SGE and AI Overviews: How AI Search Results Are Changing SEO

1. What AI Overviews and SGE are and how they generate answers

AI Overviews are AI-generated summaries that Google has been displaying above the classic organic results since the introduction of the Search Generative Experience (SGE). Instead of ten blue links, Google delivers a synthesized answer directly, based on a Gemini model that combines information from multiple sources at once. Technically, this is a retrieval-augmented generation system: the model doesn't draw freely on its training knowledge, but first searches the existing Google search index for relevant pages and only then generates a piece of running text that combines statements from several of these pages.

This architecture has an important consequence for SEO: a page still has to satisfy classic ranking signals to even become a candidate for text generation in the first place. AI Overviews therefore don't replace the existing ranking, they add an extra processing stage on top of it. Only once a page is classified as relevant in the retrieval step does the language model decide whether and how its content flows into the generated answer, and whether the page gets linked as a source.

2. AI Overviews compared to classic Featured Snippets

A classic Featured Snippet extracts a single, mostly verbatim passage from exactly one source and displays it in a box above the organic results. The selection follows a comparatively simple extraction logic: Google identifies the paragraph that most directly answers the search query and quotes it almost unchanged. An AI Overview works fundamentally differently: it generates an entirely new, paraphrased text that processes facts and phrasing from multiple sources at once and merges them into a coherent answer.

This also changes the visibility of the source itself. While a Featured Snippet shows the origin page prominently with title and URL, an AI Overview often lists sources only as a small, expandable link list at the edge or below the generated answer. AI Overviews also frequently show dynamically generated follow-up questions, which lead to further, likewise AI-generated answers without the user ever leaving the search results page. For website operators, this means: even a successful citation guarantees no comparable click-through rate to a former Featured Snippet.

3. Impact on organic click-through rate and zero-click search

Several independent studies from SEO analytics firms show a noticeable decline in organic click-through rate for search queries where an AI Overview is displayed, especially for information-oriented queries like definitions, how-tos, or comparisons. Users already get their answer on the search results page and simply have no reason to click through anymore. This effect further accelerates the so-called zero-click search, a trend that had already begun with classic Featured Snippets and knowledge panels and is now clearly amplified by generative summaries.

For highly transactional and commercial search queries, such as concrete product searches with purchase intent, the effect has so far been more moderate, since Google tends to show shopping results and ads here instead of pure text summaries. For Magento stores, this means a shift in how success is measured: pure click and session numbers from Google Analytics become less reliable as a sole KPI. A combined view, made up of impressions in Search Console, brand awareness, and whether your own domain even appears as a cited source in relevant AI Overviews, makes more sense.

4. What makes content citable for AI summaries

AI systems favor content that answers a search intent precisely and without detours. A paragraph that delivers the core statement in the first sentence or two, instead of only formulating it after a long introduction, gets extracted far more reliably than nested, marketing-heavy text. Factual specificity is also decisive: concrete numbers, data, named entities, and verifiable statements are easier for a language model to classify as solid facts than vague phrasing like "many experts recommend".

A second important factor is originality. Content that provides its own data, its own tests, or its own analysis, instead of paraphrasing content that already exists on other pages, is treated by generative systems as a more valuable source, because it feeds unique information into the model that isn't available anywhere else. Clear structure supports this effect further: headings that cover exactly one sub-question, short paragraphs, bullet lists, and tables are easier for a model to extract in isolation than long, unstructured blocks of running text without a recognizable outline.

<!-- Citable answer block: direct answer first, then supporting evidence -->
<section class="answer-block" itemscope itemtype="https://schema.org/Answer">
  <h3 id="what-is-lcp">What is a good LCP value?</h3>

  <!-- Direct, self-contained answer in the first sentences -->
  <p class="answer-block__summary">
    A good LCP value is under 2.5 seconds. This threshold comes
    directly from Google and is measured via the Chrome User Experience
    Report (CrUX) using real user data.
  </p>

  <!-- Supporting details in extractable list form -->
  <ul class="answer-block__facts">
    <li>Good: under 2.5 seconds</li>
    <li>Needs improvement: 2.5 to 4.0 seconds</li>
    <li>Poor: over 4.0 seconds</li>
  </ul>
</section>

5. Structured data and schema as a signal for source selection

Structured data following schema.org doesn't guarantee inclusion in an AI Overview, but it significantly reduces a page's ambiguity and makes it easier for search systems to capture facts in a machine-readable way instead of purely as text. FAQPage and HowTo markup structure question-answer pairs and step-by-step instructions in a format that is almost identical to what a generative model would otherwise have to painstakingly extract from running text. Product, Review, and Organization schema additionally deliver verifiable facts like prices, ratings, and company data that can flow into a generated answer as trustworthy data points.

In addition, author and provenance markup is gaining importance: author fields with Person schema, references to professional qualifications, and consistent sameAs links to author profiles provide signals that feed into the quality guidelines Google has publicly described around expertise, experience, authoritativeness, and trustworthiness. The basic rule from classic SEO still applies here: structured data must exactly reflect what is visibly present on the page for humans, otherwise you risk not only losing rich-snippet eligibility but also lower trustworthiness as a citation source.

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "What does a Magento relaunch with Hyvä Theme cost?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "A Magento relaunch with Hyvä Theme costs between 15,000 and 60,000 euros depending on scope. The decisive factors are the number of custom modules, the extent of design customization, and data migration."
      }
    },
    {
      "@type": "Question",
      "name": "How long does a Hyvä Theme relaunch take?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "A typical Hyvä relaunch takes 8 to 14 weeks, depending on catalog size, number of integrations, and testing phases before go-live."
      }
    }
  ]
}

6. Technical and content patterns for better inclusion chances

Since AI Overviews build on the regular Google search index, all technical SEO fundamentals still apply unchanged as a precondition: a page must be crawlable, indexable, and reachable within a reasonable load time before it can even become a retrieval candidate. Pages that act as a clear "single source of truth" on a topic, rather than spreading the same content across multiple thin, overlapping pages, also benefit. Canonical, consolidated content pages are easier for a language model to classify as a reliable single source than fragmented duplicate content.

Access by AI crawlers themselves also plays a role: Google uses the same Googlebot for SGE as for regular indexing, so no separate opt-in is needed. However, anyone who also wants to be cited in Perplexity, ChatGPT Search, or other AI search systems should specifically check whether AI-specific crawlers like GPTBot, Google-Extended, or PerplexityBot are allowed via robots.txt. An additional, optional llms.txt document in the root directory can also offer AI systems a curated overview of the most important pages, even though its support is still inconsistent across the industry.

# robots.txt: steer AI crawlers deliberately instead of blocking them wholesale
User-agent: Googlebot
Allow: /

# Google-Extended controls usage for Gemini/SGE training data,
# NOT regular crawling for Google Search itself
User-agent: Google-Extended
Allow: /

User-agent: GPTBot
Allow: /

User-agent: PerplexityBot
Allow: /

Sitemap: https://mironsoft.de/sitemap.xml

# llms.txt in the root directory: a curated overview for AI systems
# https://mironsoft.de/llms.txt
# # Mironsoft
# > Magento 2 and Hyvä Theme agency for high-performance online stores.
#
# ## Key pages
# - [Services](https://mironsoft.de/services): Magento development, Hyvä migration, SEO
# - [Blog](https://mironsoft.de/blog): Technical guides on Magento, SEO, and performance

7. Measuring visibility in AI Overviews without direct ranking reporting

Google Search Console so far offers no complete, standalone report that shows exactly which search queries your own domain was cited for within an AI Overview. A first starting point is the "Search Appearance" filter in the performance analysis, which now partially shows impressions for AI Overview appearances separately, even though the depth of data remains limited. Noticeable discrepancies between high impressions and unusually low click-through rate for certain keyword clusters are usually a clear indicator that an AI Overview is regularly being shown for those queries.

In addition, specialized third-party tools have emerged that specifically track SERPs with AI Overview presence, log whether your own domain gets cited, and compare this to competitor visibility. For Magento stores, it's also worth taking a look at unusual referrer patterns in the server log or in analytics, such as noticeable traffic from AI chat interfaces, which can indicate a successful citation as a source. It's important not to view these signals in isolation, but together with classic ranking data and the organic click-through-rate trend over time.

// Search Console API: query impressions by search appearance
// to approximate AI Overview presence
const { google } = require('googleapis');

async function fetchAiOverviewSignals(siteUrl, startDate, endDate) {
  const searchconsole = google.searchconsole('v1');

  const response = await searchconsole.searchanalytics.query({
    siteUrl,
    requestBody: {
      startDate,
      endDate,
      dimensions: ['query', 'searchAppearance'],
      dimensionFilterGroups: [{
        filters: [{
          dimension: 'searchAppearance',
          operator: 'contains',
          expression: 'AI_OVERVIEW',
        }],
      }],
      rowLimit: 1000,
    },
  });

  // High impressions combined with very low CTR suggest
  // frequent AI Overview appearances for this query
  return response.data.rows ?? [];
}

8. The debate: threat or opportunity for shops and publishers

Publishers and content websites have reported double-digit declines in organic traffic since the broad rollout of AI Overviews, especially for information-oriented guide content. This development has triggered several licensing disputes and negotiations between major publishers and AI providers, centered on compensation for using content as training and answer material. For ad-funded publishers whose business model depends directly on page views, this effect is immediately business-critical.

For online stores, the situation looks more nuanced. Anyone cited as a source in an AI Overview gains brand awareness and trust, even if the immediate click doesn't happen, an effect that often shows up with a delay in the form of direct or brand-related search. In addition, users who still click through after already receiving an information answer often have more concrete purchase intent, because the pure information phase is already complete. The pragmatic position held by many SEO experts is therefore: AI Overviews shift value away from pure click volume and toward more qualified, purchase-ready visitors.

9. What stays the same: E-E-A-T, technical SEO, and content quality

Despite all the changes brought by generative search results, the selection of retrieval candidates still rests on the same foundations as classic SEO: technical reachability, clean indexability, topical relevance, and backlink authority. No language model generates an answer from a page that isn't even listed in the search index or that can't be crawled due to technical errors. The basic rule remains: without a solid technical foundation, there is no chance of visibility, neither in classic ranking nor in an AI-generated summary.

Google's E-E-A-T principles (Experience, Expertise, Authoritativeness, Trustworthiness) are also gaining rather than losing importance in the AI context, because generative systems are especially prone to spreading false or unsubstantiated statements and therefore tend to favor more trustworthy sources that are recognizably written by experts. Original, well-researched content with traceable authorship, genuine hands-on experience, and verifiable facts remains the most reliable strategy, regardless of whether a user ultimately reaches a page via a classic blue link or via an AI Overview.

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Mironsoft",
  "url": "https://mironsoft.de",
  "sameAs": [
    "https://www.linkedin.com/company/mironsoft",
    "https://github.com/mironsoft"
  ],
  "founder": {
    "@type": "Person",
    "name": "Miro Sonntag",
    "jobTitle": "Magento and Hyvä Theme Developer",
    "knowsAbout": ["Magento 2", "Hyvä Theme", "Technical SEO", "Web Performance"],
    "hasCredential": {
      "@type": "EducationalOccupationalCredential",
      "credentialCategory": "Adobe Commerce Certified Professional"
    }
  },
  "areaServed": "DE",
  "knowsAbout": ["Magento 2 Development", "Hyvä Theme", "E-Commerce SEO"]
}

The following overview summarizes the concrete differences between AI Overviews and classic Featured Snippets, and what consequence follows for optimization.

10. Summary

SGE and AI Overviews are fundamentally changing search, but they don't replace the foundations of SEO, they add an additional generation stage on top of the existing search index. AI Overviews synthesize answers from multiple sources instead of quoting a single paragraph, which noticeably reduces organic click-through rate especially for information-oriented queries and further accelerates zero-click search. Content becomes citable when it is precise, factually specific, clearly structured, and marked up with clean structured data.

Since there is no direct ranking report for AI Overviews, measurement remains a combination of Search Console signals, referrer analysis, and specialized third-party tools. The debate between threat to publishers and opportunity for brand-conscious shops isn't settled yet, but regardless of the outcome, technical SEO, E-E-A-T, and original content quality remain the reliable foundation on which any form of visibility, classic or AI-generated, is built in the first place.

11. FAQ: SGE and AI Overviews in SEO