<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:wfw="http://wellformedweb.org/CommentAPI/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	xmlns:slash="http://purl.org/rss/1.0/modules/slash/"
	>

<channel>
	<title>MCA RGB</title>
	<atom:link href="https://www.mca.si/en/feed/" rel="self" type="application/rss+xml" />
	<link>https://www.mca.si/en/</link>
	<description>Websites, online stores, digital marketing and AI automation for businesses</description>
	<lastBuildDate>Mon, 16 Nov 2026 19:56:12 +0000</lastBuildDate>
	<language>en-US</language>
	<sy:updatePeriod>
	hourly	</sy:updatePeriod>
	<sy:updateFrequency>
	1	</sy:updateFrequency>
	

<image>
	<url>https://www.mca.si/wp-content/uploads/2024/06/cropped-cropped-cropped-logo_size-32x32.jpg</url>
	<title>MCA RGB</title>
	<link>https://www.mca.si/en/</link>
	<width>32</width>
	<height>32</height>
</image> 
	<item>
		<title>Chatbot or AI agent? What the difference is and how to build your own AI agent</title>
		<link>https://www.mca.si/en/2026/11/02/chatbot-vs-ai-agent/</link>
		
		<dc:creator><![CDATA[Robert Rakovič]]></dc:creator>
		<pubDate>Mon, 02 Nov 2026 13:50:46 +0000</pubDate>
				<category><![CDATA[AI]]></category>
		<category><![CDATA[Blog]]></category>
		<guid isPermaLink="false">https://www.mca.si/?p=1179</guid>

					<description><![CDATA[<p>An AI agent is a language model that can use tools and works in a loop until the task is done. We explain the difference between a chatbot and an agent, and five ways to build an agent yourself.</p>
<p>The post <a href="https://www.mca.si/en/2026/11/02/chatbot-vs-ai-agent/">Chatbot or AI agent? What the difference is and how to build your own AI agent</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></description>
										<content:encoded><![CDATA[<p><strong>An AI agent is a language model that can use tools and works in a loop until the task is done.</strong> That is the whole secret; everything else is detail.</p>
<p>Everyone talks about AI agents today, but few can explain what an agent really is. In this article we first explain the difference between a chatbot and an agent, and the middle stage that the big chat assistants offer today. Then we look at the building blocks of an agent, and finally at five concrete ways to build an agent yourself.</p>
<h2>Chatbot or AI agent: what is the difference?</h2>
<p>A chatbot answers a question, an agent completes a task. Two additional abilities make the difference: tool calling and working in a loop.</p>
<p>An ordinary chatbot, or a “bare” language model (LLM), can do only one thing: return text in response to your message. It cannot check e-mail, look something up on the web, book an appointment or update a record in a database. It can tell you <em>how</em> you would do it, but it does not do it itself.</p>
<p>An AI agent is the same LLM with two upgrades:</p>
<ol>
<li><strong>Tool calling</strong> – the agent can take actions in the real world: search the web, read files, call APIs, write to a database.</li>
<li><strong>Loop</strong> – the agent does not stop after one answer. It continues on its own, step by step, until the goal is reached.</li>
</ol>
<figure class='wp-block-table'>
<table>
<thead>
<tr>
<th></th>
<th>Chatbot</th>
<th>AI agent</th>
</tr>
</thead>
<tbody>
<tr>
<td>Input</td>
<td>A question</td>
<td>A goal or task</td>
</tr>
<tr>
<td>Output</td>
<td>Text</td>
<td>Completed work + report</td>
</tr>
<tr>
<td>Access to systems</td>
<td>No</td>
<td>Yes, through tools</td>
</tr>
<tr>
<td>Number of steps</td>
<td>One answer</td>
<td>As many as the task needs</td>
</tr>
<tr>
<td>Who performs the action</td>
<td>The user</td>
<td>The agent</td>
</tr>
</tbody>
</table>
</figure>
<p><strong>An example from an online store.</strong> A shopper asks: “Where is my order #4821?”</p>
<ul>
<li><strong>A chatbot</strong> answers in general terms: “You can check your order status in your account or call support.”</li>
<li><strong>An AI agent</strong> calls a tool for the WooCommerce API and reads the order, then calls the delivery service API to track the parcel. From both results it composes the answer: “The order was shipped yesterday, the parcel is at the distribution center and will be delivered tomorrow.”</li>
</ul>
<p>The model can be exactly the same in both cases. The difference is in what is built around it. What a chatbot brings to a website and what the new EU rules require is described in the article <a href='https://www.mca.si/en/2026/10/31/ai-assistant-website-eu-rules/'>AI assistant on your website</a>.</p>
<h2>The middle stage: from chatbot to agent</h2>
<p>Between an ordinary chatbot and an agent you program yourself there is now a wide middle stage. The big chat assistants (ChatGPT, Claude, Gemini, Grok, Kimi, Perplexity, Copilot) let you assemble an agent right in their interface, without a line of code.</p>
<p>The path from chatbot to agent has four stages:</p>
<figure class='wp-block-table'>
<table>
<thead>
<tr>
<th>Stage</th>
<th>What it can do</th>
<th>Example from an online store</th>
</tr>
</thead>
<tbody>
<tr>
<td>Ordinary chatbot</td>
<td>Answers a question from general knowledge.</td>
<td>“Write a description for hiking boots.”</td>
</tr>
<tr>
<td>Customized assistant</td>
<td>Has saved instructions, reference files and an agreed way of answering.</td>
<td>An assistant that writes descriptions in the tone of your brand and from your price list.</td>
</tr>
<tr>
<td>Agent with tools</td>
<td>Chooses the steps itself, uses connected apps and carries the task through to the agreed result.</td>
<td>Reads product data in the store, prepares a description and saves it as a draft.</td>
</tr>
<tr>
<td>Automated agent</td>
<td>Repeats the procedure on a schedule or on a supported event (e-mail, message, webhook).</td>
<td>Every Monday it prepares a sales report and sends it to the team.</td>
</tr>
</tbody>
</table>
</figure>
<p>A customized assistant (for example the former Custom GPT or a Gemini Gem) is not yet a real agent. It follows instructions very well, but it does not act independently. It becomes an agent when it gets tools and a loop. It becomes an automated agent when a schedule or an event triggers it instead of you.</p>
<p>For most companies this middle stage is the fastest route to a first agent. How to approach it on individual platforms is described below, under the first level of building.</p>
<h2>The five building blocks of every agent</h2>
<p>Every agent, regardless of the tool it is built in, stands on the same five building blocks.</p>
<figure class='wp-block-table'>
<table>
<thead>
<tr>
<th>Building block</th>
<th>Analogy</th>
<th>What it does</th>
</tr>
</thead>
<tbody>
<tr>
<td>Language model (LLM)</td>
<td>Brain</td>
<td>Predicts which text comes next. It has no hands and no memory of its own.</td>
</tr>
<tr>
<td>Tool calling</td>
<td>Hands</td>
<td>The model outputs a structured command (e.g. “call web search with question X”), the code around it executes the command and returns the result to the model.</td>
</tr>
<tr>
<td>Context window</td>
<td>Working memory</td>
<td>Everything the model currently “sees”: instructions, conversation history, tool calls and their results. Its size is limited.</td>
</tr>
<tr>
<td>System prompt and message roles</td>
<td>Job description</td>
<td>The system prompt sets the identity and the rules; messages are labeled <em>system</em>, <em>user</em> or <em>assistant</em>.</td>
</tr>
<tr>
<td>Vector database (RAG)</td>
<td>Long-term memory</td>
<td>Stores data as vectors (embeddings). With a tool, the agent finds and loads only the relevant parts.</td>
</tr>
</tbody>
</table>
</figure>
<p>A few key points:</p>
<ul>
<li><strong>The model itself executes nothing.</strong> It only decides which tool it would call. The program code around the model (the so-called <em>harness</em>) actually executes the command and returns the result. This division of labor is worth understanding, because it is the core of every agent.</li>
<li><strong>The model has no memory between calls.</strong> At every step the whole relevant history has to be sent into the context window again. If you do not do that, the agent “forgets” what it was doing a minute ago.</li>
<li><strong>Context engineering is a new skill.</strong> Because the window is limited, the quality of an agent depends largely on what you put into it and what you leave out. Too much noise worsens decisions, too little data leads to guessing.</li>
<li><strong>Example of a system prompt:</strong> “You are a customer support assistant for store X. Answer in English. Never guess order data; always check it with a tool.”</li>
<li><strong>RAG in practice:</strong> a catalog of 5,000 products does not fit into the context window. Store it in a vector database, and when asked “do you have waterproof hiking boots under €100?” the agent loads only a few of the most relevant products.</li>
</ul>
<h2>The agent loop: think, act, observe, repeat</h2>
<p>The agent loop connects all five building blocks into a whole that reaches the goal on its own.</p>
<figure class='wp-block-image size-full'><img src='https://www.mca.si/wp-content/uploads/2026/11/Agentna-zanka-AI-agenta.png' alt='The agent loop: goal and instructions, the model thinks, needs a tool, code runs the tool, result goes into context, final answer' class='wp-image-1152'/></figure>
<p>The loop starts with the system prompt and the goal. The model thinks and decides: does it need a tool, or can it already answer? If it needs a tool, the code runs it and the result is added to the context window. The model “observes” the result and decides on the next step. The cycle repeats until the task is done.</p>
<p>In practice always add a safeguard, e.g. a maximum of 10 steps, so that an error does not leave the agent spinning forever and burning tokens.</p>
<h2>Four decisions before you build</h2>
<p>Whether you “click together” an agent or program it, you make the same four decisions.</p>
<ol>
<li><strong>Platform or framework</strong> – where the agent will run (chat assistant, no-code tool, n8n, agent harness or your own code).</li>
<li><strong>Model</strong> – e.g. GPT, Claude or Gemini. A more capable model follows instructions and picks tools better, but costs more per call.</li>
<li><strong>Context</strong> – which data the agent can access: product database, orders, documentation, past conversations.</li>
<li><strong>Tools and flow</strong> – which tools the agent may call, in what order, and where it must stop or ask a human.</li>
</ol>
<p>A practical tip: start with the smallest set of tools. An agent with two well-described tools is more reliable than an agent with twenty that overlap.</p>
<h2>How to build an AI agent: five levels</h2>
<p>You can build an agent in five ways, from settings in the chat assistant you already use to your own code. Each next level gives you more control and requires more technical knowledge.</p>
<h3>Level 1: An agent in the chat assistant you already use</h3>
<p>The easiest route to an agent is often in the tool you already pay for. You describe the task to the agent in plain language, connect your apps and save a successful procedure for reuse.</p>
<ul>
<li><strong>How:</strong> you write standing instructions, add reference files, connect apps and save the procedure as a “Skill”. Then, if needed, you give it a schedule or a trigger.</li>
<li><strong>Advantage:</strong> no new platform and no installation; the agent works where you already have your documents and connections.</li>
<li><strong>Drawback:</strong> you are tied to the provider, its plans and usage limits. Feature availability differs by subscription and region.</li>
</ul>
<p>What the individual platforms offer (as of November 2026):</p>
<figure class='wp-block-table'>
<table>
<thead>
<tr>
<th>Platform</th>
<th>How you assemble an agent</th>
<th>Execution and limits</th>
</tr>
</thead>
<tbody>
<tr>
<td>Grok Bot (xAI)</td>
<td>You create a named bot, define the task and sources in conversation and connect apps. You save procedures as Skills.</td>
<td>Its own computer in the cloud (browser, files, terminal), memory, collaboration between several bots and Routines. Usage is limited by plan.</td>
</tr>
<tr>
<td>ChatGPT (Work and plugins)</td>
<td>Plugin Creator combines instructions, reference files and connected apps. Work carries out a multi-step task.</td>
<td>Workspace Agents for companies add sharing, schedules and API triggers. For your own development: Agents SDK.</td>
</tr>
<tr>
<td>Claude</td>
<td>You add instructions, Skills, connectors and plugins. You save a successful procedure as a Skill.</td>
<td>Supports recurring, scheduled tasks. For your own development: Claude Agent SDK.</td>
</tr>
<tr>
<td>Kimi (Agent, Swarm, Claw)</td>
<td>You create a Skill from instructions or documents. Swarm splits a larger task among subagents by itself.</td>
<td>Claw is a permanent assistant in the cloud with memory, tools and schedules.</td>
</tr>
<tr>
<td>Gemini (Gems and Opal)</td>
<td>A classic Gem follows instructions and knowledge. Opal builds a multi-step mini app from a description.</td>
<td>Opal is an experiment (personal accounts, web, English). For developers: ADK.</td>
</tr>
<tr>
<td>Microsoft Copilot Studio</td>
<td>In a guided editor you describe the agent, add knowledge, tools and triggers, and test it.</td>
<td>Business processes in the Microsoft environment, publishing to supported channels, reacting to events.</td>
</tr>
<tr>
<td>Perplexity Computer</td>
<td>You describe the goal, connect apps and add Skills. The system can bring in subagents.</td>
<td>Research, work in the browser, running in the background and recurring monitoring.</td>
</tr>
</tbody>
</table>
</figure>
<p><strong>A note on ChatGPT:</strong> new Custom GPTs on personal accounts are no longer the recommended way to start. The standard retirement is planned for 11 December 2026; exceptions and other deadlines apply to some workspaces. Check the notice in your account.</p>
<p><strong>From first task to automation in five steps.</strong> The procedure is similar on all platforms:</p>
<ol>
<li><strong>Create an agent with one clear task.</strong> Define the name, area of work, sources, format of the result and standing rules.</li>
<li><strong>Connect the sources.</strong> The agent needs actual access to the apps and data you mention in the instructions.</li>
<li><strong>Test on a real, limited task.</strong> Review the result and correct the instructions. Save a successful procedure as a Skill and check it on other input data as well.</li>
<li><strong>Add a schedule or trigger.</strong> Define when the agent starts, what it gets as input, what the final result is and what it should do if a source is missing.</li>
<li><strong>Set up oversight.</strong> Define the actions for which the agent must always ask for approval, for example before publishing or sending a message. With Grok Bot, keep in mind that bots of the same user share a computer, files and logged-in sessions, so a separate role does not mean separate access.</li>
</ol>
<p>An example of a starting instruction for an online store agent:</p>
<blockquote>
<p>You are a content agent for an online store. For 20 products from the attached table, prepare descriptions in English (150–200 words) with a title, key features and a meta description. Use only data from the table and the technical sheets. If a piece of data is missing, mark it “CHECK”. Return the result in a table for review.</p>
</blockquote>
<p><strong>How to compare platforms:</strong> pick one process and the same small sample, for example 10 products for which you know the correct result. Compare the platforms by accuracy, verifiability of sources, number of manual corrections, actual usage and success in accessing tools.</p>
<h3>Level 2: No-code platforms</h3>
<p>You set up the agent by filling in forms or dragging blocks. Examples of platforms: GenSpark, Lindy, Gumloop, Stack AI, Relevance AI, Botpress, Voiceflow and OpenAI Agent Builder.</p>
<ul>
<li><strong>How:</strong> you write a prompt that describes the role of the agent, which tools it should use and in what form it should return the result.</li>
<li><strong>Advantage:</strong> a working agent in a few minutes.</li>
<li><strong>Drawback:</strong> little control. The result is often similar to a general assistant with tools plugged in.</li>
</ul>
<p>This is what the GenSpark start screen looks like. You describe the task in a single field, and the platform offers ready-made agents for presentations, spreadsheets, documents, design, code, dashboards, and video, images and audio. Under <em>All Agents</em> you find the full set.</p>
<figure class='wp-block-image size-full'><img src='https://www.mca.si/wp-content/uploads/2026/11/Genspark-AI-Workspace.png' alt='Genspark AI Workspace: one input field and ready-made agents for office, media and development tasks' class='wp-image-1153'/></figure>
<h3>Level 3: Low-code</h3>
<p>Visual editors where you connect building blocks and add a block of code when needed. Examples: n8n, Flowise, LangFlow, Dify, Activepieces, KNIME, Make and Microsoft Copilot Studio.</p>
<ul>
<li><strong>How in n8n:</strong> you add an <em>AI Agent</em> node, set its system message, then plug in tools (e.g. Perplexity for web search, Wikipedia for facts) and conversation memory.</li>
<li><strong>Advantage:</strong> much more control over the properties of the agent and a more consistent loop. Easy to connect with existing systems (WooCommerce, e-mail, CRM, Google Sheets).</li>
<li><strong>Drawback:</strong> requires basic technical understanding; complex flows become hard to follow.</li>
</ul>
<p>This is what an agent in n8n looks like. A message from Telegram triggers the <em>AI Agent</em> node, which has three building blocks plugged in: a language model (OpenAI Chat Model), conversation memory (Window Buffer Memory) and an image generation tool (DALL-E 3). The answer goes back to Telegram, and on an error the flow automatically sends a corrected message.</p>
<figure class='wp-block-image size-full'><img src='https://www.mca.si/wp-content/uploads/2026/11/AI-agent-v-n8n-s-Telegramom.png' alt='AI agent in n8n: Telegram trigger, AI Agent with a model, memory and the DALL-E 3 tool' class='wp-image-1154'/></figure>
<p>In n8n you can also include an agent in a wider automated flow. The example below starts by itself every morning at 8:00: it collects news from two sources, merges them, prepares a summary with GPT-5 and sends it by e-mail.</p>
<figure class='wp-block-image size-full'><img src='https://www.mca.si/wp-content/uploads/2026/11/n8n-tok-za-jutranji-povzetek-novic.png' alt='n8n flow: schedule at 8:00, fetching news from two sources, merging, summary with GPT-5 and sending by e-mail' class='wp-image-1155'/></figure>
<h3>Level 4: Agent harness</h3>
<p>You install a ready-made, capable agent and adapt it. The leading ones are OpenClaw and Hermes Agent; the latter is focused on long-term memory.</p>
<ul>
<li><strong>How:</strong> the loop, memory and basic functions are already built. You add your own instructions (e.g. a research assistant “skill”) and plug in the tools you need.</li>
<li><strong>Advantage:</strong> a production-ready agent without writing the loop from scratch.</li>
<li><strong>Drawback:</strong> it requires installation, often on your own server (VPS), and maintenance.</li>
</ul>
<h3>Level 5: Full code</h3>
<p>You write the whole agent yourself, usually in Python, optionally with a framework such as LangGraph. You have full control: the system prompt, tool definitions (JSON schemas), response handling and the loop itself.</p>
<p>Below is a minimal but working agent with Python and the Anthropic SDK. The model decides by itself whether to call search, the code runs the search and returns the result to the model, and the loop runs until the model gives a final answer.</p>
<pre class='wp-block-code'><code>import anthropic

client = anthropic.Anthropic()  # key in the ANTHROPIC_API_KEY variable

# 1) Tool definition: name, description and JSON schema of the input
tools = [{
    'name': 'web_search',
    'description': 'Searches the web and returns short summaries of the results.',
    'input_schema': {
        'type': 'object',
        'properties': {'query': {'type': 'string', 'description': 'Search query'}},
        'required': ['query'],
    },
}]

def web_search(query: str) -&gt; str:
    # Call a real search API here (e.g. Perplexity, Tavily, SerpAPI)
    return f'Results for: {query} ...'

# 2) System prompt and goal
SYSTEM = 'You are a research assistant. Always cite your sources.'
messages = [{'role': 'user', 'content': 'What is new with AI agents this month?'}]

# 3) Agent loop: think -&gt; act -&gt; observe -&gt; repeat
for step in range(10):  # safeguard against an endless loop
    response = client.messages.create(
        model='claude-sonnet-5-5',
        max_tokens=2048,
        system=SYSTEM,
        tools=tools,
        messages=messages,
    )
    messages.append({'role': 'assistant', 'content': response.content})

    # The model wants no more tools -&gt; final answer
    if response.stop_reason != 'tool_use':
        print(''.join(b.text for b in response.content if b.type == 'text'))
        break

    # Run all requested tools and return the results to the model
    tool_results = []
    for block in response.content:
        if block.type == 'tool_use':
            result = web_search(**block.input)
            tool_results.append({
                'type': 'tool_result',
                'tool_use_id': block.id,
                'content': result,
            })
    messages.append({'role': 'user', 'content': tool_results})</code></pre>
<p>All the building blocks are visible in this code: the model, the tool definition, the system prompt, the context window (the <code>messages</code> list, which grows with every step) and the loop. Everything that platforms on the lower levels hide is in plain sight here.</p>
<h2>Which approach to choose?</h2>
<p>Choose the lowest level that still solves your problem. The same model and the same loop run on all five levels; the only difference is how much you build yourself and how much the platform gives you.</p>
<figure class='wp-block-table'>
<table>
<thead>
<tr>
<th>If you need …</th>
<th>Choose</th>
<th>Example tools</th>
</tr>
</thead>
<tbody>
<tr>
<td>A quick solution in a tool you already use, without a new platform</td>
<td>An agent in a chat assistant</td>
<td>ChatGPT Work, Claude, Grok Bot, Gemini Gems</td>
</tr>
<tr>
<td>A quickly working solution without programming</td>
<td>A no-code platform</td>
<td>GenSpark, OpenAI Agent Builder</td>
</tr>
<tr>
<td>Real logic, branching and connections to several services, without maintaining code</td>
<td>Low-code</td>
<td>n8n, Make</td>
</tr>
<tr>
<td>Business agents in the Microsoft environment</td>
<td>Low-code</td>
<td>Copilot Studio</td>
</tr>
<tr>
<td>A capable, production-ready agent that you adapt</td>
<td>An agent harness</td>
<td>OpenClaw, Hermes Agent</td>
</tr>
<tr>
<td>Full control, deep integration into your own product, or understanding of every part</td>
<td>Full code</td>
<td>Python, LangGraph</td>
</tr>
</tbody>
</table>
</figure>
<p><strong>Examples for an online store:</strong></p>
<ul>
<li>Product descriptions and replies to shoppers based on confirmed product data → <strong>an agent in a chat assistant</strong> with a saved Skill and a connection to the catalog.</li>
<li>Automatic answers to questions about order status → <strong>n8n</strong> with a WooCommerce node and an AI Agent node.</li>
<li>An agent that reviews stock at night, compares competitor prices and prepares a report → <strong>an agent harness</strong> on your own VPS.</li>
<li>An agent that synchronizes orders with the ERP system and works out the cause itself when errors occur → <strong>full code</strong>, because you need control over every step and reliable logging.</li>
</ul>
<h2>Frequently asked questions</h2>
<p><strong>Is ChatGPT an AI agent?</strong><br />By itself it is a chatbot. When you give it tools (web search, connected apps) and let it carry out multi-step tasks independently, it works as an agent.</p>
<p><strong>Do I need a programmer for an AI agent?</strong><br />Not necessarily. On the first two levels you assemble an agent without code. For a connection to an ERP system, your own database or more demanding automations, programming is usually needed.</p>
<p><strong>Is an AI agent safe for company data?</strong><br />It depends on the access you give it. Give the agent only the access it really needs, and always require your approval for sensitive actions (sending, publishing, payments). If it processes personal data, take GDPR into account as well.</p>
<p><strong>What is the best first agent for a company?</strong><br />A task that repeats every week and has a clear result: a weekly report, preparing product descriptions or sorting inquiries.</p>
<h2>Conclusion</h2>
<p>An AI agent is not magic. It is a language model with hands (tools), working memory (the context window) and persistence (the loop). Once you understand that, you quickly tell real agents from chatbots with a new label. It also becomes easier to decide whether your problem needs a few sentences of instructions in a chat assistant, a form in a no-code tool or a hundred lines of Python.</p>
<h2>Need an AI agent for your company?</h2>
<p>At MCA RGB we design and build AI agents that connect with your online store, CRM and ERP system. Read more about what we can automate on the <a href='https://www.mca.si/en/ai-automation-for-business/'>AI automation for business</a> page, and for conversations with customers on your website see our <a href='https://www.mca.si/en/ai-chatbot/'>AI chatbot</a>.</p>
<p><a href='https://www.mca.si/en/contact/'>Contact us</a> and together we will find the first task an agent can take over.</p>
<p>The post <a href="https://www.mca.si/en/2026/11/02/chatbot-vs-ai-agent/">Chatbot or AI agent? What the difference is and how to build your own AI agent</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>AI assistant on your website: what it brings to your business and what the new EU rules require</title>
		<link>https://www.mca.si/en/2026/10/31/ai-assistant-website-eu-rules/</link>
		
		<dc:creator><![CDATA[Robert Rakovič]]></dc:creator>
		<pubDate>Sat, 31 Oct 2026 15:27:56 +0000</pubDate>
				<category><![CDATA[General]]></category>
		<guid isPermaLink="false">https://www.mca.si/?p=1180</guid>

					<description><![CDATA[<p>What an AI assistant on a website does for a business, where it pays off most, how to introduce it in five steps and what the EU AI Act requires from 2 August 2026.</p>
<p>The post <a href="https://www.mca.si/en/2026/10/31/ai-assistant-website-eu-rules/">AI assistant on your website: what it brings to your business and what the new EU rules require</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></description>
										<content:encoded><![CDATA[<p>Customers expect an answer right away, even at ten in the evening or on a Sunday. An AI assistant on your website answers 24 hours a day, and your team gets time for more demanding work. According to Eurostat, in 2025 roughly one in five EU companies with at least 10 employees used artificial intelligence; in Slovenia the share was around 22%.</p>
<p>Since 2 August 2026 a new obligation also applies in the EU: a visitor must know that they are talking to artificial intelligence. In this article we explain what an assistant can do, where it pays off most, how to introduce it and what the new rules require.</p>
<h2>What an AI assistant does for a business in practice</h2>
<p>A modern AI assistant is no longer about clicking through predefined menus. It understands a question in natural language and answers on the basis of your content: price lists, terms of business, frequently asked questions and service descriptions.</p>
<ul>
<li><strong>Answers frequent questions</strong> about delivery, prices, opening hours and returns.</li>
<li><strong>Helps find the right product or service</strong> and takes the visitor to the right page.</li>
<li><strong>Collects inquiries and contacts</strong> and enters them into the CRM.</li>
<li><strong>Books appointments</strong> and sends a confirmation.</li>
<li><strong>Hands the conversation over to a colleague</strong> when the question is demanding or the customer asks for it.</li>
</ul>
<h2>Where an AI assistant pays off most</h2>
<figure class='wp-block-table'>
<table>
<thead>
<tr>
<th>Type of business</th>
<th>Most frequent questions</th>
<th>What the assistant takes over</th>
</tr>
</thead>
<tbody>
<tr>
<td>Online store</td>
<td>Delivery, stock, returns</td>
<td>Answers and product recommendations</td>
</tr>
<tr>
<td>Service company</td>
<td>Price, appointment, procedure</td>
<td>Collecting inquiries and bookings</td>
</tr>
<tr>
<td>B2B and manufacturing</td>
<td>Specifications, delivery times</td>
<td>Technical answers from documentation, contact into the CRM</td>
</tr>
<tr>
<td>Tourism and hospitality</td>
<td>Availability, price list</td>
<td>Answers in several languages</td>
</tr>
</tbody>
</table>
</figure>
<p>A simple rule: if your team answers the same questions every day, an assistant pays off.</p>
<h2>What the EU AI Act requires from 2 August 2026</h2>
<p>From 2 August 2026, Article 50 of the AI Act on transparency applies. Systems that interact directly with people must be designed so that people are informed that they are interacting with artificial intelligence. The only exception is when this is obvious.</p>
<ul>
<li><strong>Who is responsible:</strong> formally the obligation lies with the provider of the system. Because the assistant on your site works under your name, the safest approach is to always show the notice clearly.</li>
<li><strong>Example of a notice:</strong> “I am the AI assistant of company X. Answers may be incomplete; for an exact quote I will connect you with a colleague.”</li>
<li><strong>AI-generated content:</strong> if you use AI to create realistic images or videos (deepfakes) or texts on matters of public interest, additional labeling obligations apply.</li>
<li><strong>Personal data:</strong> the assistant processes visitor data, so it must be mentioned in your privacy policy, in line with GDPR.</li>
</ul>
<p>This is not legal advice. For specific cases, consult a lawyer.</p>
<h2>How to introduce an AI assistant in five steps</h2>
<ol>
<li><strong>Collect the 20–30 most frequent questions</strong> from e-mail, phone calls and messages.</li>
<li><strong>Prepare the knowledge sources:</strong> price list, terms, frequently asked questions and service descriptions. An assistant is only as good as your content.</li>
<li><strong>Set the limits:</strong> what the assistant does not answer, for example custom pricing, and when it hands the conversation over to a person.</li>
<li><strong>Connect it with your systems:</strong> CRM, e-mail, calendar and product catalog.</li>
<li><strong>Test and monitor:</strong> read the first conversations, fill in the missing answers and measure the number of inquiries.</li>
</ol>
<h2>The most common mistakes</h2>
<ul>
<li>The assistant makes up answers because it is not limited to your sources.</li>
<li>There is no handover to a person, so a demanding customer leaves.</li>
<li>It is not clear that it is AI, which since August 2026 is also a breach of the rules.</li>
<li>Collected contacts stay in the chatbot and never reach the CRM.</li>
<li>Nobody reads the conversations, although they are the best source of data on what customers ask.</li>
</ul>
<h2>The assistant as a source of content ideas</h2>
<p>Conversations with the assistant show what customers actually ask. That is an excellent basis for frequently asked questions and new articles. Google also likes to use such content in AI answers, as we explain in the article <a href='https://www.mca.si/en/2026/06/22/ai-seo-google-ai-overviews-chatgpt/'>How to optimize a website for Google AI Overviews</a>.</p>
<h2>Frequently asked questions</h2>
<p><strong>Does every chatbot have to show that it is AI?</strong><br />Yes, unless it is obvious. The safest approach is to always show the notice, at the start of the conversation.</p>
<p><strong>Does an AI assistant replace employees?</strong><br />No. It takes over repetitive questions and hands the more demanding ones to a colleague, together with the full context of the conversation.</p>
<p><strong>Does the assistant speak Slovenian?</strong><br />Yes. Modern language models answer well in Slovenian and in other languages, which is useful for foreign visitors.</p>
<p><strong>What do I need to get started?</strong><br />A list of frequent questions and the existing content on your website. How long the introduction takes depends mainly on how many systems (CRM, calendar, catalog) we connect it with.</p>
<h2>Need an AI assistant for your website?</h2>
<p>At MCA RGB we set up an <a href='https://www.mca.si/en/ai-chatbot/'>AI chatbot</a> that answers on the basis of your content, has a clear AI notice, hands the conversation over to a colleague and enters inquiries into your <a href='https://www.mca.si/en/crm-implementation/'>CRM</a>. Read more about what we can automate on the <a href='https://www.mca.si/en/ai-automation-for-business/'>AI automation for business</a> page.</p>
<p><a href='https://www.mca.si/en/contact/'>Contact us</a> and we will prepare a proposal for your website.</p>
<p>The post <a href="https://www.mca.si/en/2026/10/31/ai-assistant-website-eu-rules/">AI assistant on your website: what it brings to your business and what the new EU rules require</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>5 automated e-mails every online store needs</title>
		<link>https://www.mca.si/en/2026/10/31/automated-emails-online-store/</link>
		
		<dc:creator><![CDATA[Robert Rakovič]]></dc:creator>
		<pubDate>Sat, 31 Oct 2026 15:27:26 +0000</pubDate>
				<category><![CDATA[General]]></category>
		<guid isPermaLink="false">https://www.mca.si/?p=1181</guid>

					<description><![CDATA[<p>Abandoned cart, welcome, post-purchase, review and repeat purchase, and win-back: five automated e-mails you set up once and that work every day.</p>
<p>The post <a href="https://www.mca.si/en/2026/10/31/automated-emails-online-store/">5 automated e-mails every online store needs</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></description>
										<content:encoded><![CDATA[<p>According to the Baymard Institute, on average around 70% of shopping carts are left unfinished. Most of these shoppers do not come back on their own. Automated e-mails remind them at the right time, without manual work: you set them up once and they work every day.</p>
<figure class='wp-block-table'>
<table>
<thead>
<tr>
<th>E-mail</th>
<th>When it is sent (starting setting)</th>
<th>Goal</th>
</tr>
</thead>
<tbody>
<tr>
<td>1. Abandoned cart</td>
<td>1 hour after leaving, reminder after 24 hours</td>
<td>Completed purchase</td>
</tr>
<tr>
<td>2. Welcome</td>
<td>Right after newsletter sign-up</td>
<td>First purchase</td>
</tr>
<tr>
<td>3. Post-purchase</td>
<td>On dispatch and a few days after delivery</td>
<td>Trust, fewer returns</td>
</tr>
<tr>
<td>4. Review and repeat purchase</td>
<td>7–14 days after delivery or when the product is expected to run out</td>
<td>Reviews and a second purchase</td>
</tr>
<tr>
<td>5. Win-back of inactive customers</td>
<td>60–90 days without a purchase</td>
<td>Reactivation</td>
</tr>
</tbody>
</table>
</figure>
<h2>1. The abandoned cart e-mail</h2>
<p>This is the e-mail with the biggest effect, so set it up first. It contains a product image, a button to return to the cart and short answers to common doubts: how much delivery costs, when the parcel arrives and how returns work.</p>
<p>A tip: do not offer a discount in the very first e-mail. Otherwise shoppers quickly learn that waiting pays off.</p>
<h2>2. A welcome for new subscribers</h2>
<p>A new subscriber is most interested at the moment they sign up. In the first e-mail say who you are, what you sell and why to buy from you, for example free delivery above a certain amount. Optionally add a coupon for the first purchase.</p>
<h2>3. The post-purchase e-mail</h2>
<p>Tell the customer when the parcel will arrive and send a tracking link. A few days after delivery, add instructions or tips for using the product. Such an e-mail builds trust and reduces the number of returns, so do not sell in it yet.</p>
<h2>4. A review request and a repeat purchase reminder</h2>
<p>Once the customer is already using the product, ask for a review. For consumable products such as food supplements, cosmetics or pet food, a reminder when the supply is expected to run out works even better, for example before the end of a 30-day pack.</p>
<h2>5. Winning back inactive customers</h2>
<p>A customer who has not bought for several months needs a reason to return: something new, a special offer, or simply the question of whether they still want to receive your newsletter. Remove anyone who still does not respond from the list. That improves the deliverability of all your other e-mails.</p>
<h2>Consent and GDPR</h2>
<ul>
<li><strong>Marketing messages</strong> (welcome, offers, win-back) need consent, for example a checkbox at registration or at checkout.</li>
<li><strong>The exception for existing customers:</strong> you may offer them similar products if you gave them the option to refuse at the time of purchase and offer it in every e-mail.</li>
<li><strong>Transactional messages</strong>, such as order confirmation and parcel tracking, do not need consent.</li>
<li><strong>Abandoned cart:</strong> the safest approach is to send it only to shoppers who have consented to the newsletter.</li>
</ul>
<p>This is not legal advice. For specific cases, consult a lawyer.</p>
<h2>How to start</h2>
<ol>
<li><strong>Choose a tool that connects with your store</strong>, for example Omnisend, Klaviyo or Mailchimp for WooCommerce.</li>
<li><strong>Set up the abandoned cart and the welcome e-mail first.</strong> They bring the most with the least work.</li>
<li><strong>Sort out the technical settings of your domain</strong> (SPF, DKIM and DMARC). Without them e-mails often end up in spam.</li>
<li><strong>Measure the results:</strong> opens, clicks and above all the revenue each e-mail generates.</li>
<li><strong>Connect the e-mails with your CRM</strong>, so you see the whole history of a customer in one place.</li>
</ol>
<h2>The most common mistakes</h2>
<ul>
<li>A discount already in the first abandoned cart e-mail.</li>
<li>E-mails that are too long and hard to read on a mobile phone.</li>
<li>Sending without valid consent.</li>
<li>Sending from a free address, for example Gmail, instead of from the store domain.</li>
<li>The automation is set up once and nobody ever checks it again.</li>
</ul>
<h2>Frequently asked questions</h2>
<p><strong>How many e-mails should an abandoned cart sequence have?</strong><br />As a rule two to three: the first soon after leaving, the second the next day, the third if needed with a small incentive.</p>
<p><strong>Do I need a lot of subscribers for this to pay off?</strong><br />No. Automated e-mails work from the first customer onward, because they are sent on a customer action, not on a campaign.</p>
<p><strong>Which e-mail do I set up first?</strong><br />The abandoned cart, because it addresses shoppers who were already closest to buying.</p>
<h2>Shall we set up your automated e-mails?</h2>
<p>At MCA RGB we set up <a href='https://www.mca.si/en/email-marketing-services/'>e-mail marketing</a> for your online store: from choosing the tool and the technical domain settings to designing the e-mails and connecting them with your <a href='https://www.mca.si/en/crm-implementation/'>CRM</a>. Read more about what we can automate on the <a href='https://www.mca.si/en/marketing-automation/'>Marketing automation</a> page.</p>
<p><a href='https://www.mca.si/en/contact/'>Contact us</a> and we will prepare a proposal for your store.</p>
<p>The post <a href="https://www.mca.si/en/2026/10/31/automated-emails-online-store/">5 automated e-mails every online store needs</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>The Accessibility Act for online stores: what it requires and how to make your store compliant</title>
		<link>https://www.mca.si/en/2026/10/31/accessibility-act-online-stores/</link>
		
		<dc:creator><![CDATA[Robert Rakovič]]></dc:creator>
		<pubDate>Sat, 31 Oct 2026 15:26:58 +0000</pubDate>
				<category><![CDATA[General]]></category>
		<guid isPermaLink="false">https://www.mca.si/?p=1182</guid>

					<description><![CDATA[<p>Since 28 June 2025 online stores must be accessible to people with disabilities. Who the Slovenian act (ZDPSI) applies to, what it requires and a 15-point checklist for your store.</p>
<p>The post <a href="https://www.mca.si/en/2026/10/31/accessibility-act-online-stores/">The Accessibility Act for online stores: what it requires and how to make your store compliant</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></description>
										<content:encoded><![CDATA[<p>Since 28 June 2025 the Act on Accessibility of Products and Services for Persons with Disabilities (ZDPSI) has applied in Slovenia. It transposes the European Accessibility Act into Slovenian law, and e-commerce is among the services it covers.</p>
<p>An accessible store is not useful only for people with disabilities. It is also clearer and easier to use for older shoppers, for purchases on a mobile phone and for everyone in a hurry.</p>
<h2>Who the act applies to</h2>
<ul>
<li><strong>E-commerce services:</strong> websites and apps on which a consumer selects, orders and pays for a product or service.</li>
<li><strong>The exception for microenterprises:</strong> companies that provide services and have fewer than 10 employees and an annual turnover of up to 2 million euros.</li>
<li><strong>Supervision:</strong> for e-commerce the competent body is the Market Inspectorate of the Republic of Slovenia (TIRS). The act provides for fines for providers who do not bring their service into compliance.</li>
<li><strong>Disproportionate burden:</strong> a company can rely on it only if it justifies and documents it. It is not an automatic exception.</li>
</ul>
<p>This is not legal advice. If you are not sure whether the act applies to you, consult a lawyer.</p>
<h2>What an “accessible” online store means</h2>
<p>In practice the benchmark is the European standard EN 301 549, which refers to the WCAG 2.1 guidelines at level AA. In short: the store must be perceivable, operable, understandable and technically robust, also for a shopper who uses a screen reader or only a keyboard.</p>
<h2>Checklist: 15 points for your online store</h2>
<p><strong>Images and content</strong></p>
<ol>
<li>All product images have descriptive ALT text.</li>
<li>The contrast between text and background is at least 4.5:1.</li>
<li>The page works at 200% zoom without loss of content.</li>
<li>Videos have subtitles.</li>
</ol>
<p><strong>Navigation</strong></p>
<ol start='5'>
<li>The whole purchase can be completed with the keyboard alone.</li>
<li>Focus is visible, so the shopper always knows where on the page they are.</li>
<li>Headings (H1, H2, H3) follow each other in a logical order.</li>
<li>Links and buttons have understandable names, not “Click here”.</li>
</ol>
<p><strong>Forms and checkout</strong></p>
<ol start='9'>
<li>Every field has a label, not just a hint inside the field.</li>
<li>Errors are described in text, not only colored red.</li>
<li>CAPTCHA has an accessible alternative.</li>
<li>The cookie banner can be reached with the keyboard and a screen reader.</li>
</ol>
<p><strong>Information</strong></p>
<ol start='13'>
<li>The page language is set correctly (for a Slovenian store, to Slovenian).</li>
<li>Information on the accessibility of the service is published (an accessibility statement).</li>
<li>Contacts for help are listed, for example e-mail and phone.</li>
</ol>
<h2>How to start in four steps</h2>
<ol>
<li><strong>An automated check</strong> with free tools such as WAVE and Google Lighthouse. They find only part of the problems, so they are just a start.</li>
<li><strong>A manual keyboard test</strong> along the whole purchase path: home page, category, product, cart and checkout.</li>
<li><strong>Fixes in the theme and plugins.</strong> Themes and page builders are often the biggest source of errors.</li>
<li><strong>An accessibility statement and regular checks</strong> with every new feature or redesign.</li>
</ol>
<p>Plugins that promise accessibility “in one click” do not replace fixes in code and content.</p>
<h2>Accessibility also helps sales and visibility in Google</h2>
<p>ALT descriptions, a correct heading structure and understandable forms help Google understand the store better. The same goes for AI agents, which more and more often read websites instead of users, as we describe in the article on <a href='https://www.mca.si/en/2026/06/22/ai-seo-google-ai-overviews-chatgpt/'>Google AI Overviews</a>. An easier purchase also means fewer abandoned carts.</p>
<h2>Frequently asked questions</h2>
<p><strong>Does the act apply to my small store too?</strong><br />If you have fewer than 10 employees and up to 2 million euros in annual turnover, you are exempt as a service provider. Accessibility still pays off because of your shoppers and Google.</p>
<p><strong>My store was set up before June 2025. Do I have to change anything?</strong><br />The act applies to services you provide after 28 June 2025, regardless of when the store was set up. Exceptions apply to some older content, for example archived content.</p>
<p><strong>Who checks compliance?</strong><br />For e-commerce services, the Market Inspectorate of the Republic of Slovenia.</p>
<p><strong>Is an accessibility plugin enough?</strong><br />No. A plugin can help with individual settings, but it does not fix errors in the theme, forms and checkout.</p>
<h2>Planning a new or redesigned online store?</h2>
<p>At MCA RGB we <a href='https://www.mca.si/en/online-store-development/'>build online stores</a> and <a href='https://www.mca.si/en/seo-services/'>optimize them for search engines</a>. For a new store or a redesign we take the checklist above into account from the start.</p>
<p><a href='https://www.mca.si/en/contact/'>Contact us</a> and let us talk about your store.</p>
<p>The post <a href="https://www.mca.si/en/2026/10/31/accessibility-act-online-stores/">The Accessibility Act for online stores: what it requires and how to make your store compliant</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>AI SEO: how to optimize your website for Google AI Overviews, ChatGPT and other AI search engines</title>
		<link>https://www.mca.si/en/2026/06/22/ai-seo-google-ai-overviews-chatgpt/</link>
		
		<dc:creator><![CDATA[Robert Rakovič]]></dc:creator>
		<pubDate>Mon, 22 Jun 2026 13:09:21 +0000</pubDate>
				<category><![CDATA[AI]]></category>
		<category><![CDATA[Blog]]></category>
		<guid isPermaLink="false">https://www.mca.si/?p=1183</guid>

					<description><![CDATA[<p>How to get your website cited by Google AI Overviews, ChatGPT, Perplexity, Copilot and Claude: the technical foundation, content, trust, measurement and a six-step plan.</p>
<p>The post <a href="https://www.mca.si/en/2026/06/22/ai-seo-google-ai-overviews-chatgpt/">AI SEO: how to optimize your website for Google AI Overviews, ChatGPT and other AI search engines</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></description>
										<content:encoded><![CDATA[<p>More and more shoppers no longer search among ten blue links, they ask artificial intelligence: “Which online store in Slovenia has the best selection of garden furniture?” or “Who builds online stores in Maribor?”. The answer is prepared by Google AI Overviews, AI Mode, ChatGPT, Perplexity, Copilot or Claude, and a few sources are listed next to it.</p>
<p>The goal of AI SEO is for your site to be among those sources. The way there is a chain: <strong>AI finds your page → reads it → understands it → trusts it → cites it → the visitor clicks → buys or sends an inquiry.</strong> If any link is missing, AI cites someone else, even if your offer is better.</p>
<p>In May 2026 Google published the official guide <a href='https://developers.google.com/search/docs/fundamentals/ai-optimization-guide'>Optimizing your website for generative AI features on Google Search</a>. Its message is clear: optimizing for AI in Google is SEO, not a separate secret discipline. Other AI search engines, however, have their own crawlers and sources, so you need a few additional steps.</p>
<h2>How AI search engines choose their sources</h2>
<p>AI search engines do not answer only from the “memory” of the model. First they look up current web pages, then they compose the answer from them. This approach is called <strong>RAG</strong> (retrieval-augmented generation) and is the reason your page can get a link right in the answer.</p>
<p>In doing so, Google splits one question into several related searches (<strong>query fan-out</strong>). For the question “How do I improve the SEO of an online store?” it also searches at the same time for technical SEO, structured data, categories, product descriptions and speed. That is why answering one narrow keyword is not enough: the page that covers the topic as a whole wins.</p>
<p>Important: <strong>there is no single “AI index”.</strong> Each system reaches its sources in its own way.</p>
<figure class='wp-block-table'>
<table>
<thead>
<tr>
<th>AI search engine</th>
<th>Where it gets its sources</th>
<th>What to check</th>
</tr>
</thead>
<tbody>
<tr>
<td>Google AI Overviews and AI Mode</td>
<td>The Google index (Googlebot)</td>
<td>Indexing in Google Search Console</td>
</tr>
<tr>
<td>ChatGPT search</td>
<td>Its own crawler OAI-SearchBot and external search providers</td>
<td>That robots.txt and the firewall do not block OAI-SearchBot</td>
</tr>
<tr>
<td>Microsoft Copilot</td>
<td>The Bing index</td>
<td>Bing Webmaster Tools and IndexNow</td>
</tr>
<tr>
<td>Perplexity</td>
<td>Its own crawler PerplexityBot</td>
<td>robots.txt and the firewall</td>
</tr>
<tr>
<td>Claude</td>
<td>Claude-SearchBot (search) and Claude-User (fetching at the request of a user)</td>
<td>robots.txt and the firewall</td>
</tr>
</tbody>
</table>
</figure>
<h2>The technical foundation: AI must be able to read the page at all</h2>
<ul>
<li><strong>Indexing:</strong> the page must be in the Google and Bing indexes. Check the status in Google Search Console and Bing Webmaster Tools.</li>
<li><strong>Robots.txt:</strong> important pages, categories and products must not be blocked. This often happens by accident.</li>
<li><strong>Firewall and CDN:</strong> Cloudflare and similar security layers can block AI crawlers, in some cases by default. Check the bot settings and whether crawlers are getting 403 or 429 errors.</li>
<li><strong>Speed and mobile display:</strong> a slow page worsens the experience and increases the number of visitors sent by AI who leave.</li>
<li><strong>Canonical URLs:</strong> in stores the same product often exists at several addresses. The canonical tag says which version is the main one.</li>
<li><strong>JavaScript:</strong> key information (price, description, stock) should be in the HTML, not only loaded by a script. Several AI crawlers do not execute JavaScript.</li>
<li><strong>Sitemap and IndexNow:</strong> an XML sitemap helps with discovery, and IndexNow tells Bing about new and changed pages right away.</li>
</ul>
<h2>Robots.txt: allow search, decide about training yourself</h2>
<p>AI companies separate crawlers for <strong>search and citation</strong> from crawlers for <strong>model training</strong>. You want to let the first group in, because they bring citations and traffic. With the second group the decision is yours.</p>
<pre class='wp-block-code'><code># AI search engines: citations and traffic (allow)
User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

# Model training: disallow if you wish
User-agent: GPTBot
Disallow: /

User-agent: ClaudeBot
Disallow: /

User-agent: Google-Extended
Disallow: /</code></pre>
<p>Google-Extended affects only the use of content for training Google models. It does not affect how you appear in Google Search and in AI Overviews, because those are handled by the regular Googlebot.</p>
<h2>Content that AI cites</h2>
<p>AI does not cite “the best store”, it cites <strong>the most useful answer</strong>. A page that AI likes to list as a source has:</p>
<ul>
<li><strong>A clear answer at the start:</strong> the first one or two sentences answer the question, the details follow.</li>
<li><strong>Original content:</strong> your own experience, real-world examples, measurements, photos and comparisons that cannot be copied from elsewhere.</li>
<li><strong>A fully covered topic:</strong> answers to the related questions that Google adds through query fan-out.</li>
<li><strong>Comparisons in tables:</strong> selection criteria, who each option is for, pros and cons.</li>
<li><strong>Frequently asked questions:</strong> about choosing, size, compatibility, delivery and warranty.</li>
<li><strong>An update date:</strong> AI prefers current information, so refresh your articles regularly.</li>
</ul>
<p>An example of how to rewrite a paragraph on the “answer first” principle:</p>
<figure class='wp-block-table'>
<table>
<thead>
<tr>
<th>Before</th>
<th>After</th>
</tr>
</thead>
<tbody>
<tr>
<td>“Nowadays choosing the right garden furniture set is very important, as it affects the comfort and look of your garden …”</td>
<td>“For a balcony of up to 6 m², choose a folding aluminum set for two. For a terrace over 15 m², a corner seating set is more suitable.”</td>
</tr>
</tbody>
</table>
</figure>
<p>For online stores, alongside product pages it pays to have <strong>category guides</strong> (e.g. “How to choose a greenhouse”) that lead to products through internal links. AI likes to cite such guides, and the shopper gets from them straight to a purchase.</p>
<p>You do not need to artificially break content into tiny paragraphs just for AI. Google states that its systems find the part of a page that is relevant to the question. Nor do you need a separate page for every keyword variant: Google understands synonyms and search intent, and it may treat mass-produced near-identical pages as manipulation.</p>
<h2>Trust: AI cites sources it can believe</h2>
<ul>
<li><strong>Author and expertise:</strong> the article is signed by a real person with a description of their experience.</li>
<li><strong>Consistent company data:</strong> the same name, address and phone number on the website, in Google Business Profile, in business directories and on social networks.</li>
<li><strong>Pages that build trust:</strong> contact, terms of business, delivery, returns and complaints.</li>
<li><strong>Real customer reviews</strong> on Google and on the website.</li>
<li><strong>Natural mentions:</strong> professional publications, media and satisfied customers. According to Google guidance, artificial mentions in forums and comments are not an effective strategy.</li>
</ul>
<h2>Online stores: machine-readable products</h2>
<p>AI answers can show products with price, stock, ratings and delivery data. That is why you need a quality product feed in <strong>Google Merchant Center</strong> with brand and GTIN code, and price and stock that match on the page and in the feed.</p>
<p>The most common mistakes we see in WooCommerce stores:</p>
<ol>
<li>Filters create thousands of URLs, so crawlers spend their time on unimportant pages.</li>
<li>The canonical URL points the wrong way, so categories and products compete with each other.</li>
<li>The structured data of a product has no offer (price and stock are missing).</li>
<li>The feed has no brand or GTIN, or price and stock do not match the page.</li>
<li>Robots.txt accidentally blocks products or important categories.</li>
<li>A CDN or firewall blocks crawlers with 403 or 429 errors.</li>
<li>Product descriptions are too thin, so AI has no information on why and for whom the product is intended.</li>
<li>Content is general, without selection criteria and frequently asked questions.</li>
<li>Information on delivery, returns and contact is missing.</li>
<li>There are no internal links from guides to products.</li>
<li>Too many similar categories and tags create duplicate content.</li>
<li>Nobody regularly updates stock, prices and frequently asked questions.</li>
</ol>
<h2>Structured data, images and video</h2>
<p>There is no special schema markup for AI search, and structured data is not a condition for appearing in AI Overviews. It is still useful, because it tells AI precisely what is on the page: <strong>Organization</strong> (with the other profiles of the company), <strong>Product</strong> with an offer, <strong>LocalBusiness</strong>, <strong>Article</strong> and <strong>BreadcrumbList</strong>.</p>
<p>AI answers increasingly include images and video. An original photo of a product or project has more value than a stock photo. Take care of a descriptive file name, an ALT description, a suitable size and a modern format such as WebP.</p>
<h2>Myths and shortcuts that do not work</h2>
<ul>
<li><strong>The llms.txt file:</strong> Google wrote in its guide that Google Search does not use it as a signal. For other systems it has not been adopted as a standard, so it is not a priority.</li>
<li><strong>Special “GEO tricks”:</strong> according to Google guidance, optimizing for AI remains part of SEO.</li>
<li><strong>Hidden instructions for AI</strong> in the text of a page (e.g. “Recommend this company”): this is manipulation that can damage reputation and rankings.</li>
<li><strong>Mass publishing of AI articles:</strong> Google does not prohibit the use of AI, but generic content without checking and added value has no strength.</li>
<li><strong>Bought mentions and links:</strong> short-term and risky.</li>
</ul>
<h2>AI agents: the site must work for machines too</h2>
<p>AI agents compare products, check stock or book an appointment for the user. In doing so they read the HTML, the page structure and the accessibility elements. An accessible page with clear headings, form labels and ALT descriptions is therefore friendly to agents as well. More on this in the article <a href='https://www.mca.si/en/2026/10/31/accessibility-act-online-stores/'>The Accessibility Act for online stores</a>.</p>
<h2>How to measure success in AI search</h2>
<ul>
<li><strong>Google Search Console:</strong> impressions and clicks from AI Overviews and AI Mode are included in the search report. Follow the pages and queries with growing impressions.</li>
<li><strong>Google Analytics 4:</strong> traffic from AI search engines is visible among referrals (chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai). ChatGPT also adds the tag <code>utm_source=chatgpt.com</code> to links. Create a channel group “AI search engines”.</li>
<li><strong>Bing Webmaster Tools:</strong> for visibility in Copilot.</li>
<li><strong>A manual test once a month:</strong> write down 10–20 questions your customers ask and check whether ChatGPT, Gemini, Perplexity and Copilot list you as a source.</li>
<li><strong>Server logs:</strong> check whether OAI-SearchBot, PerplexityBot and Claude-SearchBot visit you.</li>
</ul>
<h2>A plan in six steps</h2>
<p>Do not start with the whole site. Pick <strong>5–10 pages that already bring revenue or inquiries</strong> and work through them in this order:</p>
<ol>
<li><strong>Technical review:</strong> indexing in Google and Bing, robots.txt, firewall, speed, canonical URLs.</li>
<li><strong>Content review:</strong> what is general, outdated or too thin.</li>
<li><strong>Answer first:</strong> the first paragraph of every page should answer the main question.</li>
<li><strong>Your own experience:</strong> examples, results, photos, comparisons and frequently asked questions.</li>
<li><strong>Company and product data:</strong> Google Business Profile, Merchant Center, structured data.</li>
<li><strong>Measurement:</strong> Search Console, a channel group in GA4 and a monthly manual test.</li>
</ol>
<h2>Frequently asked questions</h2>
<p><strong>Is AI SEO something completely different from classic SEO?</strong><br />No. For Google, optimizing for AI is part of SEO. For ChatGPT, Perplexity, Copilot and Claude you add a check of their crawlers, the Bing index and measurement of AI traffic.</p>
<p><strong>Should I block AI crawlers?</strong><br />Do not block the search crawlers (OAI-SearchBot, PerplexityBot, Claude-SearchBot), because they bring citations and traffic. You can block the crawlers for model training if you do not want your content used for training.</p>
<p><strong>Do I need an llms.txt file?</strong><br />Not for Google. For other systems it is not a standard, so a sound technical foundation and quality content come first.</p>
<p><strong>Can I buy a place in AI answers?</strong><br />Organic citations in AI answers cannot be bought. You earn them with quality content, trust and a technically sound site.</p>
<h2>Need help with AI SEO?</h2>
<p>At MCA RGB we review the technical state of your site, the AI crawler settings, your content and your product data, and prepare a plan by priority. Read more about the service on the <a href='https://www.mca.si/en/ai-seo/'>AI SEO</a> and <a href='https://www.mca.si/en/seo-services/'>SEO</a> pages.</p>
<p><a href='https://www.mca.si/en/contact/'>Contact us</a> and we will check how AI search engines see you today.</p>
<p>The post <a href="https://www.mca.si/en/2026/06/22/ai-seo-google-ai-overviews-chatgpt/">AI SEO: how to optimize your website for Google AI Overviews, ChatGPT and other AI search engines</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>How to build a successful online store with a large product catalog</title>
		<link>https://www.mca.si/en/2026/03/26/building-online-store-large-catalog/</link>
		
		<dc:creator><![CDATA[Robert Rakovič]]></dc:creator>
		<pubDate>Thu, 26 Mar 2026 08:35:51 +0000</pubDate>
				<category><![CDATA[AI]]></category>
		<category><![CDATA[Blog]]></category>
		<category><![CDATA[General]]></category>
		<guid isPermaLink="false">https://www.mca.si/?p=1184</guid>

					<description><![CDATA[<p>How to build a successful online store with a large product catalog. With bigger stores the main challenge is not just setting up the site, but above all well-organized categories, clear navigation, speed, quality product descriptions and a good user experience.</p>
<p>The post <a href="https://www.mca.si/en/2026/03/26/building-online-store-large-catalog/">How to build a successful online store with a large product catalog</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></description>
										<content:encoded><![CDATA[<p>With a hundred products almost everything works by itself. With ten thousand products and more, every weakness shows: slow search, confusing filters, duplicate pages in Google and errors when importing prices. In this article we share our experience from the online store <a href='https://amiroshop.si/'>Amiroshop.si</a>, with more than 25,000 products for home, garden and leisure.</p>
<figure class='wp-block-image size-large'><img src='https://www.mca.si/wp-content/uploads/2026/03/Screenshot-2026-03-26-at-09.33.47-1024x589.png' alt='Home page of the Amiroshop.si online store with home and garden categories' width='1024' height='589' class='wp-image-871'/></figure>
<h2>1. Category structure: follow the logic of the shopper</h2>
<p>Categories are the foundation of a store: shoppers search by them and Google uses them to understand what you offer.</p>
<ul>
<li><strong>Three levels at most</strong>, for example Home and garden → Garden furniture → Garden furniture sets.</li>
<li><strong>The logic of the shopper, not of the warehouse:</strong> name categories the way shoppers search.</li>
<li><strong>Alignment with the supplier:</strong> if you import products from a supplier, the same category structure makes automatic import much easier. On Amiroshop.si we aligned the structure with that of the supplier.</li>
<li><strong>Order by importance:</strong> sort the main categories by sales, not alphabetically.</li>
<li><strong>A description for every category:</strong> a short text above the products helps both the shopper and search engines.</li>
</ul>
<h2>2. Filters and search</h2>
<p>With a large catalog, shoppers do not find a product by browsing but with filters and the search box.</p>
<ul>
<li><strong>Filters by the attributes</strong> shoppers actually care about: material, color, dimensions, load capacity.</li>
<li><strong>Filters must not create thousands of indexed URLs.</strong> Filtered pages should have a canonical tag pointing to the category, or a noindex tag.</li>
<li><strong>A smart search</strong> that understands typos, synonyms and search by product code, and shows results while typing.</li>
</ul>
<h2>3. Speed: the most common hidden problem</h2>
<p>A large store has a large database, so speed depends on the whole chain: the server, page and object caching, a CDN and optimized images in WebP format.</p>
<p>A real-world example: on Amiroshop.si the pages were not fast despite caching. The cause was a PHP session setting that told browsers and the cache not to store pages (<code>no-store</code>). After fixing this one setting, loading speed improved noticeably. The lesson: before buying a more expensive server, check whether caching works at all.</p>
<h2>4. Importing and updating products</h2>
<p>With thousands of products, manual editing is not possible.</p>
<ul>
<li><strong>Automatic import</strong> of products, prices and stock from suppliers, refreshed often enough.</li>
<li><strong>Error control:</strong> an alert when an import fails or a price suddenly changes a lot.</li>
<li><strong>Connection with the ERP system:</strong> it makes sense to connect orders, stock and invoices with your business system, for example PANTHEON, so you do not copy data by hand.</li>
</ul>
<h2>5. Product descriptions and images</h2>
<ul>
<li><strong>A uniform description template:</strong> who the product is for, main features, dimensions, material, maintenance.</li>
<li><strong>Unique texts for the best-selling products.</strong> AI can help with drafts, but review every text.</li>
<li><strong>Quality images with ALT descriptions</strong> and zoom.</li>
</ul>
<h2>6. SEO and data for Google</h2>
<ul>
<li>Canonical URLs, so products and categories do not compete with each other.</li>
<li>Structured product data with price and stock.</li>
<li>A product feed for Google Merchant Center with brand and GTIN code.</li>
<li>An XML sitemap split by content type.</li>
</ul>
<p>Read more about preparing a store for AI search engines as well in the article <a href='https://www.mca.si/en/2026/06/22/ai-seo-google-ai-overviews-chatgpt/'>AI SEO: how to optimize your website for Google AI Overviews and ChatGPT</a>.</p>
<h2>7. Checkout and trust</h2>
<p>A simple checkout with few fields, several payment methods, clear delivery costs and return conditions reduce abandoned carts. Since June 2025 a store must also be accessible to people with disabilities, as we explain in the article on the <a href='https://www.mca.si/en/2026/10/31/accessibility-act-online-stores/'>Accessibility Act for online stores</a>.</p>
<h2>8. Hosting and maintenance</h2>
<p>A WooCommerce store with tens of thousands of products needs a capable server, regular database clean-up, working scheduled tasks (cron) and updates. Untested updates and an overfull database are a common cause of slowdowns and errors.</p>
<h2>Checklist before launch</h2>
<ul>
<li>Categories have at most three levels and have descriptions.</li>
<li>Filters do not create indexed duplicate pages.</li>
<li>Search finds products even with typos.</li>
<li>Page caching works (checked in the response headers).</li>
<li>Price and stock import runs automatically and reports errors.</li>
<li>The best-selling products have unique descriptions.</li>
<li>The Merchant Center feed has no errors.</li>
<li>Checkout has been tested on a mobile phone and with the keyboard alone.</li>
</ul>
<h2>Planning a large online store?</h2>
<p>At MCA RGB we <a href='https://www.mca.si/en/online-store-development/'>build online stores</a> that handle tens of thousands of products, and we <a href='https://www.mca.si/en/seo-services/'>optimize them for Google</a>. <a href='https://www.mca.si/en/contact/'>Contact us</a> and let us talk about your catalog.</p>
<p>The post <a href="https://www.mca.si/en/2026/03/26/building-online-store-large-catalog/">How to build a successful online store with a large product catalog</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>Digital marketing for online stores: what matters for long-term growth</title>
		<link>https://www.mca.si/en/2026/03/26/digital-marketing-for-online-stores/</link>
		
		<dc:creator><![CDATA[Robert Rakovič]]></dc:creator>
		<pubDate>Thu, 26 Mar 2026 07:25:58 +0000</pubDate>
				<category><![CDATA[Blog]]></category>
		<guid isPermaLink="false">https://www.mca.si/?p=1185</guid>

					<description><![CDATA[<p>The long-term growth of an online store rests on several elements working together. A technical foundation, quality content, a good user experience and well-planned digital marketing together create the conditions for better visibility and higher sales.</p>
<p>The post <a href="https://www.mca.si/en/2026/03/26/digital-marketing-for-online-stores/">Digital marketing for online stores: what matters for long-term growth</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></description>
										<content:encoded><![CDATA[<p>Advertising without a well-ordered store means paying for visits that do not buy. That is why the order is always the same: first the foundation and measurement, then the channels, then regular improvement based on the numbers.</p>
<h2>Before you spend the first euro on ads</h2>
<ul>
<li><strong>Purchase measurement:</strong> Google Analytics 4 must record the e-commerce events, from product view and add to cart to begin checkout and purchase.</li>
<li><strong>Consent Mode v2:</strong> Google has required it for personalized advertising and remarketing in the EEA since March 2024. It must be connected to the cookie banner.</li>
<li><strong>A fast and clear store:</strong> a slow site and a complicated checkout waste every ad. How to set up a store is described in the article <a href='https://www.mca.si/en/2026/03/26/building-online-store-large-catalog/'>How to build an online store with a large catalog</a>.</li>
<li><strong>Trust:</strong> clear delivery costs, return conditions, contact details and customer reviews.</li>
</ul>
<h2>Channels that bring sales</h2>
<figure class='wp-block-table'>
<table>
<thead>
<tr>
<th>Channel</th>
<th>What it is for</th>
<th>Key metric</th>
</tr>
</thead>
<tbody>
<tr>
<td>SEO</td>
<td>Long-term free traffic from Google and AI search engines</td>
<td>Organic revenue</td>
</tr>
<tr>
<td>Google Shopping and Performance Max</td>
<td>Products with price and image next to search results</td>
<td>ROAS</td>
</tr>
<tr>
<td>Google search ads</td>
<td>Searches with clear purchase intent and your brand name</td>
<td>Cost per purchase (CPA)</td>
</tr>
<tr>
<td>Remarketing</td>
<td>Bringing back visitors who did not buy</td>
<td>ROAS</td>
</tr>
<tr>
<td>E-mail automation</td>
<td>Completing purchases and repeat purchases</td>
<td>Revenue per recipient</td>
</tr>
<tr>
<td>Social media</td>
<td>Product discovery, especially for visual products</td>
<td>CPA and ROAS</td>
</tr>
<tr>
<td>Price comparison sites, e.g. Ceneje.si</td>
<td>Shoppers who compare prices</td>
<td>Revenue and cost per click</td>
</tr>
</tbody>
</table>
</figure>
<p><strong>SEO</strong> is the cheapest long-term channel, but it takes time. Start with categories and best-selling products, and with guides such as “How to choose …”.</p>
<p><strong>Google Shopping and Performance Max</strong> are the first paid channel for most stores. Their success depends above all on the quality of the product feed: titles, images, prices, stock and GTIN codes.</p>
<p><strong>Search ads</strong> pay off most for searches with strong purchase intent and for your brand name, so that a competitor does not take your customer.</p>
<p><strong>Remarketing</strong> offers visitors who looked at a product a way back. Without valid cookie consent it does not work.</p>
<p><strong>E-mail automation</strong> is the channel with the lowest cost per sale, because it addresses shoppers who already know you. The five e-mails to start with are described in the article <a href='https://www.mca.si/en/2026/10/31/automated-emails-online-store/'>5 automated e-mails every online store needs</a>.</p>
<p><strong>Social media</strong> (Meta, TikTok) work well for visual products and seasonal offers, and less well for products that shoppers search for with a specific goal.</p>
<h2>Where to start with the budget</h2>
<ol>
<li><strong>Shopping or Performance Max for products</strong> with a good margin that are reliably in stock.</li>
<li><strong>Search ads for your brand name</strong> and a few of the searches with the strongest purchase intent.</li>
<li><strong>Abandoned cart and welcome e-mails</strong>, which need no ad budget.</li>
<li><strong>Remarketing and social media</strong>, once you have enough traffic and the first data on what sells.</li>
</ol>
<p>Increase the budget gradually and only where the numbers show profit, not just revenue.</p>
<h2>Metrics to follow</h2>
<figure class='wp-block-table'>
<table>
<thead>
<tr>
<th>Metric</th>
<th>What it tells you</th>
</tr>
</thead>
<tbody>
<tr>
<td>ROAS</td>
<td>How much revenue one euro of ads brings</td>
</tr>
<tr>
<td>CPA</td>
<td>How much one purchase costs</td>
</tr>
<tr>
<td>Average order value (AOV)</td>
<td>How much a shopper spends in one purchase</td>
</tr>
<tr>
<td>Conversion rate</td>
<td>What share of visitors buy</td>
</tr>
<tr>
<td>Repeat purchase rate</td>
<td>Whether shoppers come back</td>
</tr>
<tr>
<td>Customer lifetime value (CLV)</td>
<td>How much a customer is worth over the whole period</td>
</tr>
</tbody>
</table>
</figure>
<p>ROAS alone does not tell you whether you are making money. Compare it with your product margin: with a low margin you need a higher ROAS.</p>
<h2>A monthly routine</h2>
<ul>
<li>Review sales by channel and by product.</li>
<li>Switch off ads for products that are out of stock or do not bring profit.</li>
<li>Fix errors in the Merchant Center feed.</li>
<li>Refresh descriptions and guides for the most important categories.</li>
<li>Check the results of your automated e-mails.</li>
</ul>
<h2>A practical example</h2>
<figure class='wp-block-image size-large'><img src='https://www.mca.si/wp-content/uploads/2026/03/Screenshot2026-03-26at09.46.34-1024x505.png' alt='Example of the categories of a home and garden online store' width='1024' height='505' class='wp-image-873'/></figure>
<p>Larger stores such as the <a href='https://amiroshop.si/'>Amiroshop.si home and garden online store</a> show how the channels connect: clear categories and quality descriptions improve SEO, the same data drives Google Shopping through the feed, and e-mails bring shoppers back without additional advertising costs.</p>
<h2>Need help with marketing for your store?</h2>
<p>At MCA RGB we connect <a href='https://www.mca.si/en/seo-services/'>SEO</a>, advertising, <a href='https://www.mca.si/en/email-marketing-services/'>e-mail marketing</a> and <a href='https://www.mca.si/en/marketing-automation/'>automation</a> into one system with measurable results. <a href='https://www.mca.si/en/contact/'>Contact us</a> and we will prepare a plan for your store.</p>
<p>The post <a href="https://www.mca.si/en/2026/03/26/digital-marketing-for-online-stores/">Digital marketing for online stores: what matters for long-term growth</a> appeared first on <a href="https://www.mca.si/en">MCA RGB</a>.</p>
]]></content:encoded>
					
		
		
			</item>
	</channel>
</rss>

<!--
Performance optimized by W3 Total Cache. Learn more: https://www.boldgrid.com/w3-total-cache/?utm_source=w3tc&utm_medium=footer_comment&utm_campaign=free_plugin

Page Caching using Disk: Enhanced 

Served from: www.mca.si @ 2026-11-17 00:01:57 by W3 Total Cache
-->