GEO articles: AI visibility and citation
GEO, AEO and AI SEO are one body of work measured by three questions. Here is what each means and what to actually change on the page.
How AI finds a website
An AI answer is assembled from two sources. Part of the content is already in the model's training data, and part is fetched from the web for each question by its own crawler. The second path is the one a site owner can affect today.
So GEO work starts with access. If robots.txt blocks GPTBot, ClaudeBot or PerplexityBot, every later decision stops mattering. Once access is there, the text matters most: a model quotes individual passages, so the answer has to sit directly under its own heading.
-
How do I check whether AI sees and cites my website?
A GEO audit checks three things in order. First, whether an AI crawler can reach the site and whether any content exists without JavaScript. Second, whether the page declares facts through structured data, authorship and dates. Third, whether the text is shaped for citation. Each step has its own method.
-
What is AI SEO and how does it differ from GEO and AEO?
AI SEO is the umbrella name for the whole field — adapting a website for AI-driven search. GEO is about whether a model knows and recommends you, and AEO is about whether a direct answer gets lifted from your page. Technically they overlap almost entirely, so separate task lists do not exist.
-
What is Markdown content negotiation and why should a site support it?
Markdown negotiation serves two versions of a page from one URL: HTML to a browser, and plain text without layout to any agent sending an Accept: text/markdown header. The agent receives only what it actually reads and spends several times fewer tokens. No new URLs are created for it.
-
What kind of pages does AI cite most often?
Four kinds of pages get cited most: a question with its answer directly underneath, pricing and terms with concrete numbers, definitions and glossaries, and primary documents. They share one property — you can take two or three lines out of them and those lines stay true without the rest of the page.
-
Is llms.txt actually needed, and what does a good one look like?
llms.txt is still a proposal that no vendor has publicly committed to reading, so its benefit today is small and unproven. But it is cheap. A good file is short, maintained by hand, and lists live URLs with one line of context each — not a sitemap copy, and not a list that went stale after the first restructure.
-
Do websites actually block AI crawlers?
Most of the time AI bots are neither blocked nor admitted deliberately — robots.txt simply does not name them. What happens then depends on the general rules: sometimes everything passes, sometimes everything is blocked along with the rest. Either way it is a decision nobody actually made.
-
How many websites are ready for AI search?
Usually not. Across roughly 500 audits the AI visibility side almost always stands lower than SEO. Four things are missing: a deliberate decision about AI bots, text without JavaScript, questions with answers, and verifiable facts. None of them can be bought — they are assembled inside the site.
-
How do you get into Google AI Overviews and how is it different from ranking?
AI Overviews is a Google-generated summary at the top of the results that cites a handful of sources. What gets cited is not the page ranking first but the page holding a precise, self-contained paragraph answering one specific part of the question. A tenth-place page can be cited while the first place is not.
-
How do you build a page so that AI cites it?
A model does not read a page in order. It splits the text into passages and picks the one that answers the question, which makes headings the boundaries of those passages. That produces four jobs: a question as the heading with the answer directly beneath it, 40 to 120 words per passage, JSON-LD with Organization and Article nodes, and access for answer crawlers in robots.txt. Internal links tie pages on one topic together, so a model finds a topic hub rather than a single answer.
-
Which GEO and AI SEO claims are myths?
The ten most common GEO myths are these: llms.txt is supposedly required, GEO will supposedly replace SEO, a model can supposedly be fooled with keywords, Google search will supposedly drop 25 percent, a page supposedly needs 1000 to 1500 words, content written with AI is supposedly already GEO, blocking every AI crawler is supposedly safest, FAQPage schema will supposedly earn a rich result, the result can supposedly be guaranteed, and ranking is supposedly the same as being cited. None of them survive a check against primary sources.
-
How do you measure whether GEO work is actually working?
GEO results can be measured in four ways: traffic from AI sources tracked separately from other traffic, Search Console impressions and positions, a monthly citation check against the same list of questions, and technical check results before and after the work. Three things cannot be measured: how many people read the answer without clicking, when a model will refresh its data, and why in one specific case it cited one source rather than another.
-
How do you check a GEO expert's claims before paying?
Ask for a concrete deliverable, verifiable example and measurement method. Unsupported citation guarantees or unclear scope warrant further questions. Missing robots.txt, schema or an AI mention alone does not prove that a specialist is fake.
-
Why does GEO need monitoring rather than a single audit?
An audit shows what is on the site today. But a site changes without its owner knowing: after a CMS update, after an editor's fix, after a plugin adds noindex, or when server settings change. None of these announce themselves. So what is needed is not a second identical report but monitoring: the change against the previous scan, an email only when something changed, and a history that shows when it happened.
-
How do you choose a GEO specialist in Lithuania?
Choose a GEO specialist by a clear scope, verifiable examples and a measurement method. Ask them to separate the technical audit, content work and AI answer monitoring. This guide helps compare proposals; it is not an agency ranking or an endorsement of a particular provider.
-
What is llms.txt, and do I actually need it?
llms.txt is a plain text file at the root of a site that tells AI systems what the site is and which pages to read first. It is not a standard and no vendor has publicly committed to reading it — useful, but not the first job on the list.
-
Can ChatGPT see my website, and how do I check?
Check technical access and citations in specific answers separately. Access needs rule and server evidence; a citation needs a prompt, date and real source link. One answer or an audit score cannot establish your overall visibility in ChatGPT.
-
What is AEO, and how do I get into the direct answer?
AEO (Answer Engine Optimization) is shaping a page so a direct answer can be lifted from it — a Google featured snippet or a paragraph inside an AI answer. What decides it is not length but structure: the question as a heading and the answer immediately below.
-
What is GEO optimization, and how is it different from SEO?
GEO (Generative Engine Optimization) is preparing a website so that AI assistants — ChatGPT, Gemini, Perplexity, Claude — understand it, cite it and recommend it. It differs from SEO in its goal: SEO wants the click on your link, GEO wants your answer inside the text the person reads without clicking anything.
-
Should I block AI crawlers?
Choose bot rules by purpose. You can allow AI search while restricting training use. OpenAI's search and training permissions, for example, are independent. Check each provider's documentation and server access before changing rules.
-
How do you write an FAQ so an answer can be taken from it?
Every question must be a real heading (h3), with the answer immediately below it. Headings are passage boundaries: without them a twenty-question block is one large text, and a model cannot cut out the answer to a single question.
-
How does a page get into Google's featured snippet?
Google takes the featured snippet from results that already rank on page one, then picks the one whose passage answers most precisely. So the work is twofold: first reach the top ten, then format the answer so it can be cut out cleanly.
-
What makes an AI assistant cite one source over another?
A model picks a passage, not a page. The winner is the one that answers on its own, is reachable without JavaScript, rests on a verifiable fact and carries a visible date. Position in Google results is neither a condition here nor a guarantee.
-
What do you do when AI answers take traffic from the site?
Short factual questions no longer lead to a website — the assistant answers them itself. What remains are the questions that need a decision, a price or an action. So the work moves from volume to precision: fewer visits, but closer to a purchase.
-
How does a website get into AI assistant answers?
You need three things at once: do not block AI crawlers in robots.txt, serve the content from the server without JavaScript, and structure the text so the answer to one question is one passage. A blocked crawler sees nothing, and empty HTML means an empty page to a model.
Other areas
-
SEO
Indexing, titles, structure, links and what all of it costs. Each article answers one question and rests on a primary source.
-
Speed
Core Web Vitals, the PageSpeed score and what they actually measure. A number without an address changes nothing, so we talk about specific files.
-
Security
Certificates, security headers and what a site says about itself for no reason. We demand only what matters to an ordinary website.
-
Benchmarks
A score means nothing without a comparison. Here is what ordinary looks like, what good looks like, and what separates them.