Słownik SEO i GEO

Refers to the visible area in a browser without scrolling.
Originally, the meaning comes from the print area and describes the print fold in printing.
The opposite is "Under the fold"

AEO stands for Answer Engine Optimization. Answer engines are systems that respond to a question with a direct answer instead of a list of links: featured snippets, voice assistants, Google AI Overviews or AI chats such as ChatGPT and Perplexity.

The focus is on question-and-answer structure. Phrase questions the way users ask them and answer them directly in the first sentence that follows. Details, examples and sources come afterwards. FAQ sections, meaningful subheadings and structured data make it easier for answer engines to find the right passage.

AEO, GEO and LLMO are often used interchangeably. AEO is the older term and also includes non-generative answer formats such as featured snippets.

AI agents are systems that independently carry out multi-step tasks on behalf of users: they research, compare offers, fill in forms or make bookings and purchases. In agentic search, it is no longer the person who searches, but an agent that evaluates the results and acts on them.

For agents to be able to use a website, it must be easy for machines to read and operate. This includes complete product data with price and availability, structured data, clearly labelled forms and buttons, clean HTML and as few obstacles as possible, such as unnecessary pop-ups. Accessibility measures also help, because many agents evaluate pages in a similar way to screen readers.

AI citations are source references with which AI search systems and AI chats back up their answers, usually as a link, footnote or source card. A citation is the most direct form of visibility in generative answers because it can lead users to your page.

A distinction is made between citations (your URL is linked as a source) and mentions (your brand is named in the answer text without a link). Both are recorded separately when measuring AI visibility.

Passages are easy to cite if they answer a question on their own, contain concrete figures, definitions or comparisons, and have a recognisable source, a date and a responsible person. Your own data, studies and surveys have a good chance of being cited because they are not available elsewhere.

AI crawlers are bots that AI providers use to fetch web pages, either to collect training data for their models or to load current content when a user asks a question. Many providers use separate user agents for this, so training and search can be controlled individually.

AI Mode is a separate Google search interface in which queries are answered in a conversation with an AI model. Unlike AI Overviews, which appear above the regular results, AI Mode largely replaces the results list with a detailed answer containing source links. Follow-up questions are possible, and the context of the conversation is kept.

Technically, AI Mode breaks a query down into many sub-queries (see Query fan-out) and combines the results. As a result, pages can appear as a source that do not rank for the original query but answer a sub-question well.

Google launched AI Mode in the US in 2025 and has since expanded it to further countries and languages.

AI Overviews are AI-generated summaries that Google shows above the classic search results. They answer a query in a few paragraphs and link the sources the answer is based on. They appear only when Google assumes that they add value to classic search.

AI referral traffic consists of visits that reach your website via links in AI answers, for example from ChatGPT, Perplexity, Gemini or Copilot. In web analytics tools, they appear as referrals from domains such as chatgpt.com, perplexity.ai, gemini.google.com or copilot.microsoft.com.

Create a separate channel group in your analytics tool so these visits can be evaluated separately. There are two limitations: clicks from Google AI Overviews and AI Mode are counted as regular organic Google traffic. In addition, some visits from AI apps are transmitted without a referrer and appear as direct traffic.

For many websites the volume is still low. However, these visitors have often already narrowed down their choice, because the AI answer has put the topic into context.

AI search engines answer queries with a response written by a language model and name their sources. Well-known examples are ChatGPT Search, Perplexity, Microsoft Copilot, Google Gemini and Google's AI Mode.

The systems differ in where they get their information from. Some operate their own web index, others rely on search indexes such as Bing or Google, or combine both. This is why a page can be cited frequently in one system and not at all in another.

Check regularly how your brand and your topics appear in the systems that matter to your audience (see AI visibility), and make sure their crawlers are allowed to access your pages (see AI crawlers).

AI SEO is an umbrella term with two meanings. On the one hand, it refers to using AI tools for SEO tasks, such as keyword clustering, content briefs or analysing large data sets (see AI tools for SEO). On the other hand, it refers to optimising websites for AI-powered search systems. In this sense, AI SEO is synonymous with GEO, AEO or LLMO.

When reading articles or offers, check which of the two meanings is intended.

In 2025, there are numerous AI tools that support SEOs – from content optimization to data analysis. Use these tools to recognize patterns (e.g., keyword groupings via cluster analysis) or automatically create meta descriptions. But don't blindly rely on AI: always check the results manually. Used correctly, AI tools can accelerate routine tasks and uncover new optimization potentials.

AI visibility describes how often and in what form a brand, product or website appears in the answers of AI systems. It is the counterpart to classic visibility in search results, but it is not measured through positions but through mentions and source citations in answers.

Use AI tools (e.g., ChatGPT) for content creation with caution. You should always check and adapt AI texts for accuracy, uniqueness, and added value. Google treats AI content no differently than human-written content – what matters is that it is helpful and high-quality.

Search engines no longer use AI only for ranking but also for presenting results. Google shows AI Overviews above the search results and offers a conversational search experience with AI Mode. Bing integrates Microsoft Copilot, and AI chats such as ChatGPT or Perplexity search the web themselves.

Instead of a list of links, users receive a summarised answer with source references. For your page to be considered as a source, your content should answer questions precisely, be clearly structured and contain verifiable information. Optimising for this is called GEO (Generative Engine Optimization).

A backlink is a link from another website to yours. For search engines, links have always been a signal for which pages exist and how they relate to each other. Google says that PageRank, the link-based method from its early days, continues to be part of its core ranking systems, even though it has changed considerably since launch.

In 2023 Microsoft integrated AI answers based on OpenAI's GPT models into Bing search, now under the name Microsoft Copilot. Copilot summarises search results and links to the sources it used.

The Bing index matters beyond Bing itself, because other AI services use Bing search data. Register your website in Bing Webmaster Tools, submit a sitemap and use IndexNow so that new and updated pages are picked up quickly.

Black Hat SEO refers to prohibited methods like cloaking, hidden texts, or massive link buying to manipulate rankings. Such techniques may bring short-term success but often lead to penalties by Google. Stick to Google guidelines and rely on sustainable optimization – short-term tricks are not worth it and endanger your visibility long-term.

Brand mentions are references to a brand, product or person on other websites, in forums, review platforms, media or social networks, with or without a link.

For AI systems they are an important signal. A language model learns from its training data which brands are named in which context. If your brand is frequently mentioned together with a topic, the probability increases that a model names it for relevant questions. In live search, too, AI systems often draw on comparison articles, best-of lists, forums such as Reddit and review platforms.

Unlike classic link building, unlinked mentions also count here. What matters most is the topical context, the credibility of the source and a consistent spelling of the brand name.

A strong brand also pays off in SEO. When users specifically search for your brand, Google gets the signal that your website is trustworthy and well-known. Therefore, build your brand awareness – for example through social media activities, PR, and recommendations. Additionally, users in search results are more likely to click on familiar names, which increases your click-through rate and can thus indirectly positively influence your ranking.

Regularly check your website for broken links (404 errors). Broken internal links make crawlers' work difficult and frustrate users. Repair or remove broken links promptly, and set 301 redirects if necessary when pages have been moved. Error-free linking signals quality and care.

Chunking is the process of splitting documents into smaller text passages (chunks) that AI systems store, evaluate and retrieve individually. AI search systems usually do not compare entire pages with a query but individual passages. The passage that best matches the question is the one that gets cited.

This leads to a few principles for structuring text:

• Each section covers one subtopic under a meaningful subheading.
• The key statement comes at the beginning of the section.
• Sections can be understood without the rest of the page. Avoid references such as "as described above" and name the subject instead of writing only "it" or "this product".
• Tables, lists and definitions can easily be taken over as a unit.

Google has followed a similar approach in classic search since 2021 with passage ranking.

Freshness can be a ranking factor for certain topics. Regularly review older content and keep it up to date. Updated publication dates, fresh examples, or new insights show Google that your site is maintained and remains relevant for users.

Remove or revise outdated and weak content (content pruning). Pages that bring no traffic or are not convincing in quality can affect the overall performance of your website. By cleaning out old content, you improve the average quality and focus of your site, which can positively affect ranking.

The quality of your content is crucial. Write comprehensive, well-structured texts that provide value to the reader. Avoid phrases and filler words – every sentence should be useful. The higher the content quality, the better user satisfaction and ranking will be.

Structure your content with headings (H1-H6), paragraphs, and lists. A clear structure helps both your readers and search engines to quickly grasp the content. Use meaningful headings that summarize the topic – so crawlers and users immediately understand what the following section is about.

In conversational search, users phrase their queries as complete questions or tasks and refine them in a dialogue. The system keeps the context of the conversation and relates later questions to earlier ones.

As a result, queries become longer and more specific. Instead of "running shoes test", a query might read: "Which running shoes are suitable for a beginner weighing 90 kg on asphalt?". AI chats, Google's AI Mode and voice assistants encourage this type of search.

Content should therefore take specific situations, target groups and conditions into account and also answer obvious follow-up questions (see Long-tail keywords and Voice search).

The Core Web Vitals are three Google metrics for the user experience of a page: Largest Contentful Paint (LCP, load time until the main content), Interaction to Next Paint (INP, response time to input) and Cumulative Layout Shift (CLS, visual stability). Google's ranking systems take them into account, but the relevance of the content counts for more.

The crawl budget is the number of pages Google can and wants to fetch on your website. It results from two quantities: how much your server can handle (crawl capacity) and how much Google wants to fetch your pages (crawl demand). For most websites it is no issue, but for very large websites it helps decide how fast new and changed pages reach the index.

Duplicate content means that the same or very similar content can be reached at several addresses, on the same website or on different ones. Most of the time it arises unintentionally from the technology. Google does not punish it across the board, but it picks one version for the search results, and that is not necessarily the one you want.

E-E-A-T stands for Experience, Expertise, Authoritativeness and Trustworthiness. It is the yardstick by which Google's human quality raters judge whether a page and its content are reliable. E-E-A-T is not a ranking factor with a value of its own.

Embeddings are numerical representations of text, images or other content as vectors, i.e. long sequences of numbers. Content with a similar meaning lies close together in this vector space, even if it uses different words.

AI search systems convert both the query and text passages into embeddings and look for the passages with the greatest similarity (vector search). This way, a system can match the query "cheap car insurance" with a text about "affordable motor cover".

As a result, repeating a keyword exactly loses importance. What matters more is that a passage covers a topic clearly and completely (see Semantic search).

Google analyzes texts for entities (people, places, things). Use clear names and terms so the search engine can properly classify your content. When Google recognizes and links the entities on your page (Knowledge Graph), it increases your thematic relevance.

Focus on evergreen content: content that remains relevant long-term. Such timeless guides or tutorials bring continuous traffic without you having to constantly revise them. Make sure that evergreen topics are still occasionally checked and slightly updated so they appear current and remain competitive.

Featured snippets are highlighted results above the organic hits that provide a direct answer (often in the form of text excerpts, lists, or tables). To achieve a featured snippet, answer common questions about your topic concisely in your content. Use clear question-answer formats or lists. If Google rates your answer as particularly helpful, it may place it as a featured snippet.

GEO (Generative Engine Optimization) covers all measures that help content appear in the answers of generative AI systems such as ChatGPT, Perplexity, Gemini, Microsoft Copilot or Google AI Overviews. The goal is for an AI system to link your website as a source, adopt its statements or name your brand.

Google adjusts its algorithm several times a year with Core Updates. These core updates often focus on content quality and relevance. If you are affected by an update, check your site for possible weaknesses in content, user experience, or E-E-A-T and optimize accordingly. A sustainable content strategy protects you best from negative update effects.

Google Discover shows users personalized content even without active search. To appear here, create content with high utility value and appealing images above all. Freshness and E-E-A-T also play a role. Use interesting headlines (no clickbait) and meta descriptions so your post makes people curious in Discover. Note: Google Discover brings unpredictable traffic, but high-quality, visually appealing content increases the chance of inclusion.

Google MUM (Multitask Unified Model) understands complex queries across texts, images, and languages. Therefore, provide diverse content (text, images, videos) and cover a topic comprehensively. The more holistically you treat a topic, the better MUM can assign your site to matching search queries.

Google Search Console is a free tool from Google that shows how your website performs in Google Search. It provides clicks, impressions and queries from search, reports indexing and crawling problems, and lets you submit sitemaps and inspect individual pages. The data comes directly from Google, which is why it counts as the most important data source for SEO.

Grounding refers to connecting an AI answer to verifiable sources. Instead of answering only from its training knowledge, the model relies on current documents, such as search results, and refers to them.

Grounding is meant to reduce hallucinations and keep answers up to date. The sources a system selects for this determine which websites are linked in the answer. Content with clear, verifiable statements, a date, an author and references is well suited as a basis.

The H1 tag basically shows the search engine what topic the page is about.
It is a myth to believe that the website may have only exactly one H1 tag.
Google itself has explained that multiple H1 tags are fine:

https://www.youtube.com/watch?v=UILcrHaO7m4&feature=youtu.be&t=8m59s

<!-- Correct heading structure for SEO -->
<h1>Main topic of the page</h1>
  <h2>Subtopic 1</h2>
    <h3>Detail on subtopic 1</h3>
  <h2>Subtopic 2</h2>
    <h3>Detail on subtopic 2</h3>

A hallucination is a statement by an AI model that sounds plausible but is factually wrong or made up, for example incorrect figures, non-existent sources or invented product features.

This matters for businesses when AI systems output wrong prices, opening hours, services or ratings. The risk decreases if clear, consistent and easy-to-find information about your business is available: on your website, in structured data, in your Google Business Profile and in industry directories.

Check regularly what AI systems say about your brand and correct contradictions at the source.

Hreflang is a declaration with which you tell search engines which pages contain the same offering for other languages or countries. Google can then show users the fitting language version and does not have to treat the versions as duplicates.

Images can bring considerable traffic via Google image search. Therefore, optimize every image: use a descriptive file name and fill the alt attribute with a meaningful, concise text that describes the image content. Compress images to an appropriate file size so loading time remains low. With structured data (ImageObject) and an image sitemap, you can provide Google with additional info about your images.

IndexNow is a protocol with which you can immediately inform search engines (including Bing and Yandex; Google does not participate) about new or updated content. If you support it, your website proactively sends URLs to participating search engines. This way, new content outside of Google – especially at Bing – is indexed significantly faster.

Information gain describes how much new information a piece of content provides compared with other content on the same topic. Google holds a patent on this; whether and how the concept is used in ranking has not been officially confirmed.

For AI answers the idea is relevant nonetheless. If ten pages contain the same statement, an AI system needs only one of them as a source. Pages with their own data, practical examples, tests, first-hand experience or a well-founded assessment offer something that is missing elsewhere.

Before writing, check what the existing top results already cover and add what is missing there.

Link your pages internally in a meaningful way to signal to Google the importance of individual subpages. Use descriptive anchor texts for internal links that reflect the target topic. Strong internal linking helps distribute link juice within your website and improves the findability of new or deep content.

If your website is heavily based on JavaScript, pay attention to SEO aspects. Google can render JS, but there may be delays in indexing. Make sure important content is not only loaded later via JS or provide a fallback (e.g., SSR or dynamic rendering). Check with tools like the URL Inspection tool in Search Console whether Google sees your JS content.

The keyword that Google should respond to.

Multiple pages of your website should not compete for the same keyword. Check if you are practicing cannibalization – i.e., if different subpages target the same search query. It is better to merge such content or clearly differentiate it so Google clearly recognizes which page should rank for which keyword.

The Knowledge Graph is a knowledge base in which search engines store entities such as people, companies, places and products, as well as their relationships to one another. Google uses it for knowledge panels, among other things; other providers operate comparable databases.

For AI systems, a knowledge graph serves as a factual basis. It helps resolve ambiguities and verify statements. If your company is recorded there as a distinct entity, AI answers are more likely to give correct information about its name, location or services.

You can support this with structured data of the type Organization or Person, links to official profiles via the sameAs property, a well-maintained Google Business Profile and consistent information in directories and databases such as Wikidata, provided their notability criteria are met.

In link building (offpage SEO), you should proceed sustainably. Create content that others like to link to (linkbait), or do targeted outreach to build backlinks. Avoid unnatural link patterns and purchased links – Google increasingly recognizes manipulative linkbuilding methods and punishes them. Instead, rely on PR, guest posts, or cooperations to get real recommendations.

A large language model (LLM) is an AI model trained on very large amounts of text, from which it has learned how words and sentences follow one another. On this basis it can understand, summarise and rephrase text. Examples are GPT (OpenAI), Claude (Anthropic), Gemini (Google) and Llama (Meta).

LLMs are the foundation of AI chats and AI search features. Their built-in knowledge ends at the cut-off date of their training data (see Training data). They obtain current information through a web search while answering a request (see RAG).

An LLM captures relationships of meaning rather than individual keywords. Clear language, unambiguous terms and logically structured paragraphs make it easier to reproduce your content correctly.

LLMO stands for Large Language Model Optimization. The term describes optimising content and brands for large language models (LLMs), such as the models behind ChatGPT, Claude, Gemini or Llama.

LLMO concerns two levels. The first is the knowledge a model brings from its training data. What matters here is how often and in what context your brand is mentioned on the web. The second is content that a model retrieves live when answering a request (see RAG). Here, findability, crawlability and easily quotable passages count.

In practice, LLMO largely overlaps with GEO and AEO.

The llms.txt is a proposed plain-text file in Markdown format that sits in the root directory of a website (/llms.txt). It is meant to give language models a compact overview of the most important content: a short description and a list of key pages with links.

For local businesses, "Local SEO" is crucial. Maintain your Google Business Profile (formerly Google My Business) with current data, reviews, and photos. Optimize your website for local keywords (including place names) and present name, address, phone number (NAP) consistently. Local directory entries and reviews can improve your local ranking so you become visible in Google map search (Local Pack).

Long-tail keywords are longer, more specific search queries with lower search volume but often high purchase or action intent. By covering such specific keywords, you can more easily achieve top rankings and attract very targeted visitors. Create content that specifically addresses these niche queries.

If your website violates Google's guidelines, a manual action (penalty) can be imposed. Then your ranking is drastically worsened or pages are completely removed from the index. To avoid this, refrain from unfair methods and monitor Search Console for warnings. If you receive a manual action, fix the problems (e.g., remove spam links or thin content) and submit a reconsideration request to have the penalty lifted.

Meta tags are specifications in the source code of the HTML document. They contain, among other things, information for search engines.

Attention: Google does not read the meta tag "keywords". So you can save time and completely omit this tag!

<!-- Meta description: optimal length 150-160 characters -->
<meta name="description" content="This is an optimal meta description with 150-160 characters. It describes the page content precisely and contains the main keyword.">

Mobile-first indexing means that Google uses the mobile version of your content for indexing and ranking, fetched with the smartphone crawler. What is missing from the mobile page cannot count for search, even if it is visible on desktop.

There are many myths surrounding SEO and Google.

For example, Google does not crawl the meta tag keyword and multiple H1 tags on a page are quite ok.

See Google video on H1 tag:

https://www.youtube.com/watch?v=UILcrHaO7m4&feature=youtu.be&t=8m59s

With the noindex directive you tell search engines that a page should not be included in the search index. You set it as a meta tag in the head of the page or as an HTTP header. Unlike the robots.txt, noindex does not control crawling but whether a fetched page appears in the results.

Onpage describes all measures taken to optimize the website with regard to ranking and/or a keyword.

Page Experience includes factors like loading speed, mobile usability, security, and page stability. Google evaluates user experience: consistently good page experience across the site can improve your ranking. Therefore, pay attention to fast servers, mobile optimization, HTTPS, and avoid intrusive interstitials.

The Google Pagespeed is an indicator of how fast your website loads.
Google uses the value to consider how valuable your site is in terms of speed.

You can check your site's PageSpeed here:

https://developers.google.com/speed/pagespeed/insights/?hl=en

It's not just about the size of your website and the speed of your internet connection.

Much more important is: Which files block rendering in the browser. How fast does the website need until it displays content. etc.

For paginated pages (e.g., category pages 1, 2, 3), there is no special markup for Google since the removal of rel="next/prev". Make sure users and crawlers can easily reach the relevant page. Good internal linking (e.g., "Show all products") and clear page numbering help Google fully capture your content despite pagination.

Google personalizes search results according to the user's location, search history, and interests. While you have little direct influence on this, ensure overall strong signals (local relevance, brand mentions, social media presence). This way the chances are good that your site is also visible in personalized results.

Prompt tracking is the regular, automated querying of defined prompts in AI systems in order to record changes in their answers. It is the equivalent of rank tracking in classic search engine optimisation.

Unlike keywords, prompts have no reliable search volume data, because the providers do not publish such figures. Build your prompt list from real questions, for example from customer enquiries, support tickets, Google Search Console, "People also ask" boxes and forums. Add variants with different wording and for different stages of the buying decision: informational questions, comparisons and concrete requests for recommendations.

AI answers are not deterministic; the same question does not always produce the same answer. Evaluate trends and averages across several queries.

Query fan-out is a technique in which an AI search system splits a user query into several sub-queries, runs them in parallel and merges the results into one answer. According to Google, the method is used in AI Mode and in AI Overviews.

Example: the question "Which e-bike battery is suitable for long tours?" can be broken down into sub-queries on capacity, range, weight, charging time and price.

Ranking for the main keyword is therefore not enough. Content that covers a topic completely, including its typical sub-questions, has more opportunities to be selected as a source for individual sub-queries (see Topic authority).

Retrieval-Augmented Generation (RAG) is a method in which a language model retrieves relevant documents before answering and bases its answer on that content. The documents come, for example, from a search index, a database or a live web search.

The process consists of three steps: the request is translated into search queries, the system retrieves matching text passages (retrieval), and the model writes an answer with source references (generation).

Most AI search engines work on this principle. A page can only be cited if it is found in the retrieval step. Clean indexing, short loading times, server-rendered text and self-contained sections increase the chance that your content is retrieved and used.

Rich snippets are enhanced search result displays with additional information (star ratings, images, price information, etc.). They often arise from structured data on your site. With correctly implemented schema markup, you increase the likelihood that Google displays such rich snippets for your pages – which can lead to a higher click-through rate.

The robots.txt is a text file in the root directory of a website (/robots.txt) with which you tell crawlers which areas they may fetch and which not. It controls crawling, not indexing, and usually also contains the reference to the sitemap.

Make sure your content meets the user's search intent. For every keyword, consider what the searcher really wants to know or do. If you provide the right answer or solution, the visitor stays longer on your page and Google rates your site as more relevant.

Google today better understands the context and meaning of words. Make sure to treat a topic holistically instead of just repeating a keyword. Use synonyms and topic-related terms so search engines can capture the semantic connection and recognize your content as comprehensive.

I assume that Google will increasingly use Natural Language Processing (NLP) to interpret texts and capture their content. This leads to the following recommendations for good SEO text:

1.) When creating texts, write grammatically clean, clear, and above all simple. Don't use nested sentences. They belong in a good dissertation, but not on a webpage. Also avoid many subordinate clauses. What is good for readers cannot be bad for search engines.

2.) Avoid personal pronouns. Name the entity (unit, brand) by name so Google can clearly understand what is meant.

3.) Only use adverbs and adjectives if they are really important for understanding a sentence.

In summary, completely avoid waffle and blah blah to do Google and thus also readers a favor.

Summary: Become a brand and an authority!!

SERP is the English abbreviation for Search Engine Result Pages. SERPs are nothing other than the organic search results that are shown when you have entered a search term on Google. So they are not the ads delivered by Google Ads.

Share of model is a metric that indicates the share of mentions a brand receives in AI answers within an industry or topic area. It transfers the principle of share of voice to generative AI systems.

Example: if 100 prompts about "web hosting for small businesses" produce 400 provider mentions in total, and your brand is named 60 times, your share of model is 15%.

The metric shows how a model positions your brand relative to the competition. It depends heavily on the selection of prompts and the AI systems examined. Only compare values based on the same prompt list.

A sitemap is a file that tells a search engine which pages of a website it should know about. Usually it is an XML file called sitemap.xml in the root directory. It does not replace linking and guarantees neither crawling nor indexing, but it helps to find pages and to report changes.

Optimize your snippets (page title and meta description) for a higher click-through rate. Write concise, meaningful titles with the main keyword preferably at the front. The meta description should make people curious and briefly summarize the topic – it may directly address the user. An attractive snippet sets you apart from other hits and can bring more clicks to your site despite the same ranking.

Structured data is additional information in the source code of a page that tells search engines, in a defined vocabulary (usually Schema.org), what the page is about: an article, a product, an organization, an event. It makes content machine-readable and can lead to enhanced displays in the search results, called rich results.

Show Google that you comprehensively cover a topic. Create thematically related posts and link them internally to form a content hub. When your website offers many high-quality contents on one area, you are perceived as an authority – this can improve your ranking for all related keywords.

Training data is the text, images and other content an AI model was trained on. For large language models, a significant share comes from publicly accessible web pages, books, scientific articles and program code.

Knowledge from training ends at a certain date, the knowledge cutoff. The model only knows about events, products or prices after that date if it retrieves them via a web search. Changes to your website therefore only affect training knowledge with a new model generation, but affect AI search much faster.

With many providers, you can control via robots.txt whether your content may be used for training (see AI crawlers). An opt-out does not prevent your brand from entering training data through mentions on other websites.

Refers to the not immediately visible area.
In SEO, this is the area where speed can be optimized when mobile pages only load images when they come into the visible area. This ensures that the "Above the fold" area (i.e., the immediately visible area) loads faster. Google will give you plus points for this and your page will rank better.

Even though Google officially doesn't confirm user signals like click-through rate or dwell time as direct ranking factors, they can play an indirect role. A low bounce rate and longer visit duration signal that your content is relevant and engaging. Therefore, optimize the usability and content of your site so visitors like to stay and interact. Satisfied users lead to better rankings in the long term.

Videos are prominently displayed in search results and on YouTube. Use video SEO: create an optimized title, a meaningful description, and tags for your videos. Embed your videos meaningfully on your site and use video schema.org markup. Transcripts or subtitles help search engines understand video content. Through video SEO, you can gain visibility both on YouTube and in Google search results.

More and more users use visual search (e.g., Google Lens) to find information about images. Make sure your images are SEO-optimized: use descriptive file names, alt texts, and surrounding text that explains the image content. This way search engines can capture the context. With visual search, users can find your site if your image matches the searched object.

In voice search, users often ask longer, naturally formulated questions. Therefore, optimize your content also for complete questions and colloquial formulations. A good strategy is to create FAQ sections that briefly answer typical W-questions. This increases the chance that voice assistants (e.g., Siri, Google Assistant) will read your site as an answer.

Keyword density was yesterday. Anyone who wants to rank well on Google today needs somewhat better content than writing the word: MyKeyword 10 times in a row in the content.

The solution, thanks to the legendary blog article by Karl Kratz, is: wdf*idf.
Until there is something better, the wdf*idf formula is considered the jack-of-all-trades.

So the ratio of the occurrence of a term in relation to its weighting in the text. Basically a mathematical formula.

Of course, wdf*idf optimized content is no guarantee for a TOP ranking on Google. As with everything in SEO, the overall mix is what matters here too.

A clear website structure helps search engines with indexing. Organize your content hierarchically (homepage → categories → subpages) and create clear navigation. A good structure ensures that crawlers find all important pages and users can easily navigate.

For more and more search queries, users get the answer directly on the Google results page (zero-click). This includes featured snippets, knowledge panels, weather, currency converters, etc. For websites, this means: still provide valuable information, but don't always expect a click. Optimize your content so it provides value both on your site and directly in search results (e.g., through concise summaries or structured data).