Back to overview
    Studies & benchmarks

    From Schema Markup to AI Recommendation: What Most Websites Are Still Missing

    GEO Expert9. November 2026Updated 10. November 20269 min read
    From Schema Markup to AI Recommendation: What Most Websites Are Still Missing

    Nearly every online shop today has structured data on board: the shop system delivers it automatically, the SEO plugin is active, and the validator shows green. Yet many of these shops remain invisible as soon as customers ask ChatGPT, Perplexity, or Gemini for recommendations. Our analysis of 5,733 websites reveals some reasons why: while the technical shell is present almost everywhere, the signals that AI systems use to establish identity and trust are often missing. The average audit score is 59.9% – only 2.8% of websites reach 80% or more, while 42.6% remain below 60%. What separates the majority from the top group is not expensive relaunches, but a few targeted additions to existing markup. In this post, we explain the three most common gaps and show specifically how to close them.

    Why AI Systems Read Differently Than Google

    To understand the gaps, it helps to look at how an AI assistant compiles a recommendation. If someone asks, "Where can I buy good running shoes online?", the system in the background must clarify several questions – including whether a shop is reputable enough for a recommendation:

    1. Who is this provider? Do the website, review profile, and company listing belong to the same business?
    2. What exactly is being offered? Which product, at what price, is it in stock?
    3. Can the provider be trusted? Are there verifiable facts and reviews?

    A human answers these questions by clicking through the website. An AI system, on the other hand, prefers to rely on information it can uniquely assign. If this uniqueness is missing, the answer becomes uncertain – and uncertain sources are mentioned less frequently. Structured data is therefore not an end in itself, but the language in which you answer these three questions in a machine-readable way.

    The Paradox: The Technology is There, the Content is Missing

    71% of the audited websites fully meet our check for structured data. A completely different picture emerges as soon as it comes to content:

    What helps AI systemsFully met
    Structured data generally present71.1%
    About Us page with verifiable facts24.3%
    Company linked to official profiles (sameAs)19.1%
    Reviews readable as structured data5.4%

    The reason is obvious: shop systems generate the basic framework automatically. However, no plugin can know which profiles belong to your company, what makes you stand out, and where your reviews are located. You must add this information yourself.

    Gap 1: Corporate Identity Remains Unclear

    Why it matters

    AI systems work with so-called entities: uniquely identified companies, people, and products. For such a system, your shop is initially just a domain. Whether "mustershop.de", the review profile "Mustershop GmbH", and the Google Business Profile refer to the same company is something it has to deduce itself – and with similar names, this assignment can fail.

    With the sameAs property, you answer this question directly: "These profiles belong to us." This allows reviews, mentions, and company data from other locations to be assigned to your company.

    What the data shows

    Only 19% of websites fully link their company to official profiles. Even among websites with product offerings, the figure is only 22%.

    How to implement it

    Add the list of your real profiles to the Organization markup on your homepage:

    {
      "@context": "https://schema.org",
      "@type": "Organization",
      "name": "Mustershop GmbH",
      "url": "https://www.mustershop.de",
      "logo": "https://www.mustershop.de/logo.png",
      "sameAs": [
        "https://www.trustedshops.de/bewertung/info_XXXXXXXX.html",
        "https://www.linkedin.com/company/mustershop",
        "https://www.instagram.com/mustershop"
      ]
    }
    

    Best Practice:

    • Quality over quantity. Only link profiles that are actively maintained and clearly belong to you. An orphaned account from 2018 does not help.
    • Review profiles first. A profile with real customer reviews is a strong proof of trust because it exists independently of your website.
    • Keep names consistent. Use the same company name in the markup as in the legal notice (Impressum) and on the profiles.

    Gap 2: Product Data is Incomplete

    Why it matters

    When AI assistants compare products or give purchase recommendations, they need reliable facts: price, currency, availability. Without this information, the system risks passing on outdated or incorrect information – and prefers to rely on sources where the data is complete.

    Product identifiers such as GTIN/EAN and the brand play a special role. They are the unique key by which a product is recognized across different websites. Only then can test reports, price comparisons, and reviews be assigned to your offer without a doubt.

    What the data shows

    Among websites with product markup, parts of the offer data are missing for a good one in five (21%). Product identifiers and brand are not fully specified on as many as 41% of sites.

    How to implement it

    A complete product markup on the product detail page looks like this:

    {
      "@context": "https://schema.org",
      "@type": "Product",
      "name": "Aero 3 Running Shoe",
      "gtin13": "4012345678901",
      "brand": { "@type": "Brand", "name": "Musterbrand" },
      "offers": {
        "@type": "Offer",
        "price": "89.95",
        "priceCurrency": "EUR",
        "availability": "https://schema.org/InStock",
        "url": "https://www.mustershop.de/aero-3"
      }
    }
    

    Best Practice:

    • Check a real product page. Open the source code (not the rendered page) and search for gtin, price, and availability. What is missing there, many AI crawlers will not see.
    • Match visible content and markup. Price and availability in the markup must match the visible offer. Discrepancies do more harm than good.
    • Take identifiers from the ERP system. The GTIN is usually available anyway – it just isn't being output to the shop template.

    Gap 3: The About Us Page Says Little Concrete

    Why it matters

    To answer the question "Is this provider trustworthy?", AI systems look for verifiable facts. Sentences like "We are your partner for quality and passion" contain no such facts – they can neither be confirmed nor refuted. Year of foundation, company headquarters, responsible persons, or specialization, on the other hand, can.

    What the data shows

    Only 24% of websites fully meet our About Us page check. Often the page is missing entirely, or it consists primarily of general advertising claims.

    How to implement it

    A before-and-after example:

    Before: "For us, the customer is the focus. We offer the best quality at fair prices."

    After: "Mustershop GmbH was founded in 2011 in Freiburg and specializes in running equipment. Our team of twelve employees – including three certified running shoe consultants – handles over 40,000 orders annually. The managing director is Anna Muster."

    Best Practice – this information belongs on the page:

    • Full company name and headquarters
    • Year of foundation and brief development history
    • Specialization: What are you the right address for?
    • Responsible persons or the team
    • Verifiable evidence such as certificates, awards, or memberships

    Your Three-Step Roadmap

    1. Link Identity (approx. 30 minutes). Add sameAs links to your real profiles in the Organization markup of the homepage.
    2. Audit a Product Page (approx. 1 hour). Check price, currency, availability, GTIN, and brand in the source code and add missing fields in the template – the change will then affect all products.
    3. Rewrite About Us Page (approx. 2 hours). Replace advertising claims with the facts mentioned above.

    None of these steps require a relaunch. All three start where most websites in our benchmark are wasting potential.

    Where Does Your Website Stand?

    The average is a guide, not a goal. Which of the three gaps is open for you and where you can gain the most will be shown by your own evaluation.

    A better audit score means better technical requirements. No one can guarantee a mention in AI answers.


    Methodology: Data as of 11/09/2026. For each website, the latest complete run since the introduction of the current test catalog (06/25/2026) was evaluated, totaling 5,733 websites; websites audited multiple times count once, aborted runs are excluded. Product rates refer to the 2,109 websites with recognized product markup. These are voluntarily audited websites, not a representative market sample. We deliberately do not show a timeline, as the composition of the audited websites has changed over the months.

    Put this knowledge to work for your website

    Check the technical foundations of your AI visibility for free and get concrete next steps.

    „The AI Trust Audit is a huge help for us: it shows crystal clear what LLMs and AI search engines pay attention to. A clear recommendation for every online retailer who wants to become and stay visible in AI recommendations!“
    Gerd Riekers · Owner · porzellan-exklusiv.de