# Citeable > Citeable makes any website citable by AI answer engines (ChatGPT, Perplexity, Gemini, Claude). You paste your site URL: Citeable crawls it, then generates the files — a structured llms.txt, an llms-full.txt, Q&A markup in schema.org format (JSON-LD), and an AI-friendly robots.txt — that you drop onto your site so AI engines can read, understand, and cite it. One-time payment per site, no subscription. --- ## What Citeable is Source: https://citeable.eu/en Citeable is a tool that makes a website citable by AI answer engines. It analyzes a site's public content, then generates the files you add to your site: an llms.txt, an llms-full.txt, and Q&A markup in schema.org format (JSON-LD). Together they give ChatGPT, Perplexity, Gemini and Claude a clean, structured version of who you are and what you offer, so they can read, understand and cite the site directly in their answers. This matters because AI answer engines quote only a handful of sources per question: if your pages aren't machine-readable, you're skipped even when you're the best result. Citeable works on any public site — WordPress, Shopify, Framer, custom — needs no access to your admin, and shows a before/after citability score so you can see the gain. It costs from EUR 39 as a one-time payment, with no subscription, and your files stay regenerable for life. How it works: (1) You paste your site URL on the home page. (2) Citeable crawls the site's public pages — no admin access, no changes to the site. (3) Citeable generates the files: a clean llms.txt, an llms-full.txt with the full text of your key pages, and Q&A schema.org markup (JSON-LD). (4) You drop the files onto your own site. Result: a site that ChatGPT, Perplexity, Gemini, and Claude can read, understand, and cite. What makes Citeable different: Free llms.txt generators produce a minimal file from the sitemap. Citeable crawls the real content AND adds Q&A schema.org markup — that second part is what actually makes a site citable. It works on any public site: WordPress, Wix, Shopify, Framer, or fully custom builds. No subscription: one-time payment per pack of sites, files regenerable for life via the payment email. --- ## Pricing Source: https://citeable.eu/en#pricing - Solo — EUR 39 (one-time): 1 site, llms.txt + llms-full.txt + robots.txt, Q&A markup (schema.org), lifetime regeneration. - Pro — EUR 89 (one-time): 3 sites, everything in Solo, priority support. Recommended plan. - Agency — EUR 199 (one-time): 10 sites, everything in Pro, white-label client report. --- ## Frequently asked questions Source: https://citeable.eu/en#faq ### Does Citeable touch my site? No, never. Citeable only reads your site's public pages — the same way any visitor or search engine would — then generates two files that you deploy yourself: an llms.txt and Q&A markup in schema.org format. At no point do we ask for access to your hosting, your back office or your database, and we don't change a single line of your existing site. You stay fully in control: you review the generated files, then choose whether to publish them or not. That's also why Citeable works whatever your tool is — WordPress, Shopify, Framer, Webflow or a fully custom site — with no plugin or integration to install. ### What if I lose my files later? You can't lose them permanently. Your files are regenerable for life, at no extra cost. If you switch computers, delete a folder or rebuild your site, just enter the email you used at payment on the “Recover my files” page: we send you a secure link to recreate them identically. You can also sign in to your Citeable account (by magic link, no password) to re-download, regenerate or add a site at any time. Your payment unlocks permanent access, not a throwaway file — the whole point is that a customer coming back months later never feels forced to pay again. ### What kind of site does it work on? Any public site, whatever technology produced it: WordPress, Wix, Squarespace, Shopify, Framer, Webflow, Notion or a hand-coded site. There's nothing to install on the site side — no plugin, no extension, no admin access. You simply deploy the two generated files: the llms.txt at the root of your domain, and the Q&A markup in the of your homepage. If you can add a file to your site and paste a snippet into the header (which every modern host allows), you can use Citeable. The only cases it can't cover are pages behind a login or paywall, which AI engines can't read anyway. ### There are free llms.txt generators, why pay? Because llms.txt alone isn't enough to make you citable. Free generators produce a minimal llms.txt, often just a list of links pulled from your sitemap. Citeable does two more things. First, we crawl your real content to write a structured llms.txt — title, summary, sections, useful links — that AI engines actually understand. Second, and most importantly, we generate the Q&A markup in schema.org format: tagged questions and answers that ChatGPT, Perplexity or Gemini can quote word-for-word in their responses. That Q&A layer, absent from free tools, is what turns “readable” into “citable.” You also get a before/after score and files regenerable for life, for a single payment. ### Is it really a one-time payment, no subscription? Yes, a single payment, with no recurring charge at all. Each pack unlocks a number of sites (1, 3 or 10 depending on the plan) and lifetime access to their files: you can regenerate and re-download them as often as you like, never paying again. There's no trial that turns into a subscription, no card to watch, no automatic renewal. We made that choice deliberately: making a site citable is a one-off action, not a service to rent by the month. If one day you want to make a new site citable beyond your pack, you buy an extra credit — when you actually need it, not before. Your files are yours. --- ## What is Generative Engine Optimization (GEO)? A plain-language guide Source: https://citeable.eu/en/guides/what-is-generative-engine-optimization ### What is Generative Engine Optimization (GEO)? Generative Engine Optimization (GEO) is the practice of structuring a website's content so it gets read, understood and cited by AI answer engines such as ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews. Where classic SEO optimizes to appear in a list of blue links, GEO optimizes to become the source an AI quotes and attributes inside its answer. The term was popularized by a 2023 academic study (Aggarwal et al., presented at KDD 2024) that measured, across 10,000 queries, which content modifications increase a source's visibility inside generated answers. One clarification: bare 'GEO' is ambiguous (geography, the GEO Group stock, genomics databases) — in a marketing context it always means generative engine optimization. ### What's the difference between GEO and SEO? SEO and GEO share a common base (a crawlable, indexed, fast, well-structured site) but aim at two different outcomes. SEO optimizes ranking in a list of links: the goal is the click. GEO optimizes inclusion in a generated answer: the goal is the citation. Three concrete differences. First, the unit: SEO thinks at the page and domain level (authority, backlinks), GEO thinks at the passage level — a self-contained paragraph that answers a specific question can be cited even from a small site. Second, the winning format: SEO rewards keywords and links, GEO rewards verifiable facts, sourced statistics and direct answers. Third, the channel: SEO targets Google's SERP, GEO also targets ChatGPT, Perplexity and Claude, which have no 'page one' but pick 3-4 sources. GEO does not replace SEO — it stacks on top, because AI engines often retrieve their candidates from classic search indexes. A site invisible in SEO stays invisible in GEO. ### Why does GEO matter now? Because clicks are moving from links to answers, and the numbers are steep. Seer Interactive measured a 61% drop in organic click-through rate when a Google AI Overview is present; Ahrefs measured a 58% reduction in clicks on the top result; Pew Research Center found users click a traditional link roughly half as often when an AI summary appears. The flip side is the opportunity: brands cited inside AI answers capture a disproportionate share of the remaining attention. If the AI answers on the user's behalf and cites only a few sources, being one of those sources is no longer a bonus — it is the condition for existing. And because answer-engine adoption is recent, the field is still lightly contested: sites that start early gain a measurable head start. ### Which GEO techniques actually work? The reference GEO study (Aggarwal et al., KDD 2024) tested nine strategies across 10,000 queries. The winners are concrete and reproducible: adding statistics improved a source's visibility by around 41%, adding quotations from sources by around 28%, and citing credible external sources also produced clear gains — with the biggest lift for sites that don't already rank first. Keyword stuffing, the classic SEO reflex, did nothing or hurt. In practice this means five levers: (1) explicitly allow AI crawlers in robots.txt; (2) publish an llms.txt file at your root, a structured summary of your site; (3) add Q&A schema.org markup (JSON-LD) exposing machine-readable question-and-answer pairs; (4) structure pages as questions with direct, self-contained 40-160 word answers backed by real facts and numbers; (5) show an accurate updated date and dateModified, because freshness is a signal. The first three levers are mechanical and fast; the last two are editorial and make the durable difference. ### GEO, AEO, SEO — are they the same thing? They overlap but are not identical. SEO (Search Engine Optimization) targets ranking in classic search engines. AEO (Answer Engine Optimization) targets providing the direct answer to a question — historically Google's featured snippets, now extended to AI answers. GEO (Generative Engine Optimization) specifically targets inclusion and citation inside content generated by large language models. In practice, many people use GEO and AEO interchangeably, because the techniques overlap heavily: direct answers, structured content, schema.org data, verifiable facts. The useful distinction is not terminological but strategic: are you optimizing for a click from a list (SEO), a boxed answer in a SERP (AEO), or a citation inside an AI conversation (GEO)? For most sites, the right move is to treat all three as a continuum — apply the shared fundamentals, then add the signals specific to generative engines. ### How do you start with GEO today? Start with an audit: are your pages crawlable by AI bots, do you have an llms.txt, are your answers structured and sourced? Then apply the levers in increasing order of effort — robots.txt, llms.txt, schema.org markup, editorial restructuring, freshness signals. Two ways to do it: manually, following the llms.txt specification and writing the JSON-LD yourself, which takes time and needs upkeep; or automatically, with a tool that crawls your site and generates the files. That is exactly what Citeable does: you paste your URL, it analyzes your public pages and generates both your llms.txt and your Q&A schema.org markup, consistent with each other, for a one-time payment and regenerable for life. Honest scope: nobody controls the answer engines, so no tool or agency can guarantee a citation — it is a best-efforts obligation. What GEO guarantees is that your site is as readable and citable as possible; the study numbers explain why that is worth doing. --- ## How to get cited by ChatGPT: a step-by-step guide Source: https://citeable.eu/en/guides/how-to-get-cited-by-chatgpt ### How does ChatGPT decide which sites to cite? In two stages. First it retrieves candidate pages: ChatGPT leans heavily on Bing's index and on live browsing. At this stage your site must be crawlable, indexed and readable without JavaScript. Then it selects, among the retrieved pages, the passages that best answer the question, and cites their sources. This second stage happens at the passage level, not the domain level: a self-contained paragraph that answers the question directly can be cited even from a small site. The practical consequence: you're not trying to 'rank first', you're trying to be the easiest passage to lift. Write each paragraph so it stands on its own and answers one question, and you match exactly what the selection stage looks for. ### What are the concrete steps to get cited by ChatGPT? Seven steps, from the most mechanical to the most editorial. One: allow GPTBot, OAI-SearchBot and ChatGPT-User in your robots.txt (blocking them removes you from the pool). Two: publish an llms.txt file at your root, a structured summary of your site. Three: add Q&A schema.org markup as JSON-LD on your key pages. Four: structure your pages as real questions with direct, self-contained 40-160 word answers — the first sentence should answer, not introduce. Five: back every important claim with a fact, a number or a verifiable source; the GEO study shows statistics and citations clearly increase visibility. Six: keep your content fresh, with a visible updated date. Seven: make sure content is server-rendered, because several AI crawlers don't execute JavaScript. These seven steps guarantee nothing — nobody controls ChatGPT — but they align your site with everything the engine can recognize and lift. ### What kind of content does ChatGPT lift the most? Specific, verifiable, self-contained passages. Concretely: an answer that starts by answering ('No, an llms.txt is not enough to get cited, because…') rather than by setting context; a claim backed by a number and its source; a paragraph that makes sense on its own, without having read the rest of the page. Princeton's GEO study and Search Engine Land's analyses converge: statistics, source citations and direct phrasing are the traits that most increase the odds of being lifted. Conversely, vague marketing ('the best solution on the market'), keyword stuffing and long unstructured blocks are ignored. The right mental image: write as if an AI assistant will copy-paste a single one of your paragraphs into its answer and attribute it to your site. That paragraph must be accurate, self-contained and attributable. If each of your paragraphs passes that test, you're optimized for citation. ### Can anyone guarantee you'll be cited by ChatGPT? No, and be wary of anyone who promises it. Nobody controls ChatGPT: its answers vary by question, by day, by model version and by mode (live search or internal knowledge). What you do control is being in the candidate pool and being the easiest source to cite. Mechanical changes — robots.txt, llms.txt, schema.org markup — take effect as soon as engines re-crawl your site, often within days to a few weeks. To measure, keep it simple: ask the engines the questions your customers actually ask and see who they cite, then repeat regularly. That is also Citeable's logic: a best-efforts obligation, not a guarantee of results. The tool takes your URL, crawls your site and generates both your llms.txt and your Q&A schema.org markup, consistent with each other, to make you as citable as possible — the rest depends on the engines, which nobody controls. --- ## llms.txt generator: how they work and what to look for Source: https://citeable.eu/en/guides/llms-txt-generator ### What is an llms.txt generator? An llms.txt generator is a tool that automatically creates a site's llms.txt file: a structured Markdown summary, placed at the root of the domain, that AI answer engines (ChatGPT, Perplexity, Gemini, Claude) read to understand who you are and where your important pages live. There are two families of tools. Minimal generators build the file from your sitemap: they list your URLs with their titles, quickly and for free, but without a real summary of your content. Full generators crawl your actual pages, extract the content and write a file with a title, a summary, thematic sections and annotated links — much closer to what the llms.txt specification recommends. The difference matters: an llms.txt that mirrors your sitemap adds almost nothing for an AI engine, whereas a file built on your real content gives it a clean, usable version of your site. ### Should you generate llms.txt by hand or with a tool? Both work; the right choice depends on site size and available time. By hand: you read the specification at llmstxt.org and write the Markdown yourself — an H1 title, a summary blockquote, sections of links with a one-line description each. For a small site, budget an hour or two; the real difficulty isn't technical but editorial, because a good llms.txt needs genuine summaries of your real content, not a plain URL list. You also have to keep it updated as the site changes. With a tool: you paste your URL and the generator produces the file in minutes by crawling your public pages. It's faster, consistent, and regenerable when your site changes. The deciding criterion isn't 'free or paid' but 'minimal or full': a tool that just lists your sitemap is no better than doing it by hand, whereas a tool that crawls your content and adds schema.org markup does work few people do well by hand. ### Is an llms.txt generator enough to get cited? No, and it's the most common trap. The llms.txt makes your site readable — it gives the engine a clean summary of who you are — but it does not, by itself, make your content quotable word for word. AI answer engines cite passages that directly answer a specific question, and that depends on the shape of your pages, not just a file at the root. You need two things together: the llms.txt for discovery and comprehension, and Q&A schema.org markup (JSON-LD) on your pages to expose question-and-answer pairs the machine can lift as-is. A generator that only produces the llms.txt optimizes half the path. That's why a good tool generates both artifacts from the same crawl, guaranteeing they're consistent with each other — the summary and the markup tell the same story, with no drift. ### How do you generate an llms.txt (and Q&A markup) with Citeable? Citeable is a full generator: you paste your site's URL, it crawls your public pages with no admin access and no changes, then from that same crawl it generates the two files that make a site citable — a structured llms.txt following the specification and Q&A schema.org markup as JSON-LD, ready to drop in. Because both come out of the same analysis, they're consistent by construction. You also get a before/after citability score to see the gain. Payment is one-time, per pack of sites, with no subscription, and the files are regenerable for life as your site evolves. Honest scope: nobody controls the answer engines, so the tool doesn't guarantee a citation — it guarantees your site is as readable and citable as possible, which is exactly what the llms.txt and Q&A markup are for. --- ## GEO vs SEO vs AEO: the difference, and which to prioritize Source: https://citeable.eu/en/guides/geo-vs-seo-vs-aeo ### SEO, AEO, GEO — what's the one-line difference for each? SEO (Search Engine Optimization) optimizes to rank in a list of links on a search engine: the goal is the click. AEO (Answer Engine Optimization) optimizes to provide the direct answer to a question — historically Google's featured snippets, now also assistant answers: the goal is to be the boxed answer. GEO (Generative Engine Optimization) optimizes to be cited inside an answer generated by an AI like ChatGPT, Perplexity, Gemini or Claude: the goal is the attributed citation. All three share a common base — a crawlable, indexed, fast, well-structured site — but aim at three different moments of the journey: appearing in a list, occupying the answer box, or being the source an AI lifts. None replaces the others; they form a continuum where the fundamentals are shared and the specific signals stack on top. ### Comparison table: SEO vs AEO vs GEO Across five key dimensions: Goal — SEO targets the click, AEO targets the boxed answer, GEO targets the citation in an AI answer. Surface — SEO targets Google's list of links, AEO featured snippets and People Also Ask, GEO the answers of ChatGPT, Perplexity, Gemini, Claude and AI Overviews. Optimized unit — SEO works the page and domain, AEO the answer block for a question, GEO the self-contained citable passage. Main levers — SEO: keywords, backlinks, domain authority; AEO: structured data, concise answers, question-answer format; GEO: llms.txt, Q&A schema.org markup, sourced facts, freshness. Success metric — SEO: position and CTR; AEO: presence in the box; GEO: presence and fidelity of the citation. In practice GEO and AEO overlap so much that many use them as synonyms; the real dividing line is between optimizing for a click (SEO) and optimizing to be the answer (AEO + GEO). ### Do they overlap, and do you have to choose? They overlap heavily and you don't have to choose: the fundamentals are shared, so doing one well already advances the others. A crawlable, indexed site serves SEO and the retrieval stage of GEO alike. Direct, structured answers serve AEO and GEO citability alike. Schema.org data serves all three. The right strategy isn't to pit the acronyms against each other but to build in layers: first the technical fundamentals (crawl, speed, structure, server-rendered HTML), then structured data and direct answers, then the signals specific to generative engines (llms.txt, Q&A markup, sourced facts, freshness). What you prioritize depends on where your traffic comes from and goes: if your prospects increasingly find you through AI answers rather than blue links, GEO rises in your priorities — but never by abandoning the SEO fundamentals that gate everything else. ### Where should you start, given your goal? If your traffic mostly comes from Google's classic links, start with SEO: technical fundamentals, quality content, authority. If you're chasing featured snippets and position zero, add AEO: question-answer structure, structured data, concise answers. If you notice ChatGPT, Perplexity or Gemini answering on your behalf and citing your competitors, prioritize GEO: allow AI crawlers, publish an llms.txt, add Q&A schema.org markup, and rewrite your key pages as direct, sourced answers. The good news is these efforts don't cannibalize each other — each layer reinforces the others. For the most mechanical part of GEO, a tool like Citeable generates the llms.txt and Q&A schema.org markup from your real content in minutes, consistent with each other, which hands you the most technical step and lets you focus on the editorial side. --- ## How to get cited by Perplexity AI Source: https://citeable.eu/en/guides/how-to-get-cited-by-perplexity ### How does Perplexity select its sources? Perplexity works like a search engine paired with a language model: for each question it runs a real-time search, retrieves a handful of pages, then writes an answer citing its numbered sources explicitly. Two gates to pass. First gate, retrieval: your page must be indexed and accessible to PerplexityBot and Perplexity-User — if they're blocked in your robots.txt, you don't exist to it. Second gate, selection: among the retrieved pages, Perplexity picks the ones whose passage answers the question directly and reliably. Unlike a classic engine, it doesn't reward domain authority alone: a small site whose paragraph answers cleanly can be cited ahead of a big vague site. This two-gate system is what makes Perplexity attackable even for a recent site — provided you're accessible and have crisp passages. ### Which signals does Perplexity value most? Three, more pronounced than in other engines. Freshness: Perplexity often shows the date of its sources and favors recent content, so a real updated date and an accurate dateModified weigh more. Reciprocal sources: Perplexity favors pages that themselves cite credible sources — a guide backed by studies or official data earns trust and gets cited more easily. Extractable structure: Perplexity readily lifts lists, tables, definitions and short, crisp answers, because these are easy units to insert into an answer. In practice, write each section as a self-contained 40-160 word answer that starts by answering, back it with a fact or source, and keep your content current. These three signals compound: recent, sourced, structured content ticks exactly the boxes Perplexity knows how to recognize. ### How do you concretely prepare your site for Perplexity? Four actions, in order. One: allow PerplexityBot and Perplexity-User in your robots.txt, and check that your key pages are indexed and server-rendered. Two: publish an llms.txt at your root, to give Perplexity a clean summary of your site and important pages. Three: add Q&A schema.org markup (JSON-LD) on your pages, to expose question-and-answer pairs Perplexity can lift as-is. Four: restructure your key pages as real questions with direct, sourced, dated answers. These four actions cover both gates — accessibility for retrieval, crisp fresh passages for selection. For the technical steps (llms.txt and Q&A markup), a tool like Citeable generates them from your real content in minutes, consistent with each other, which lets you focus on the editorial quality of the answers. Honest reminder: nobody controls Perplexity, so the goal is to maximize your chances, not to guarantee a citation. --- ## What is llms.txt? The file that lets AI engines read your site Source: https://citeable.eu/en/guides/what-is-llms-txt ### What is an llms.txt file? llms.txt is a plain-text file, written in Markdown, placed at the root of a website, that gives AI systems a clean, curated summary of what the site is about: who is behind it, what it offers, and where its most important pages are. It is an open standard proposed in September 2024 by Jeremy Howard, co-founder of Answer.AI; the specification is published at llmstxt.org. The problem it solves is simple: language models have limited context windows, and a typical web page buries the actual content under navigation, scripts and cookie banners. llms.txt hands the model the essential version directly. Where robots.txt talks to crawlers about permissions, llms.txt talks to language models about meaning. For an AI answer engine, it is the shortest path to understanding a site correctly. ### Where does the file go, and what does it contain? The file lives at the root of your domain, at /llms.txt — for example https://citeable.eu/llms.txt. Its format is deliberately simple Markdown: a single H1 with the site's name, a blockquote summarizing what the site does, then H2 sections containing lists of links to key pages, each with a one-line description. The specification also allows an optional section, conventionally named 'Optional', for secondary links a model can skip when its context is tight. A companion convention, llms-full.txt, goes further and inlines the full text of the important pages into one file. The essential rule: it is written for machines that read like fast, literal humans — clear naming, a real summary, and links that point to pages actually worth reading. ### Do AI engines actually read llms.txt? Partly — and honesty matters here. llms.txt is a proposed standard: neither OpenAI nor Google has officially committed to using it for ranking or citations. What is verifiable today: AI crawlers such as GPTBot, ClaudeBot and PerplexityBot do request /llms.txt on sites that publish one, and a growing number of companies — Anthropic, Zapier, Cloudflare, and most documentation platforms such as Mintlify — publish the file. It is also immediately useful for live browsing: when ChatGPT, Perplexity or Claude fetch your site in real time to answer a question, a clean llms.txt is the shortest path to being understood correctly rather than paraphrased from a noisy HTML page. The cost-benefit is lopsided in your favor: one static text file, no downside, and a head start if adoption keeps growing. ### What's the difference between llms.txt, robots.txt and sitemap.xml? They answer three different questions. robots.txt handles permission: which bots may crawl which parts of your site — it says nothing about content. sitemap.xml handles inventory: the list of URLs you want indexed, with dates — useful for coverage, silent about meaning. llms.txt handles comprehension: what your site is about, in a form a language model can load in one pass and actually understand. The three are complementary, not competing. A site that wants to be visible to AI engines should have all three: a robots.txt that explicitly allows AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended), a sitemap for classic indexing, and an llms.txt for meaning. Removing any one of them weakens a different link in the chain: access, discovery, or understanding. ### Is llms.txt enough to get cited by AI engines? No — and that is the most common misconception. llms.txt makes your site readable; it does not, by itself, make it quotable. AI answer engines cite passages that answer a specific question directly, in a self-contained way. That is a content-shape problem: pages structured as questions with direct answers, plus structured data — Q&A markup in schema.org format (JSON-LD) — that exposes those question-and-answer pairs in a machine-readable form on your pages. That combination is what turns 'the AI can read me' into 'the AI can quote me word for word'. It is also exactly the pair Citeable generates: a clean llms.txt built from your real content, plus the Q&A schema.org markup to place on your pages. ### How do you create an llms.txt file? Two ways. Manually: read the specification at llmstxt.org, write the Markdown yourself — one H1, a blockquote summary, sections of links with a one-line description each — and keep it updated when your site changes. For a small site with a handful of pages, budget an hour or two; the hard part is editorial: an llms.txt that merely mirrors your sitemap adds nothing, the value is in genuine summaries of real content. Automatically: Citeable does it from a URL — it crawls your public pages, extracts the actual content, and generates both a structured llms.txt and the Q&A schema.org markup, for a one-time payment. Either way: publish the file at /llms.txt, check that it loads, and reference your best pages, not all of them. --- ## How do AI engines pick which sites to cite? Source: https://citeable.eu/en/guides/how-ai-engines-pick-sites-to-cite ### How do AI engines decide which sites to cite? In two stages. First, retrieval: the engine gathers candidate pages, either from a search index (ChatGPT leans heavily on Bing, Gemini on Google) or by fetching pages live at answer time. At this stage, classic visibility matters: your site must be crawlable, indexed and readable. Second, selection: the language model reads the retrieved pages and picks the passages that best answer the user's question, then cites their sources. This stage is passage-level, not site-level — the model quotes the paragraph that answers, not the best-known domain. That two-stage funnel is the whole playbook: being accessible and understandable gets you into the candidate pool; having self-contained passages that answer real questions directly is what converts candidacy into citations. Miss the first stage and you are invisible; miss the second and you are read but never quoted. ### Why does being cited matter so much now? Because clicks are moving from links to answers, and the numbers are steep. Seer Interactive measured organic click-through rates across thousands of informational queries: when a Google AI Overview is present, organic CTR drops by 61%. Ahrefs measured the same effect on the top-ranking result: a 58% reduction in clicks. Pew Research Center found Google users click a traditional link roughly half as often when a search returns an AI summary. The flip side is the opportunity: Seer's data also shows brands cited inside AI answers earn about 35% more organic clicks than brands that are not. The traffic is not disappearing uniformly — it is being rerouted to the handful of sources the answer engines quote. The game is no longer to rank among ten blue links; it is to be the source the answer cites. ### What kind of content do AI engines prefer to cite? The best public evidence is the GEO study (Aggarwal et al., KDD 2024), which tested nine optimization strategies across 10,000 queries sent to generative engines. The winners were concrete: adding statistics improved a source's visibility in AI answers by around 41%, adding quotations from sources by around 28%, and citing credible external sources also produced clear gains — up to 30-40% combined, with the biggest lift for sites that do not already rank first. Keyword stuffing, the classic SEO reflex, did nothing or hurt. The pattern behind those numbers: answer engines favor passages that are specific, verifiable and self-contained — a claim with a number and a source is easier to quote confidently than a vague marketing sentence. Write paragraphs that each answer one question, directly, with evidence, and you match exactly what the selection stage is looking for. ### Do classic SEO signals still matter? Yes, at the retrieval stage — with one important twist. AI engines mostly discover content through search indexes and their own crawlers, so the fundamentals still gate everything: crawlable pages, indexation, decent titles, server-rendered HTML (several AI crawlers execute little or no JavaScript), reasonable load times. The twist: the selection stage is passage-level, so domain authority buys less than it does in classic SEO. A small site whose paragraph answers the question directly can be cited ahead of a big site whose page merely mentions the topic — that is the opportunity for independents and small businesses. Watch the reverse, though: blocking AI crawlers in robots.txt removes you from the pool entirely. GPTBot, ClaudeBot, PerplexityBot and Google-Extended each obey their own user-agent rules, and many sites block them without realizing it. ### What should you change on your site to get cited more? Five things, in increasing order of effort. One: explicitly allow AI crawlers in your robots.txt — one line per bot. Two: publish an llms.txt at your root, so engines get a clean summary of who you are and what you offer. Three: add structured data — Q&A markup in schema.org format (JSON-LD) that exposes machine-readable question-and-answer pairs on your key pages. Four: reshape those pages as questions with direct, self-contained answers of roughly 40 to 160 words, backed by real facts, numbers and dates. Five: show a visible updated date and an accurate dateModified — freshness is a signal answer engines use. The first three are mechanical and fast; the last two are editorial and make the durable difference. ### Can anyone guarantee you'll be cited by AI? No — and be wary of anyone who promises it. Nobody controls ChatGPT, Perplexity, Gemini or Claude: answers vary by question, by day and by engine, and the criteria keep evolving. What you do control is being in the candidate pool and being the easiest source to quote — the two stages described above. Mechanical changes (robots.txt, llms.txt, schema.org markup) take effect as soon as engines re-crawl your site, typically within days to a few weeks. To measure progress, keep it simple: ask the engines the questions your customers actually ask, and see who they cite — then re-ask regularly. That is exactly Citeable's logic: a best-efforts obligation, not a guarantee of results; we make your site as readable and citable as possible, honestly, and the studies above are why it is worth doing. --- ## llms.txt vs schema.org Q&A markup: which one do you need? Source: https://citeable.eu/en/guides/llms-txt-vs-schema-qa ### What's the difference between llms.txt and Q&A schema.org markup? They work at two different levels. llms.txt is a single file, placed at the root of your domain, that gives AI engines a structured summary of your whole site: who you are, what you offer, where your important pages live. It operates at the site level and serves discovery and comprehension. Q&A schema.org markup is JSON-LD embedded in the code of individual pages: it exposes question-and-answer pairs in a format machines read without ambiguity. It operates at the page level and serves citability — it cuts your content into units an answer engine can lift as-is. The shorthand that sums it up: llms.txt makes you readable, Q&A markup makes you quotable word for word. One introduces your site; the other structures your answers. ### What does Q&A schema.org markup actually do? schema.org is a structured-data vocabulary created in 2011 by Google, Microsoft, Yahoo and Yandex to describe web page content in a machine-readable format. Q&A markup is one application of it: a JSON-LD block in the page's HTML that explicitly declares 'here is a question, here is its accepted answer'. The machine no longer has to interpret layout — every answer is labeled, self-contained and attributable. An honesty note: since 2023, Google rarely shows FAQ rich results in classic search pages for ordinary sites. But that is about SERP display widgets, not machine readability. For an AI answer engine parsing your page, the markup remains an unambiguous citation unit — exactly the shape it is looking for: a precise question, a direct answer, a clear source. ### Which one should you add first? If you have five minutes: llms.txt, because it is a single file to drop in and it covers the whole site at once. If you have an hour: both, because they each cover a different half of the path to citation. Answer engines work in two stages — retrieve and understand candidate pages, then select the passages to cite. llms.txt acts on the first stage: it makes your site easy to discover and understand. Q&A markup acts on the second: it makes your answers easy to select and quote as-is. Deploying only one optimizes half the funnel: readable but not quotable with llms.txt alone, structured but poorly summarized with markup alone. So the real answer is: both, generated from the same content so they stay consistent. ### Do the two formats overlap or conflict? Neither — as long as you keep them consistent. They do not say the same thing: llms.txt summarizes and points to your pages, the markup structures the content of those pages. There is no possible conflict between the formats themselves — an engine can read both, and each helps it differently. The real risk is editorial: drift. An llms.txt that promises what your pages don't say, or Q&A markup whose answers contradict the visible text of the page, destroys exactly the trust you are trying to build — answer engines cross-check sources, and inconsistency is paid in non-citation. The practical rule: both files should be generated from the same real content, and regenerated together when the site changes. That is a strong argument for automating the pair rather than hand-maintaining each piece. ### How does Citeable generate the two together? Citeable crawls your site's public pages — no admin access, nothing modified — then extracts the real content: your offer, your answers, your practical information. From that single crawl it generates both files: a structured llms.txt following the specification (title, summary, sections, annotated links) and the Q&A schema.org markup as JSON-LD, ready to drop onto your pages. Because both come out of the same analysis, they are consistent by construction — no drift between the summary and the markup. Payment is one-time, per pack of sites, and the files are regenerable for life as your site evolves: re-run the generation from your account and the pair stays in sync. Honest scope: nobody controls the answer engines, so this is a best-efforts obligation — making your site as readable and citable as possible, cleanly. --- ## Schema markup for Google AI Overviews: what actually helps Source: https://citeable.eu/en/guides/schema-markup-ai-overviews ### Does schema markup get you into AI Overviews? No, not directly — and be wary of anyone who claims it does. Google has never confirmed schema.org markup as a ranking factor for AI Overviews, and its documentation frames structured data as an aid to understanding and display, not as a lever for position. What markup actually does is remove ambiguity. Instead of letting the machine guess, from your layout, where the question is and where the answer is, you declare it explicitly. And answer engines — AI Overviews included — favor content they can extract cleanly and attribute safely. So markup doesn't rank you higher; it makes your content easier to lift once you're already in the pool of candidate sources. It's a citability multiplier, not a ranking button. The confusion comes from there: the two are correlated (well-structured content is often well-cited), but one doesn't mechanically cause the other. ### Which schema types actually matter for AI answers? Five carry most of the value. FAQPage and QAPage expose question-and-answer pairs: the shape closest to what an answer engine wants to cite. Article (or BlogPosting) locates your content, its author and its date — useful for freshness and E-E-A-T. Organization and Person, linked by sameAs, say who you are and tie your site to an identifiable entity. Product carries price, availability and reviews for product pages. Breadcrumb clarifies site structure. The rest (HowTo, Event, LocalBusiness) is useful depending on your business. One essential honesty note: since 2023, Google has sharply reduced FAQ and HowTo rich results in classic search pages for ordinary sites. But that decision is about SERP display widgets, not the machine readability of the markup. For an answer engine parsing your page, a marked-up question-answer pair remains an unambiguous citation unit — exactly what it is looking for. ### Markup helps machines read you, not rank you — why that matters Because it sets realistic expectations and stops you wasting effort. If you think markup ranks, you'll pile schema types onto a hollow page hoping for a miracle — and nothing will happen. If you understand it helps machines read, you use it for what it's worth: making good content extractable. So the correct sequence is: first content that deserves to be cited (a real, direct, honest answer), then markup that exposes it cleanly. Markup on emptiness produces nothing; markup on substance multiplies its citability. It's also why consistency is non-negotiable: markup whose answers contradict the visible text of the page is worse than no markup, because engines cross-check their sources and an inconsistency destroys trust. ### What do AI Overviews actually pull from? AI Overviews rest on Google's index and ranking systems: Google retrieves relevant pages, then a model synthesizes an answer from the passages it judges reliable, and cites its sources. Three traits recur in what gets lifted: self-contained, extractable passages (a direct answer that stands without the rest of the page), corroboration (information confirmed by several reliable sources travels better than an isolated claim), and freshness (recent sources are favored, especially on moving topics). Markup helps with the first trait — it cuts and labels your passages — but replaces neither corroboration (which comes from your off-site authority) nor freshness (which comes from your update cadence). In other words, markup optimizes the shape of your answers; the other two traits depend on what you do around them. ### How does Citeable generate the Q&A markup without breakage? Citeable crawls your site's public pages — no admin access, nothing modified — then generates, from your real content, a Q&A schema.org markup in JSON-LD ready to drop in, plus a consistent llms.txt file alongside it. Because both come out of the same crawl, the marked-up answers reflect the actual text of your pages: no drift between what the machine reads and what the human sees. Three checks remain, always yours: paste the JSON-LD in the right place in the HTML, validate it (via Google's rich results test), and regenerate it when your content changes. Honest scope, repeated everywhere on the site: clean markup makes you citable, it does not guarantee you a slot in an AI Overview — nobody controls the engines. It's a best-efforts obligation: making your content as readable and extractable as possible, cleanly. --- ## GEO agency vs GEO tool: which should you choose? Source: https://citeable.eu/en/guides/geo-agency-vs-tool ### GEO agency or GEO tool: what's the actual difference? The difference comes down to one word: duration. A GEO tool does the technical, on-page work once — it crawls your site and generates the deliverables (llms.txt, Q&A schema.org markup) that you deploy. It's one-off, scoped, cheap, and you stay in control. A GEO agency sells ongoing support: audit, editorial strategy, content production, link-building, monitoring of the answer engines, monthly reporting. It's a recurring retainer that covers not only the on-page half but also the off-page half — authority, freshness, content — that no tool does for you. The shorthand: a tool delivers the setup, an agency carries the effort over time. A tool is a purchase; an agency is a relationship. Choosing really means answering one question: do you need a clean base once and for all, or a team pushing your visibility month after month? ### What does a GEO tool do, and what does it cost? A GEO tool automates the mechanical part that's identical for everyone: making your site readable and quotable by answer engines. Concretely, it crawls your public pages, extracts your offer and your answers, then generates a structured llms.txt file and a consistent Q&A schema.org markup, ready to deploy. On price, it's a one-time payment per site — often a few tens to a few hundred euros depending on the number of sites — with no subscription. It's the right choice when your need is a clean base: you want the right files, well made, without spending a day doing it by hand or getting the format wrong. Its limit is owned openly: the tool doesn't deploy for you, doesn't build backlinks, doesn't write your content and guarantees no citation — it gives you the best possible base, the rest is yours. Citeable is a tool in this category: one crawl, two consistent deliverables, one payment, regenerable for life. ### What does a GEO agency do, and what does it cost? An agency sells human time and expertise over the long run. Beyond the technical setup, it takes on what no tool automates: defining a content strategy, writing articles that answer your market's real questions, building off-site authority (PR, partnerships, backlinks), keeping a freshness cadence, and tracking your visibility in ChatGPT, Perplexity or AI Overviews with regular reporting. On price, it's a recurring monthly retainer, generally from several hundred to several thousand euros per month depending on ambition and market. It's expensive because it's continuous, largely manual work — precisely the off-page half that decides ranking. The risk to know: some agencies mostly bill for the technical setup a tool does for a fraction of the price; an agency's real value is in content, authority and strategy, not in generating the files. ### When does an agency make sense? When GEO is a strategic, ongoing stake for you, and you don't have the hands to carry it in-house. Typically: an established company with a marketing budget, a competitive market where being cited by AI has real commercial value, a need for regular content and sustained link-building, and no one internally to do it. In that case, the retainer mostly buys what you can neither automate nor do yourself for lack of time: editorial production, the relationships that create authority, and expert monitoring of fast-moving engines. If your GEO has to move forward every month and you can fund it, an agency is the right vehicle. Conversely, paying a recurring retainer just to obtain technical files is a bad trade. ### When is a tool enough (and can you combine both)? A tool is enough when your need is to lay a clean base without committing to a monthly budget: a freelancer, a local business, an SMB, a brochure site or a small e-commerce that wants to be readable and quotable without hiring an agency. You generate the files, you deploy them, you write good answers yourself, and you check now and then whether the AIs cite you — the off-page checklist stays within reach when the site is small. And crucially, it's not an exclusive choice: the most effective combination is often to use a tool for the technical setup — done once, done well, cheap — then invest your budget or time where it truly matters, content and authority, either yourself or with an agency if you scale up. Many serious agencies actually use tools for the mechanical part and focus their hours on strategy. Tool and agency don't fight over the same work: one builds the base, the other makes it last. --- ## GEO perfect: the complete checklist (what a tool can't do for you) Source: https://citeable.eu/en/guides/geo-perfect-checklist ### What does being 'GEO perfect' actually mean? GEO perfect doesn't mean 'guaranteed cited' — nobody controls ChatGPT, Perplexity, Gemini or Claude. It means being the easiest source to cite on your topic's questions: present in the candidate pool engines retrieve, then written as self-contained passages they can lift verbatim. The work splits into two halves. The first is on-page and mechanical: a robots.txt that allows AI bots, an llms.txt at your root, Q&A schema.org markup, server-side rendering, direct 40-160 word answers. A tool generates this half in one crawl. The second is off-page and ongoing: deploying the files, building backlinks and brand mentions, publishing fresh content on a cadence, showing real authors and real reviews, and measuring who cites you over time. No tool can do this second half for you — and it's where ranking is won or lost. ### What does a generator like Citeable do — and where does it stop? Citeable takes your URL, crawls your site and generates two consistent deliverables: an llms.txt file that summarizes your site for engines, and Q&A schema.org markup as JSON-LD that makes your answers quotable verbatim. That's the technical part — the part that's the same for everyone and takes hours to do by hand without mistakes. There the honest promise stops. It doesn't deploy the files for you (you paste them in the right place on your hosting), it doesn't create backlinks to you, it doesn't write your next article, it doesn't earn real customer reviews, and it doesn't query the engines every week to see if they cite you. Saying so is the opposite of a weakness: it's the line between what can be automated and what needs your brand, your time and your reputation — none of which can be outsourced. ### Step 1 — Deploy the files where engines look An llms.txt sitting on your disk does nothing. The file must be served at your domain root, at the exact address your-site.com/llms.txt, publicly and as plain text. The Q&A schema.org markup must be injected into the HTML of the relevant pages, inside a script tag of type application/ld+json, and present in the server-rendered HTML — not added afterward by JavaScript, because several AI crawlers don't execute JS. Depending on your platform this ranges from copy-paste (a public folder on Vercel, Netlify or a static site) to a few clicks in a theme (WordPress, Framer, Webflow). Once live, verify two things: open your-site.com/llms.txt in a browser to confirm it loads, and test a page in Google's Rich Results Test to confirm the JSON-LD is detected. Until this step is done, everything else is invisible. ### Step 2 — Build off-site authority (the real ceiling) This is the highest-leverage step and the only one no tool will do for you. AI engines largely inherit the web's trust signals: the more credible sites link to and mention you, the more likely you are to enter the source pool and come out cited. Four concrete levers. One: brand mentions, even without a link — being named on pages engines already read (articles, comparisons, industry forums) creates the 'this entity exists and matters' signal. Two: niche directories and lists — for an AI/GEO topic, llms.txt registries and GitHub awesome-lists are self-serve repositories that accept submissions. Three: editorial backlinks — a guest post, original data others cite, a resource useful enough to be linked on its own. Four: launching — Product Hunt, Indie Hackers, BetaList, a newsletter in your field, which create a first wave of dated mentions. Aim for quality and relevance, never bulk link buying, which is counterproductive and penalized. ### Step 3 — Keep content fresh, on a cadence Freshness is a signal engines — Perplexity above all — weight heavily, and it's one that decays on its own if you don't touch it. An llms.txt and markup generated once freeze your site at a date; six months later, a competitor who republishes regularly looks more alive. Three moves are enough. One: when you genuinely revise a page, update its dateModified field (and regenerate your llms.txt to reflect the new content) — never cheat by changing the date without changing the substance, engines cross-check against the actual content. Two: publish new answers to the real questions your customers ask, one at a time, structured like the others (a real question, a direct self-contained answer). Three: hold a sustainable cadence — one addition a month beats ten at once then nothing. GEO is not a set-and-forget setting; it's a surface you maintain. ### Step 4 — Earn real E-E-A-T signals E-E-A-T — experience, expertise, authoritativeness, trustworthiness — is what separates a source engines judge citable from anonymous text. Three concrete signals, all verifiable and therefore never to be faked. One: the author. Attach a real name to your content, with a short bio establishing why that person is credible on the topic, and link that profile to their public accounts (LinkedIn, X) via schema.org's sameAs field — engines consolidate these identities to assess authority. Two: evidence. Back your claims with facts, numbers and citable sources — Princeton's GEO study shows statistics and citations clearly raise the odds of being lifted. Three: real reviews and feedback. An aggregateRating or Review markup only has value if it rests on real reviews; inventing testimonials or counters is illegal in many countries, detectable, and destroys the very trust you're trying to build. The rule is simple: every E-E-A-T signal must be true, because every E-E-A-T signal is verifiable. ### Step 5 — Measure who cites you (and correct) You can't optimize what you don't measure, and there's no reliable 'Search Console for AI' yet. The manual method is simple and honest: list the ten to twenty questions your customers actually ask, put them periodically to ChatGPT, Perplexity, Gemini and Claude in search mode, and note who gets cited — you, a competitor, no one. Three readings. If you appear nowhere, the problem is upstream: crawlability, indexing, or the candidate pool (redo steps 1 and 2). If competitors appear and you don't on the same question, compare: is their answer more direct, better sourced, fresher? If you appear, note the phrasing that got lifted — that's the model to duplicate across your other answers. Redo the exercise monthly: engines, models and competitors move, and so does your standing. It's this measure-correct loop, not a frozen setting, that gets you close to GEO perfect. ### Can you ever truly be 100% GEO perfect? No, and that's the honest conclusion. 'Perfect' here describes the part you control: making your site as easy to retrieve and cite as possible, then maintaining that lead. What you don't control — how each engine weights its signals, the exact question a user types, the model version of the day — stays out of reach, for you and for any vendor. That's why no serious tool promises a citation: the commitment is a best-efforts obligation, not a guarantee of results. The right way to see Citeable in all this: it does the on-page half for you, fast and error-free — a consistent llms.txt and Q&A schema.org markup in one crawl — so your energy goes where it actually counts, on steps 2 through 5 that only you can run. Do both halves, and you're as close to GEO perfect as today's web allows. --- ## Contact Source: https://citeable.eu/en Email: hello@citeable.eu