SEO for AI: How to Rank Your Website in ChatGPT, Gemini and Google AI

If two years ago your only goal was to appear on Google's first page, the landscape has shifted. More and more users now ask their questions directly to ChatGPT, Gemini, Perplexity or the Google AI Overviews, and they no longer click through anywhere. SEO for AI (often referred to as AEO or GEO) is the adaptation of traditional ranking practices to this new context and, for a Catalan SME, it is no passing trend: it makes the difference between AI surfacing your brand when someone asks 'which SEO agency would you recommend in Barcelona?' and the answer sending the client to a competitor.

What SEO for AI is

SEO for AI is the set of techniques that make a brand, a service or a piece of content get cited or used as a source within the answers generated by conversational models (ChatGPT, Gemini, Claude, Perplexity) and the new AI formats inside Google, such as AI Overviews. You'll also see the acronyms AEO (Answer Engine Optimization) and GEO (Generative Engine Optimization). They have nuances, but in day-to-day work they describe the same problem: your content has to be easy for a machine to summarise and cite with confidence.

In our experience auditing Catalan SMEs, it is common to find sites that perform well on classic Google but never appear inside a ChatGPT conversation. The cause is almost always the same: the content is written 'for SEO', not to answer the specific questions a user would ask an AI.

How ChatGPT, Gemini and Perplexity choose answers

Each engine has its own method, but they all draw from three similar wells:

• Their own training: public web content up to a cut-off date.
• Real-time search: Bing for ChatGPT, Google for Gemini, in-house indexes for Perplexity.
• Trust signals: domain reputation, content clarity, structured data and external mentions.

When someone asks 'recommend an SEO agency in Catalonia', the model summarises the sources it considers most reliable and tries to cite them. If your site is the clearest and the most cited by others, you get in. If your text needs interpretation, it skips you.

Differences from classic SEO

What I often see is that marketing teams don't know how to measure this new layer and confuse 'not appearing on Google' with 'not existing for AI'. They are different battles.

7 actions to appear in AI answers

1. Direct answer up front: the first 60-80 words of every article should answer the main question without padding. Models often extract that first pass.
2. H2 headings as real questions: 'How much does it cost…?', 'What happens if…?', 'Why…?'. These are the forms people use with a conversational AI.
3. Structured data (Schema.org): FAQPage, HowTo, Article, LocalBusiness, Organization. They give the model context to cite you confidently.
4. Cite sources and be citable: publish your own figures, cases and clear opinions with your name or your company's. AIs tend to cite content with identifiable authorship.
5. External mentions (not just backlinks): sector directories, local Catalan press, niche podcasts. AI reads mentions, not only links.
6. A serious 'About' page: who you are, where you are, years of experience, team. It boosts EEAT and helps models treat you as a reliable source.
7. Fast, indexable site: bots (GPTBot, Google-Extended, ClaudeBot, PerplexityBot) drop content that loads slowly or needs heavy JS rendering.

If you'd like help applying these actions to your site, we offer a free SEO audit that already includes a specific AI visibility layer. We tell you where you stand and what to fix first.

Common mistakes I see in audits

• Blocking AI bots in robots.txt: legitimate if you want to protect content, but blocking everything kicks you out of the conversation. I recommend allowing GPTBot and Google-Extended on the sections where you want visibility.
• Content fully inside JavaScript: single-page apps without server-side rendering. Conversational bots make little rendering effort.
• Long, diluted articles: 2,500 words without a clear structure leave the model with nothing usable.
• No schema: sites with plenty of useful content but no FAQPage or HowTo waste many citation chances.
• No name behind it: articles signed 'Team', no author page. Low EEAT and a lower chance of being cited.

How to measure AI visibility

There is no official Search Console yet, but useful indicators do exist:

• Manual probes: run your own searches in ChatGPT, Gemini, Perplexity and Google AI Overviews using your business keywords. Note whether you are cited, how you are described and in which position.
• Server logs: check visits from GPTBot, Google-Extended, ClaudeBot and PerplexityBot. If they don't come in, they can't cite you.
• External mentions: monitor your brand name with Google Alerts or tools such as BrandMentions.
• Tagged traffic: some models add an utm_source parameter when they send a click; it is still partial, but it is starting to show up in Analytics.

Practical cases in Catalonia

In recent projects we have seen clear patterns. A dental clinic in Tarragona that already ranked well on Google started appearing on Perplexity and Gemini once we rewrote its treatment pages with a question-answer format and FAQPage schema. An e-commerce site in Sabadell with a busy blog but no visible authors raised its AI visibility within six weeks just by adding an author page and a biography signed by its technical expert. And in a Girona law firm, simply allowing GPTBot and Google-Extended (which were blocked by default) made the firm start showing up inside answers about local regulations.

The common thread: not a single new article. They rewrote and marked up better what they already had. I always recommend starting there before investing in more content.