More and more users are asking ChatGPT, Gemini or Perplexity directly instead of opening Google. And when they do open Google, they often find AI Overviews providing answers before any organic results appear. User search behaviour is changing, and so are the answers they receive. This shift has the greatest impact on informational traffic.
In this context, terms such as GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization) have emerged, accompanied by countless guides promising “the new way to rank”. Google has just published an official guide clarifying what is hype and what actually works to appear in its generative AI features. In this article, we review what Google says and which actions to prioritise depending on your situation.
What is SEO for AI and how does it differ from traditional SEO?
SEO for AI is the set of practices aimed at getting models such as ChatGPT, Gemini, Perplexity, Claude or Copilot to cite your website as a source when generating their responses.
Several names for the same thing are circulating online:
- GEO: Generative Engine Optimization
- AEO: Answer Engine Optimization
- LLMO: Large Language Model Optimization
Each term focuses on a slightly different aspect, but in practice they all refer to the same thing.
Google has been clear in its official documentation: optimising for generative AI search is still SEO, not a separate discipline. Its AI features, AI Overviews and AI Mode, rely on the same ranking and quality systems as traditional search. This rules out much of the speculation and many of the techniques being marketed as “GEO-exclusive”.
The real difference lies in the end goal. Traditional SEO aims for the top positions in organic search results and measures success through clicks. SEO for AI aims to be included in the answer paragraph generated by the model, and success is often measured through mentions without clicks.
The foundations, however, remain the same: useful content, a clear technical structure and demonstrable authority.
How do ChatGPT, Gemini and other AI tools decide which websites to rank?
Not all AI tools work in the same way. There are three different architectures, and understanding them is important because each requires a different approach. Treating all models as if they worked identically is the most common mistake in generic guides.
AI tools with real-time web browsing
ChatGPT Search, Perplexity, Gemini or Claude with Search enabled, and Copilot, browse the web at the time of the query. When a user asks a question, the model performs a search, reads the most relevant pages and generates a response with citations.
This means that to appear, you need three things: your content must be properly indexed in the search engines they use as sources (Google for Gemini and AI Overviews, Bing for ChatGPT and Copilot, and proprietary indexes for Perplexity), their bots must be able to access your website without restrictions, and your content must be valuable and authoritative, written by real people who are experts in their field. If you meet all these requirements better than your competitors, you are already well on your way.
AI tools without browsing enabled by default
Claude, or ChatGPT without the search option enabled, respond based on their training. In other words, based on what they learned when the model was trained.
Here, the logic changes completely: appearing does not depend on recent indexing, but on your brand’s presence in sources that the model considered reliable during its training. This is a long-term effort, taking months rather than immediate actions.
The special case of Google AI Overviews
AI Overviews are the summarised answers that appear at the top of Google search results, generated by Gemini using Google’s own index. This is the most important case for any business with an audience in Spain, because it directly affects the searches that already bring you traffic today.
Google explains in its documentation that, to generate these responses, it does not make anything up: it takes content from pages that already rank in its index and summarises it. In addition, it runs several related searches simultaneously in the background to expand the context before responding. As a result, if your website ranks well on Google, you have a good chance of appearing in AI Overviews without any extra work. On the other hand, when you appear, users get the answer without necessarily clicking through.
What factors determine whether an AI cites your website (and what mistakes leave you out)?
With the theory clear, let’s get straight to the point. These are the six factors that really make a difference, with important nuances from Google’s official position where applicable.
1. Unique, useful and non-generic content
This is the most important factor, according to Google itself. A post titled “Benefits of digital marketing for your business” cannot compete with one titled “What changed in our results when we stopped running Meta ads for 3 months”. The first brings together information that can already be found on a hundred websites. The second provides something that only someone who experienced it can share.
The official recommendation is clear: create content that offers something a generative model could not easily produce. If your article can be written simply by asking ChatGPT, you will not rank or be cited. The key here is first-hand experience, verifiable data and real examples.
2. Clear and readable structure
AI tools process your content better when it is well organised: separate paragraphs, descriptive headings with keywords, lists where appropriate and tables when making comparisons. Google confirms this in its guide and adds an important nuance: there is no need to artificially break your content into small blocks specifically so that AI can “understand it better”. Its systems can understand content within a long page without any problems.
Write for your human audience, not for the model. If the result is readable for a person, it can be processed by an AI.
3. Verifiable and up-to-date data
Content with real data carries more weight than vague content. Writing “many companies are adopting AI” is not the same as saying “6 out of 10 marketing agencies in Spain already use ChatGPT in their daily work (source: IAB Spain study, 2026)”. The second statement is something an AI can cite; the first adds nothing.
Three things you should always include: specific figures, a source for each data point and a visible update date within the article itself.
4. Demonstrable brand authority
AI tools without real-time browsing (Claude, ChatGPT without search) depend almost entirely on how the rest of the web mentions you during their training: Wikipedia, industry publications, established technical forums and specialised directories.
Important: do not fabricate authority, build authority. Google explicitly warns against techniques aimed at “seeking inauthentic mentions”: buying reviews, creating fake accounts on forums, paying for comments on Reddit or Quora, or setting up satellite blogs that link to your website. Its systems detect these artificial patterns and ignore or penalise them.
Real authority is built through genuine presence: publishing content that others want to cite, participating in real industry discussions and appearing in the media because a journalist sought you out. It is slower, but it is the only approach that AI tools recognise as legitimate. Fabricated mentions not only fail to work: they damage your brand in the medium term.
5. Solid technical foundations
A fast website, a good mobile experience, crawlable content and reasonable HTML. Nothing new compared to traditional SEO. Google is unequivocal: if your website meets the technical requirements for Search and is indexed, it meets the requirements to appear in generative AI features.
Check Search Console regularly and resolve indexing errors. And verify that Google-Extended is not blocked in your robots.txt if you want to appear in Google’s AI experiences.
6. Allowing AI bots to access your website
Many websites block AI bots in their robots.txt, sometimes because of a hosting provider’s decision and sometimes because of settings inherited from security plugins. If a bot cannot read your website, it cannot cite you.
The main bots you should allow:
User-agent: GPTBot
Allow: /
User-agent: OAI-SearchBot
Allow: /
User-agent: ChatGPT-User
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: Google-Extended
Allow: /
User-agent: CCBot
Allow: /
Check your robots.txt today. If you have a Disallow: / directive for any of these user agents, you are out of the game before you even start.
Common mistakes that leave you out
Those are the six factors that make a difference. On the other hand, these are the most common mistakes we see in AI visibility audits. Fix them before investing in advanced tactics:
| Mistake | Why it happens | Solution |
| Blocking GPTBot, ClaudeBot, Google-Extended | Hosting providers or security plugins block them by default | Edit robots.txt and allow all AI bots |
| Critical content in JavaScript | AI bots render JS less effectively than Google | Serve HTML on the initial render |
| Information only available in PDFs | PDFs are black boxes for models | Always publish an equivalent HTML version |
| Not being indexed in Bing | ChatGPT and Copilot depend on Bing | Verify your domain in Bing Webmaster Tools |
| Generic content without data | AI tools ignore vague claims | Cifras concretas + fuentes + fechas visibles |
| Not updating old content | AI tools prioritise fresh content | Review schedule every 6 months |
| Paying for tools with Google’s “internal data” | Nadie tiene acceso a esas señales | Use Search Console as the official source |
What does Google say DOES NOT work? Common GEO myths
In recent months, dozens of “GEO tactics” have circulated, promising visibility in ChatGPT, Gemini or AI Overviews. Google has officially clarified which ones do not work. Reviewing them will help you avoid wasting time and budget on hype.
1. Google does not use the llms.txt file
The llms.txt file is a plain text file placed in the root directory of a domain (yourdomain.com/llms.txt) to tell language models which content on your website is a priority and how to interpret it. It works as a complementary guide to robots.txt, but specifically aimed at AI tools.
The problem is that Google has stated it explicitly: its Search completely ignores llms.txt and other “AI-specific” text files. You do not need to create one to appear in AI Overviews or any other Google generative AI feature. Whether or not it exists on your website makes no difference from Google’s perspective.
¿Entonces para qué sirve? Puede tener sentido si trabajas con sistemas de terceros que sí lo consultan (algunos ecosistemas específicos lo consideran).
2. Structured data is not mandatory
The SEO community has spent months insisting on implementing FAQPage, Article, HowTo and other schemas specifically for AI Overviews. Google clarifies: structured data is not mandatory for generative AI search. It is still recommended for other reasons (rich results in the SERP), but it is not the key factor it has been marketed as for GEO.
That said, FAQPage and Organization still provide value because they structure information that is useful for direct citations by AI tools with browsing capabilities, such as ChatGPT or Perplexity. Google’s official position is that they are not mandatory, not that they are useless.
3. Third-party tools with Google’s “internal metrics”
Ninguna herramienta de terceros tiene acceso a los datos internos de Google. No third-party tool has access to Google’s internal data. If a tool promises you a “Google AI visibility score” based on privileged signals, it is lying. Google has officially stated this.
There is only one reliable source for measuring your presence in AI Overviews, and it is free: Search Console.
How to optimise for each AI specifically?
Each model has its own particularities. Before going into detail, this table summarises which sources each AI relies on, which bot you should allow and where you should invest your time first:
| IA | Main source | Bot to allow | Priority |
| ChatGPT / ChatGPT Search | Bing + proprietary index | GPTBot, OAI-SearchBot, ChatGPT-User | High |
| Google AI Overviews | Google index | Google-Extended | High |
| Gemini | Google index + proprietary training | Google-Extended | Medium (also covered through AI Overviews) |
| Perplexity | Proprietary index + web crawling | PerplexityBot | Medium |
| Copilot | Bing | Microsoft/Bing bots | Media (incluida vía ChatGPT) |
| Claude | Datos con los que fue entrenado | ClaudeBot | Low (long-term work) |
Below, we look at each one in detail, with a specific action you can take this week.
SEO for ChatGPT and ChatGPT Search
ChatGPT Search relies on Bing as its main source and supplements it with GPTBot for its own index. The critical difference: ranking well on Bing is essential, and hardly anyone works on it.
Action this week: go to Bing Webmaster Tools, verify your domain and check that it is indexed. If it is not, import your property from Google Search Console (Bing allows you to do this in one click).
SEO for Google AI Overviews
It relies on Google’s traditional index combined with Gemini. If you already rank in the top 10 for a keyword, you have a high probability of being cited in the corresponding AI Overview. Google recommends using the Generative AI Performance report in Search Console to measure this.
Action this week: identify which queries in your industry already trigger AI Overviews (Semrush and Ahrefs explicitly flag this in their reports) and prioritise those pages to strengthen unique and useful content.
SEO for Perplexity
Perplexity usa su propio bot (PerplexityBot) y combina rastreo propio con índices externos. Valora especialmente fuentes con datos originales, estudios propios o cifras verificables.
Action this week: if your website publishes any original data (your own surveys, benchmarks, case studies with figures), make sure it is well structured and has a visible date. Perplexity cites this type of content before more general sources.
SEO for Gemini (outside Search)
Gemini relies on Google’s index and its own training. The bot to allow is Google-Extended, which is the specific opt-in for training Google’s models.
Action this week: in practice, working on Gemini separately adds very little extra value compared to working on AI Overviews. Prioritise AI Overviews; Gemini is included in the package.
SEO for Claude
Claude does not browse the web by default in most of its products: it responds based on its training. Appearing here requires sustained brand mentions in reliable sources over months or years.
Action this week: allow ClaudeBot in your robots.txt for future versions with real-time browsing, and work on your presence in sources Anthropic considers reliable (Wikipedia, major industry publications, technical forums). Do not expect immediate results.
SEO for Copilot
Almost identical to ChatGPT: it relies directly on Bing. If you have worked on your Bing presence for ChatGPT, you already have Copilot covered.
Action this week: the same as for ChatGPT. Work on Bing and kill two birds with one stone.
How to measure whether your website appears in AI tools?
Measuring GEO is not like measuring SEO. Clicks no longer tell the whole story because many mentions happen without the user ever clicking through. These are the four methods we combine at VILAX to get an accurate picture.
Manual recurring prompts
It is the most straightforward method and costs nothing. Define between 10 and 15 prompts that an ideal customer would use to find businesses in your sector and run them monthly in ChatGPT, Gemini, Perplexity and Claude. Record whether your brand appears and in what context.
The types of prompts you should include:
- Category: “Best [service] for [use case]”.
- Comparison: “[Brand A] vs [Brand B]”.
- Problem-solution: “How do I [task] without [restriction]”.
- Local: “Who offers [service] in [city]”.
Generative AI Performance report in Search Console
Search Console includes a specific Performance report for Google Search’s generative AI features. It shows impressions, clicks and CTR from AI Overviews and AI Mode. It is the only official source and is free.
Enable it, review it monthly and compare it with your traditional performance reports. The gap between generative impressions and clicks tells you how much brand visibility you are building without direct cost.
GA4: traffic from AI tools
Although the volume may be small, some users do click through from AI responses to your website. You can filter this traffic in GA4 by creating a segment with the following referral sources: chatgpt.com, perplexity.ai, gemini.google.com, claude.ai and copilot.microsoft.com.
This traffic is low in volume but high in intent: users arriving from an AI have already received an implicit recommendation. It tends to convert better than general organic traffic.
Third-party tools (used wisely)
There are specialised tools for monitoring AI visibility: Profound, Peec AI, Otterly, Ahrefs Brand Radar and Semrush AI Toolkit.
An important warning: Google has clarified that no third-party tool has access to its internal systems. If one promises an “exclusive Google AI score” based on supposed internal signals, that is marketing. Use them as a complement, not as your single source of truth. For most SMEs just starting out, the manual method + Search Console + GA4 covers 80% of the value.
SEO plan for AI if you have limited time: where to start
If you have made it this far and are thinking “OK, this all sounds good, but I don’t have 40 hours to implement it”, this is the minimum viable plan.
This week (3 actions):
- Audit your robots.txt and make sure no AI bots are blocked.
- Verify your domain in Bing Webmaster Tools and confirm that it is indexed.
- Enable and review the Generative AI Performance report in Search Console to establish a baseline.
This month (3 actions):
- Identify your 5 pages with the most organic traffic and strengthen their unique content (original data, real examples, first-hand experience).
- Define 10 prompts related to your industry and start testing them monthly in ChatGPT, Gemini and Perplexity.
- Build your presence in reliable industry sources (publications, established forums, Wikipedia where applicable) to develop brand authority in the medium term.
With these six actions, you cover 70% of the real impact of SEO for AI. The rest is refinement and maintenance.
Conclusion
SEO for AI is neither a passing trend nor the death of traditional SEO. Google itself has clarified that, from its perspective, it is still SEO. What changes is not the foundations (useful content, solid technical foundations, demonstrable authority), but how answers are presented to users and how visibility is measured.
Businesses that work on both layers, well-executed traditional SEO and specific optimisation for other AI tools with browsing capabilities (ChatGPT, Perplexity, Claude), will be the ones that capitalise on this shift. Those that focus on implementing every GEO trend without a clear strategy will waste time and budget.
At VILAX, we analyse whether your website already appears in the main AI tools, what is missing and which actions to prioritise depending on your industry. If you would like a review, get in touch.
