the same as receiving a visit — CTR drops when an AI summary is present, and publishers' referral traffic is under pressure.\n\n## Contents\n\n- [SEO and GEO aren't the same job](#seo-e-geo-nao-sao-o-mesmo-job)\n- [What Google says about optimizing for AI](#o-que-o-google-diz-sobre-otimizar-para-ia)\n- [GEO with method: what the papers show](#geo-com-metodo-o-que-os-papers-mostram)\n- [Citation is not a click](#citacao-nao-e-clique)\n- [llms.txt and other \"AI SEO\" myths](#llmstxt-e-outros-mitos)\n- [Where marketing fits in](#onde-fica-o-marketing)\n- [Uncertainties](#incertezas)\n\n<a id=\"seo-e-geo-nao-sao-o-mesmo-job\"></a>\n\n## SEO and GEO aren't the same job\n\nSEO, in the classic sense, is about being discovered: crawling, indexing, and eligibility in search results. GEO — Generative Engine Optimization — emerged as an academic term for optimizing content toward visibility in systems that retrieve information and synthesize responses with attribution, a different job from holding a stable position on a results page. The paper that coined the term (Aggarwal et al.) describes generative engines this way: retrieval followed by synthesis by a language model, with inline citations. [[1]](#ref-1) A more recent paper, FeatGEO (ACL 2026), reinforces this distinction: answer engines perform selective citation, they don't allocate ranked SERP slots, and it proposes optimizing for interpretable structural and content features before rewriting text. [[2]](#ref-2)\n\nThe market, however, sells both names as if they were a brand-new discipline built purely for marketing. It's worth separating the academic method — which exists, with clear lab-grade limits — from the commercial pitch, which promises to \"rank in ChatGPT\" without showing how to measure it.\n\n<a id=\"o-que-o-google-diz-sobre-otimizar-para-ia\"></a>\n\n## What Google says about optimizing for AI\n\nFor the Google ecosystem — today the most documented case — the official answer is direct: optimizing for generative AI features in Search is optimizing for the Search experience, because these features are built on the same core quality and ranking systems. The guide describes how these features use grounding/RAG (retrieving from the Search index and generating with supporting links) and \"query fan-out,\" and lists among the myths to avoid: `llms.txt`, special AI markup, splitting content into \"chunks\" for AI, and rewriting text only for agents — structured data is not required for generative AI features. [[3]](#ref-3) The \"AI features and your website\" documentation confirms there is no additional technical requirement beyond being indexed and eligible for a snippet. [[4]](#ref-4) John Mueller's post on the Search Central Blog (May 2025) extends the same longstanding advice — content designed for people, good page experience, technical crawling and indexing requirements — to AI Overviews and AI Mode; structured data remains useful for machine-readable sharing when it matches the visible content, not as a citation trick. [[5]](#ref-5)\n\nThis doesn't mean AI citation is irrelevant — only that, at Google, it is not a separate technical discipline with its own requirements. Google's VP of Search, Liz Reid, described AI responses in Search as bringing visible links, source citation, and attribution embedded in the text, explicitly as reinforcing the web, \"not the web or AI — both.\" [[6]](#ref-6) Search Console also began offering performance reports for generative AI features (impressions by page, country, device, and date) — an official visibility measure, not a third-party \"citation\" score. [[7]](#ref-7)\n\n<a id=\"geo-com-metodo-o-que-os-papers-mostram\"></a>\n\n## GEO with method: what the papers show\n\nThe GEO paper reports a relative visibility gain \"of up to 40%\" — but that number comes from a benchmark of the authors' own making (GEO-bench, about 10 thousand queries), with a specific setup: Google's top-5 results reprocessed by a GPT-3.5-style pipeline. It is not a public ChatGPT ranking API, nor a number that automatically repeats itself in production. In the same experiment, stuffing text with keywords (keyword stuffing) performed poorly against strategies using statistics, quotes, and citations — a clue about the type of content that tends to get cited, not a formula. [[1]](#ref-1) FeatGEO uses a different benchmark (GEO-Bench) and reaches similar conclusions about structure and interpretable language. [[2]](#ref-2)\n\nNeither paper tests whether these results transfer to ChatGPT search, Perplexity, or Claude in production — that remains an open question. Treating \"up to 40%\" as a business target is the most common mistake: it is a bench result, valid within the scope of the experiment, not a prediction of traffic or revenue.\n\n<a id=\"citacao-nao-e-clique\"></a>\n\n## Citation is not a click\n\nEven when a site is cited or linked, that does not equal receiving a visit. The Pew Research Center, in a correlational analysis of Google users' behavior, found an organic-link click-through rate of about 8% when an AI summary appeared in the results, versus about 15% when it did not; only around 1% of users clicked any link inside the summary itself. [[8]](#ref-8) Digiday reporting, citing Digital Content Next data, reported publishers seeing a median of approximately -10% year-over-year in referral traffic from Google (with cases between -1% and -25% in the sample), associated with — not necessarily universally caused by — AI Overviews. [[9]](#ref-9) Another Digiday report, citing Similarweb, put zero-click news searches rising from about 56% to 69%, with growth in traffic from AI platforms insufficient to offset that loss. [[10]](#ref-10)\n\nThe quality of the citation itself is also a fragile metric: a 2023 study (Liu et al.) measured per-sentence citation recall at around 51,5% across the generative search engines evaluated at the time of the snapshot — nearly half of the generated sentences lacked an adequate verifiable citation. [[11]](#ref-11) For managers, the practical implication is to measure Search Console, real sessions, and conversion — not a synthetic \"mentions scoreboard.\"\n\n<a id=\"llmstxt-e-outros-mitos\"></a>\n\n## llms.txt and other \"AI SEO\" myths\n\nBeyond Google's mythbusting, two independent studies from SEO vendors reinforce the same skepticism about `llms.txt`: Ahrefs, in a sample of about 137 thousand sites, found that 97% of published `llms.txt` files received no requests at all in the measured month; SE Ranking, in a sample of about 300 thousand domains, found no observable correlation between publishing `llms.txt` and being cited by LLMs. [[12]](#ref-12) [[13]](#ref-13) \"No correlation\" is not definitive proof that the file never helps — but, today, it does not support treating it as a priority deliverable.\n\nRobots policy also deserves separation by job: OpenAI documents GPTBot (training collection), OAI-SearchBot (search indexing), and ChatGPT-User (real-time user-requested search) as distinct bots, with their own controls via `robots.txt`; Anthropic documents how to block its crawler. [[14]](#ref-14) [[15]](#ref-15) Blocking training is not the same decision as blocking search/citation — and edge networks such as Fastly report that AI crawlers already account for a relevant share of the observed bot traffic, which weighs on bandwidth and origin costs even before any decision about citation. [[16]](#ref-16)\n\n<a id=\"onde-fica-o-marketing\"></a>\n\n## Where marketing fits in\n\nPart of the commercial discourse around GEO promises figures that aren't backed by public method — for example, claims of conversion far above average, or of low overlap between chatbot citations and Google's top 10, without a published replicable methodology. This kind of claim appears here only as a portrait of the market, never as a basis for decisions. Trade-press reporting, citing veteran SEO professionals, describes much of the GEO discourse as traditional SEO dressed up as a new discipline [[17]](#ref-17) — a real market tension, not a scientific verdict. Both the optimistic and the skeptical claims appear here as perspective, not as fact: the factual basis of this text remains Google's official documentation, the papers with method, and the independent studies cited above.\n\n<a id=\"incertezas\"></a>\n\n## Uncertainties\n\n- There is no public list of the exact signals Google uses to choose which pages appear as support in AI Overviews, beyond the core quality systems.\n- It is not known how (or whether) the GEO-bench and FeatGEO results transfer to ChatGPT search, Perplexity, or Claude in production.\n- There is no public study that converts synthetic \"mention\" or \"AI visibility\" metrics into publishers' real impressions, clicks, or revenue.\n- It is not possible to calculate, with the available data, the net marginal cost of allowing or blocking each bot (training vs. search) for a specific site.\n\n## References\n\n1. <a id=\"ref-1\"></a>[Aggarwal et al. — GEO: Generative Engine Optimization](https://arxiv.org/abs/2311.09735). arXiv. Accessed August 07, 2026.\n2. <a id=\"ref-2\"></a>[Liu & Xu — FeatGEO](https://aclanthology.org/2026.acl-long.929/). ACL Anthology. Accessed August 07, 2026.\n3. <a id=\"ref-3\"></a>[Google Search Central — Optimizing your website for generative AI features on Google Search](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide). Google. Accessed August 07, 2026.\n4. <a id=\"ref-4\"></a>[Google Search Central — AI features and your website](https://developers.google.com/search/docs/appearance/ai-features). Google. Accessed August 07, 2026.\n5. <a id=\"ref-5\"></a>[Google Search Central Blog — Succeeding in AI search](https://developers.google.com/search/blog/2025/05/succeeding-in-ai-search). Google. Accessed August 07, 2026.\n6. <a id=\"ref-6\"></a>[Google — AI in Search: Going beyond information to insight](https://blog.google/products-and-platforms/products/search/ai-search-driving-more-queries-higher-quality-clicks/). Google. Accessed August 07, 2026.\n7. <a id=\"ref-7\"></a>[Google Search Central Blog — New Search Generative AI performance reports in Search Console](https://developers.google.com/search/blog/2026/06/gen-ai-performance-reports). Google. Accessed August 07, 2026.\n8. <a id=\"ref-8\"></a>[Pew Research Center — Google users are less likely to click on links when an AI summary appears in the results](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/). Pew Research Center. Accessed August 07, 2026.\n9. <a id=\"ref-9\"></a>[Digiday — Google AI Overviews linked to publisher referral traffic drops, new data shows](https://digiday.com/media/google-ai-overviews-linked-to-25-drop-in-publisher-referral-traffic-new-data-shows/). Digiday. Accessed August 07, 2026.\n10. <a id=\"ref-10\"></a>[Digiday — In graphic detail: AI platforms are driving more traffic, but not enough to offset zero-click search](https://digiday.com/media/in-graphic-detail-ai-platforms-are-driving-more-traffic-but-not-enough-to-offset-zero-click-search/). Digiday. Accessed August 07, 2026.\n11. <a id=\"ref-11\"></a>[Liu et al. — Evaluating Verifiability in Generative Search Engines](https://arxiv.org/abs/2304.09848). arXiv. Accessed August 07, 2026.\n12. <a id=\"ref-12\"></a>[Ahrefs — We Analyzed 137K Sites: 97% of llms.txt Files Never Get Read](https://ahrefs.com/blog/llmstxt-study/). Ahrefs. Accessed August 07, 2026.\n13. <a id=\"ref-13\"></a>[SE Ranking — llms.txt vs LLM citations analysis](https://seranking.com/blog/llms-txt/). SE Ranking. Accessed August 07, 2026.\n14. <a id=\"ref-14\"></a>[OpenAI — Crawler / bot documentation](https://developers.openai.com/api/docs/bots). OpenAI. Accessed August 07, 2026.\n15. <a id=\"ref-15\"></a>[Anthropic — Does Anthropic crawl data from the web, and how can site owners block the crawler?](https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler). Anthropic. Accessed August 07, 2026.\n16. <a id=\"ref-16\"></a>[Fastly — New Fastly threat research reveals AI crawlers make up almost 80% of AI bot traffic](https://www.fastly.com/press/press-releases/new-fastly-threat-research-reveals-ai-crawlers-make-up-almost-80-of-ai-bot). Fastly. Accessed August 07, 2026.\n17. <a id=\"ref-17\"></a>[Digiday — GEO hype busted: How it differs (and how it doesn't) from SEO](https://digiday.com/media/geo-hype-busted-experts-call-it-more-seo-than-new-discipline/). Digiday. Accessed August 07, 2026.\n"
}