In short. Use a relevant, representative image in schema.org metadata or og:image, while treating Google's thumbnail selection as automated.
Definition: A publisher can suggest the image used as a page thumbnail in Google Discover by declaring a relevant, representative image in schema.org markup or in the
og:imagemeta tag. Google describes this as an influence on automated image selection, rather than a fixed selection rule.
Key facts
- Google says the feed can use either schema.org markup or
og:imageto specify a large image that is relevant and representative of the page. This can influence which image is chosen as the thumbnail. Google's Discover documentation gives this guidance. - Google's image SEO documentation says image selection is completely automated and can use metadata sources such as
primaryImageOfPage, an image on the main entity, orog:image. - The Open Graph protocol defines
og:imageas an image URL that represents the page object within the graph. - Schema.org describes ImageObject as an image file and defines
representativeOfPageas indicating whether an image represents the page's content.
Which schema.org image can a publisher provide?
Google's image documentation accepts a URL or an ImageObject for the schema.org primaryImageOfPage property. It also accepts an image URL or ImageObject attached to the main entity through mainEntity or mainEntityOfPage. The page supplies this WebPage example:
{
"@context": "https://schema.org",
"@type": "WebPage",
"url": "https://example.com/url",
"primaryImageOfPage": "https://example.com/images/cat.png"
}
For an article-like entity, Google's example uses BlogPosting, mainEntityOfPage, and an image URL:
{
"@context": "https://schema.org",
"@type": "BlogPosting",
"mainEntityOfPage": "https://example.com/url",
"image": "https://example.com/images/cat.png"
}
The Schema.org definition also provides contentUrl for the actual bytes of a media object and representativeOfPage for whether the image represents the page content.
What should the suggested image look like?
Google's image guidance says to choose an image relevant to and representative of the page, avoid a generic image such as a site logo, an image with text, or an extreme aspect ratio, and use high resolution when possible. The Discover page separately recommends a large, relevant image that is at least 1200 px wide, has more than 300,000 total pixels, uses a 16x9 aspect ratio, and is enabled by max-image-preview:large or AMP. Its example says that a 16:9 image measuring 1280x720 pixels has 921,600 total pixels.
Google tries to crop the image automatically for the feed. When a publisher crops an image, Google's guidance says the crop should be well positioned for wide orientation and retain important details in the version named by og:image.
How does og:image express the suggestion?
Google's image documentation gives this minimal form:
<meta property="og:image" content="https://example.com/images/cat.png">
The Open Graph protocol defines optional properties for the image URL, a secure URL, MIME type, pixel width, pixel height, and alternative text. Its full example is:
<meta property="og:image" content="http://example.com/ogp.jpg" />
<meta property="og:image:secure_url" content="https://secure.example.com/ogp.jpg" />
<meta property="og:image:type" content="image/jpeg" />
<meta property="og:image:width" content="400" />
<meta property="og:image:height" content="300" />
<meta property="og:image:alt" content="A shiny red apple with a bite taken out" />
The protocol says that when several og:image tags are present, the first tag from top to bottom receives preference during conflicts. It also says that structured properties follow their root og:image tag.
Related terms
FAQ
Does this markup determine the thumbnail?
No. Google's image documentation says image preview selection is completely automated. The metadata gives the publisher a documented way to influence the image considered for the preview.
Can a site logo serve as the suggested image?
Google's Discover guidance advises avoiding generic images such as a site logo in schema.org markup or the og:image tag. A relevant image that represents the page is the documented choice.
Is schema.org markup required for eligibility?
No. The same Discover documentation says indexed content that meets its content policies is automatically eligible, and that special tags or structured data are not required.
Does og:image replace the schema.org option?
The image documentation presents og:image and schema.org metadata as alternative sources for a preferred image. It does not publish a precedence rule between those two sources.
Sources
- Google Search Central, "Get on Discover", 9 March 2026
- Google Search Central, "Image SEO Best Practices", 2 March 2026
- Open Graph protocol, "The Open Graph protocol", undated
- Schema.org, "ImageObject", undated
Check how a page reads to answer engines. Open the GEO score analyzer