What you should know about GEO and AEO
What is GEO optimization and how is it different from SEO
GEO (Generative Engine Optimization) prepares your site to be used by a neural network as a source of an answer. SEO fights for a position in a list of links, GEO fights for the model to cite you inside generated text.
The practical difference: in search a user sees ten links and chooses, while a neural network answer cites only two or three sources — making the cut decides everything. Classic ranking factors do not disappear: the page must still be accessible, indexable and fast. New ones are simply added.
- Machine-readable structure: JSON-LD the model can extract facts from without guessing.
- A direct answer in the first sentences — models cite the start of a block, not conclusions at the end.
- Explicit trust signals: author, update date, links to primary sources.
- Access for AI crawlers that neither robots.txt nor a WAF blocks.
What is AEO and why separate it from GEO
AEO (Answer Engine Optimization) handles how easily a page text can be turned into an answer: question headings, short definitions, lists and tables. GEO handles how technically understandable the site is to a machine; AEO handles the form of the content itself.
The split matters because different people fix them. A developer closes GEO issues in one release: add markup, fix canonical, open bots. AEO issues are closed by an editor, and that is weeks of work. A single score mixing both does not tell you whom to go to.
What is llms.txt and is it mandatory
llms.txt is a file in the site root with a short project description and a list of key materials in Markdown. No search engine is required to read it, and it does not affect ranking. It is a cheap safety net, not a requirement.
Treat it as a README for machines: it guarantees nothing, but costs ten minutes of work and makes the site structure explicit. RANKA checks for a heading, a description and at least two links to same-domain materials — a one-line file is not useful.
How do I check whether ChatGPT can see my site
Looking into robots.txt is not enough. You must request the page with a specific bot User-Agent and compare the response to a normal browser one: usually crawlers are blocked not by robots.txt but by Cloudflare, a WAF, or a 403 rule on User-Agent.
RANKA does exactly this: it sends requests as GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, YandexBot and others, and compares response codes and content volume to a baseline. Cases where a bot gets a browser-check page or a trimmed version are flagged separately.
The exception is Google-Extended and Applebot-Extended: they have no dedicated User-Agent, only robots.txt tokens — the crawl is done by Googlebot and Applebot. They can only be checked via robots.txt, and the report says so honestly.
Which Schema.org markup do you need to appear in AI answers
At minimum Organization and WebSite on the homepage, BreadcrumbList on inner pages, plus a type matching the page: Article for materials, Service for services, Product for goods, FAQPage for question-and-answer sections.
More important than coverage is how filled the fields are. An Organization without url and sameAs links the brand to nothing; an Article without author and datePublished gives the model no way to judge trust and freshness. RANKA shows not only which type is missing but which specific fields are empty on each page.
About FAQPage specifically: Google has barely shown rich results for it since 2023, so as a snippet tool it is exhausted. For language-model parsing it is still useful — it ties a question to an answer explicitly.
How does the RANKA check work
You paste up to 10 URLs of one site, the service loads each page, parses the HTML and returns 20 checks with scores, priority and fix effort. No generative models are used, no tokens are spent.
- Only the URLs you entered are checked — there is no automatic crawl of the site.
- Each check returns points, a priority (from critical to low) and an effort estimate.
- Non-applicable checks are marked as such and do not hurt the score: if your selection has no articles, you do not lose points for a missing publication date.
- The final score is the average of GEO and AEO weighted 50/50, not a sum of all points.
- The report is saved under a permanent link you can hand to a developer.
Why does my React site show zero words and zero markup
Because the audit reads HTML without executing JavaScript — exactly like most AI crawlers do. If content appears only after scripts run, neural networks will see an empty page.
This is not a tool limitation but a reproduction of real conditions. GPTBot, ClaudeBot and PerplexityBot generally do not execute JS. The fix is server-side rendering or pre-rendering key pages. RANKA detects such sites and warns with a separate block above the report.
How much does the check cost
Nothing. RANKA is free, works without registration and does not collect contact data. The limits — 10 pages per check and a per-address request limit — keep the service available for everyone.