AI SEO (also called GEO, generative engine optimization) means structuring your website's content so that AI-powered search – Google's AI Overviews, ChatGPT search or Perplexity – can find it, understand it and cite it as a source in their answers. It doesn't replace classic SEO; it builds on it. The goal isn't just a blue link in the results, but being the page behind the answer the machine writes.
The term "GEO" took off after a 2023 research paper (Aggarwal et al., KDD 2024). Since then, English-language advice on what AI "likes" has multiplied. Data for Hungarian search is scarce, so we measured it ourselves: who does Google's AI Overview cite for Hungarian, service-type searches?
- On 22 October 2026 we pulled Google results pages for 12 Hungarian service keywords. All 12 showed an AI Overview.
- Of the 29 source domains cited in those AI Overviews, 7 (24%) ranked in the organic top 3 and 15 (52%) appeared anywhere on the first organic page. 14 sources (48%) were not on the first organic page at all.
- For "ai seo", "llms.txt" and "seo audit", not a single .hu domain appeared on the first organic page.
- A later request for the same keyword returned three entirely different sources. Track trends, not one-off snapshots.
How did we measure?
We used the DataForSEO SERP API to request Google Hungary's desktop results page with Hungarian language settings on 22 October 2026, once per keyword. The 12 keywords (in Hungarian): website design, webshop development, marketing agency, search engine optimization, SEO agency, ai seo, ai marketing, llms.txt, custom software development, software development, seo audit and AI marketing agency. Together they account for roughly 7,800 searches a month (DataForSEO, based on Google Ads monthly averages).
On each results page we recorded three things: whether an AI Overview appeared, which domains it cited, and what the first organic page looked like (8–10 organic results per page). We matched at domain level: we stripped the www. prefix and mapped subdomains to their root domain (for example, app.neilpatel.com counts as neilpatel.com).
The limits are worth stating plainly. One point in time, desktop only, 12 keywords: this is not a representative sample, it's an indicative snapshot. For two keywords (search engine optimization, custom software development) the API reported an AI Overview but returned no source domains, so we left them out of the source analysis. Source figures are therefore based on 10 keywords.
How common are AI Overviews in Hungarian search?
In our sample, they appeared on all 12 results pages. That holds for the highest-volume keyword, "weboldal készítés" (website design, about 4,400 searches a month), and for the small "ai marketing ügynökség" (AI marketing agency, about 70). So they don't only show up for "what is it, how does it work" questions, but also where people are looking for a provider and where ads compete for attention.
That doesn't mean every Hungarian search has one. It does mean that for a Hungarian service business, the AI Overview is no longer an experiment: it's the block at the top of the page that people read before they click anything.
As for how it works, Google's official guidance says AI Overviews and AI Mode may use a "query fan-out" technique: issuing multiple related searches across subtopics and data sources to build the answer. That's why the set of cited links can be wider and more varied than the classic ten blue links.
Who does the AI Overview cite?
Across the 10 keywords where a source list was available, we counted 29 cited domains (2–3 per keyword). Of these:
- 7 (24%) ranked in the organic top 3,
- 15 (52%) appeared somewhere on the first organic page,
- 14 (48%) were not on the first organic page at all.
Per keyword: in 5 cases at least one AI source was in the top 3, while in 4 cases none of the sources appeared on the first organic page (website design, webshop development, llms.txt, seo audit).
| Keyword (Hungarian) | Monthly searches | AI sources | In top 3 | On page 1 | .hu on page 1 |
|---|---|---|---|---|---|
| weboldal készítés | 4,400 | 3 | 0 | 0 | 6/9 |
| webshop készítés | 880 | 3 | 0 | 0 | 9/9 |
| marketing ügynökség | 720 | 3 | 2 | 3 | 9/10 |
| keresőoptimalizálás | 590 | n/a | – | – | 7/10 |
| seo ügynökség | 260 | 3 | 1 | 3 | 6/10 |
| ai seo | 170 | 3 | 1 | 1 | 0/8 |
| ai marketing | 170 | 3 | 0 | 1 | 2/8 |
| szoftverfejlesztés | 170 | 3 | 1 | 3 | 6/8 |
| llms.txt | 140 | 2 | 0 | 0 | 0/8 |
| egyedi szoftverfejlesztés | 110 | n/a | – | – | 6/8 |
| seo audit | 90 | 3 | 0 | 1 | 0/9 |
| ai marketing ügynökség | 70 | 3 | 2 | 3 | 6/10 |
| Total | 7,770 | 29 | 7 (24%) | 15 (52%) | – |
Source: FlowProject measurement, DataForSEO SERP API, Google Hungary, desktop, 22 Oct 2026. Monthly searches are DataForSEO estimates. "n/a": an AI Overview appeared, but the API returned no source list.
One pattern stands out. For agency and provider searches (marketing agency, SEO agency, AI marketing agency, software development), all three AI sources came from the first organic page. For "build me one" searches, none did: the AI Overview for "weboldal készítés" (website design), for instance, cited honlapbirodalom.hu, wordpress-honlapkeszites.hu and virto.hu, none of which were on the first organic page.
Our reading (a hypothesis, not a proven link): a "build me a website" search hides lots of sub-questions – how much does it cost, which platform, how long does it take – and query fan-out looks for sources on those. A page that answers one sub-question well can get cited even if it doesn't rank near the top for the main term.
Where is Hungarian expert content missing?
For three keywords – "ai seo" (about 170 a month), "llms.txt" (about 140) and "seo audit" (about 90) – not one .hu domain appeared on the first organic page. International English-language sites filled it: semrush.com, moz.com, neilpatel.com, reddit.com, ahrefs.com and the like. The AI Overview sources were English too; for "llms.txt", the standard's own site (llmstxt.org) and developer.chrome.com. For "ai marketing", 2 of 8 first-page domains were Hungarian.
In practice, someone asking about these topics in Hungarian today usually gets an answer built on English sources. These are small keywords and won't bring a flood of traffic. But where there's no Hungarian expert content, a good, specific answer in Hungarian faces less competition. That's part of why we wrote this piece.
How is AI SEO different from classic SEO?
Google's position is clear. According to its guidance on AI features, there are no additional requirements to appear in AI Overviews or AI Mode, and no special optimization is needed. You don't need new machine-readable files, AI text files or special schema.org markup. The requirement is the same as before: the page must be indexed and eligible to appear in Search with a snippet. Search Console counts traffic from AI features within overall search traffic, under the "Web" search type, without a separate breakdown.
From Google's perspective, then, AI SEO is good SEO. In practice, a few things still differ:
- The unit is different. Classic SEO ranks pages. AI answers lift out passages: a paragraph, a definition, a number. Pages that make it easy to extract a clear statement win.
- The success metric is different. It's not just rankings and clicks, but whether you get mentioned. Search Console doesn't show that separately today.
- More engines matter. According to OpenAI, ChatGPT search relies on third-party search providers and content supplied by partners; in 2023 Microsoft announced that Bing would become ChatGPT's default search experience. OpenAI's crawler documentation says ChatGPT search surfaces sites via OAI-SearchBot, and sites that block it in
robots.txtwon't be shown in ChatGPT search answers (they can still appear as navigational links). Perplexity searches the web in real time and adds numbered source citations to every answer. A Google-only SEO strategy can easily miss all of this.
We covered the technical groundwork – speed, semantic HTML, structured data, crawlers – in a separate article.
What makes a page citable?
In May 2025 Google published a dedicated post on performing well in its AI search experiences. The key points: unique, non-commodity content that readers find helpful; important content available as text; structured data that matches the visible content; and the fact that nosnippet, max-snippet and similar controls also limit how your content can appear in AI features.
On top of that, here's what we see in our own work (experience, not a measured correlation):
- A direct answer under the question. A question-style heading, a 2–3 sentence answer that stands on its own, then the detail.
- Original data, with method and date. One number that can only be found on your site is worth more than ten rephrased generalities. The method makes it credible; the date shows how fresh it is.
- Local language, local context. Prices in local currency, local platforms, local regulation: an English-language page can't provide that for a Hungarian searcher.
- A consistent entity. Company name, address and description match everywhere: on the site, in structured data and on external profiles.
- Access for crawlers. Googlebot, Bingbot and OAI-SearchBot allowed in
robots.txt, and key content readable without JavaScript.
On llms.txt: Google explicitly says you don't need such a file for AI Overviews. What it's good for elsewhere, and how to write one, is covered in our llms.txt guide.
How do you track whether AI cites you?
The most important takeaway from our measurement isn't the 24%; it's the volatility. On our first request for "ai marketing ügynökség" (AI marketing agency), the AI Overview cited contented.hu, aimarketingugynokseg.hu and marketingai.hu. On a later request for the same term, it cited meraki.hu, marketingkalkulator.hu and seoklikk.hu. All three sources had changed.
So a manual spot check ("I searched and we're in it") proves nothing, and neither does the opposite. A reliable picture only emerges from repeated measurement with identical settings:
- Keyword list. 10–30 questions and phrases, worded the way your customers actually search.
- Regular requests with identical settings. Same country, language and device every time. A SERP API makes this automatic; manual searches are skewed by personalization.
- What to record. Whether there's an AI Overview, which domains it cites, whether you're among them, and your organic position.
- Trends, not snapshots. At least 6–8 weeks of data before drawing conclusions.
- ChatGPT and Perplexity. Ask the same questions regularly and note which pages they link as sources.
We do this weekly with our own Marketing.OS system: it pulls the results pages for tracked keywords, stores the AI Overview sources, and flags when we or a client get cited, or drop out. The data in this article comes from the first round of that measurement.
What to do now: a checklist
- Check in Search Console that your key pages are indexed and don't carry
nosnippetor an overly strictmax-snippet. - Check
robots.txt: can Googlebot, Bingbot and OAI-SearchBot get in? Register your site in Bing Webmaster Tools as well. - List the 10–20 questions your customers ask before they contact you.
- Put a 2–3 sentence direct answer at the top of key pages, under question-style headings.
- Add original data wherever you can: prices, turnaround times, measured results, with method and date.
- Make sure structured data matches what's visible on the page.
- Measure weekly with identical settings, and decide based on trends only.
Frequently asked questions
What is AI SEO?
AI SEO is the umbrella term for the work that gets a website found, understood and cited as a source by AI-powered search – Google's AI Overviews, ChatGPT and Perplexity. It builds on classic SEO fundamentals, but measures success by citations and mentions as well as rankings.
What's the difference between AI SEO and GEO?
In practice they're synonyms. GEO (generative engine optimization) comes from academic usage, AI SEO from industry shorthand. Both refer to getting into the answers of generative search engines.
Do I need special optimization for Google AI Overviews?
Google says no. There are no extra requirements, AI files or special markup; the page needs to be indexed and snippet-eligible, and regular SEO principles do the rest.
How do I get into ChatGPT's answers?
Allow OAI-SearchBot in robots.txt, make sure your site is well indexed in Bing too, and write content that makes it easy to extract a clear statement. There's no guarantee: answer sources vary from question to question and over time.
Is llms.txt required?
No. Google explicitly doesn't ask for such a file for AI Overviews. It can be a cheap addition, but it doesn't replace good, indexable content.
Can I see AI Overview traffic in Search Console?
Not separately. Google counts clicks from AI features under the "Web" search type, together with other organic traffic. If you want to know whether you're being cited, you have to measure it yourself.
How long until I see results?
There's no reliable general figure, and anyone promising a specific timeline isn't working from your data. Because sources change week to week, the effect only shows up across several weeks of consistently measured data.
If you'd like to know whether your keywords trigger AI Overviews, who gets cited, and what's missing for you to be cited too, request a free SEO and AI visibility audit. We work from your keywords, not a template.