On March 2, 2026, Google added a section on specifying a preferred image with metadata to its image SEO best practices and its Discover documentation. If Google shows the wrong image beside one of your pages, the first fix to try is to name one image the same way in three places.

Google says it uses both schema.org markup and the og:image meta tag as sources when it determines thumbnails in Google Search and Discover. Google still selects images automatically, so this metadata can influence the choice without guaranteeing it.

Start by choosing one image that represents the page, then use its absolute URL in primaryImageOfPage, in the image property of the page's main entity, and in og:image.

Key takeaways

  1. Name one image three timesUse the same absolute image URL in primaryImageOfPage, in the main entity's image property and in the og:image tag.
  2. Google reads bothGoogle says it uses both schema.org markup and the og:image tag when it determines thumbnails in Search and Discover.
  3. Influence, not controlImage selection stays automated, so this metadata can influence the choice without guaranteeing a thumbnail, ranking or Discover appearance.
  4. Choose a representative imageAvoid a site logo or text-heavy artwork. Pick a relevant, high-resolution image that appears in a standard <img> element, not only as a CSS background.
  5. Keep the file reachableConfirm crawlers can request the image, with no robots.txt block, login wall or noindex header. Allow time for Google to recrawl before judging the result.

What Google changed on March 2, 2026

Google's documentation updates log lists the March 2 change as "Added preferred image best practices." The entry says the section on specifying a preferred image with metadata went into two places: the image SEO best practices and the Discover documentation. The image SEO page is marked as last updated on March 2, 2026.

The log entry explains the reason in Google's own words: "Based on feedback, we're clarifying that Google uses both schema.org markup and the og:image meta tag as sources when determining image thumbnails in Google Search and Discover."

That wording is a clarification of what Google reads, and it names two surfaces, Google Search and Discover. Google's image documentation describes image selection as automated, which is why the guidance talks about influence. The update leaves one thing in your hands: making sure the page points at the image you would pick yourself.

A man with a man bun and a floral shirt photographs a plated dish at a kitchen counter with a laptop open beside him
Figure 1. Pick the photo that best represents the page, then name it the same way everywhere.

What Google documented on March 2, 2026

  1. March 2, 2026Google's documentation update log lists "Added preferred image best practices."
  2. March 2, 2026Google says it added a preferred-image section to its image SEO best practices and Discover documentation.
  3. March 2, 2026Google says schema.org markup and the og:image tag are both sources it uses when determining thumbnails in Search and Discover.
  4. March 2, 2026The Google image SEO page is marked "Last updated 2026-03-02 UTC."

Three places to name the preferred image

Google's guidance covers three metadata paths. The first is primaryImageOfPage on a WebPage, the schema.org property that indicates the main image on a page. The second is image on the page's main entity, such as a BlogPosting, connected to the page with mainEntity or mainEntityOfPage. The third is the Open Graph og:image meta tag in the page's <head>.

Google's documentation shows each as a short example. The WebPage version carries a primaryImageOfPage URL, the BlogPosting version carries mainEntityOfPage and an image URL, and the HTML version is a single tag:

html
<meta property="og:image" content="https://example.com/images/cat.png">

You do not need to treat the three as competing options. They describe the same fact from different angles, and the point of the guidance is that they agree. Pages that name one image in og:image and a different one in image give Google two answers to the same question.

For this guidance, primaryImageOfPage is optional. It is a valid schema.org property for WebPage, and Google documents it as one way to specify a preferred image, but it is not a separately documented Google rich result requirement. The same goes for og:image, which is optional for Google's purposes and useful to every consumer that reads Open Graph.

Before and after: one image URL in all three places

This first example describes an article but names no image anywhere. Replace the example.com URLs and the author details with your own page's facts when you adapt it.

json
{
  "@context": "https://schema.org",
  "@type": "BlogPosting",
  "@id": "https://example.com/articles/blue-garden#article",
  "mainEntityOfPage": "https://example.com/articles/blue-garden",
  "headline": "How to Grow Blue Garden Herbs",
  "datePublished": "2026-02-20",
  "author": {
    "@type": "Person",
    "name": "Taylor Example",
    "url": "https://example.com/authors/taylor-example"
  }
}

The next version puts one image URL into the WebPage entity and the BlogPosting entity, and the two entities point at each other with @id references. Use your own absolute image URL in both places, and give the image its true dimensions or leave the dimension fields out.

json
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "WebPage",
      "@id": "https://example.com/articles/blue-garden#webpage",
      "url": "https://example.com/articles/blue-garden",
      "name": "How to Grow Blue Garden Herbs",
      "primaryImageOfPage": {
        "@type": "ImageObject",
        "url": "https://example.com/images/blue-garden-herbs.jpg",
        "contentUrl": "https://example.com/images/blue-garden-herbs.jpg",
        "width": 1600,
        "height": 900
      },
      "mainEntity": {
        "@id": "https://example.com/articles/blue-garden#article"
      }
    },
    {
      "@type": "BlogPosting",
      "@id": "https://example.com/articles/blue-garden#article",
      "mainEntityOfPage": {
        "@id": "https://example.com/articles/blue-garden#webpage"
      },
      "headline": "How to Grow Blue Garden Herbs",
      "image": "https://example.com/images/blue-garden-herbs.jpg",
      "datePublished": "2026-02-20",
      "author": {
        "@type": "Person",
        "name": "Taylor Example",
        "url": "https://example.com/authors/taylor-example"
      },
      "publisher": {
        "@type": "Organization",
        "name": "Example Garden Journal",
        "url": "https://example.com"
      }
    }
  ]
}

The meta tags go in the page's <head>, and the <img> element goes in the page body where visitors see it. Both use the same URL as the JSON-LD above.

html
<meta property="og:image" content="https://example.com/images/blue-garden-herbs.jpg">
<meta property="og:image:alt" content="Blue garden herbs growing in a raised bed">

<img
  src="https://example.com/images/blue-garden-herbs.jpg"
  width="1600"
  height="900"
  alt="Blue garden herbs growing in a raised bed">

The Open Graph protocol lists og:image as a required basic property for an Open Graph object. Its structured image properties, such as og:image:width and og:image:type, are optional, and og:image:alt is recommended when an image is specified.

For more complete JSON-LD to copy, see 25 JSON-LD schema examples you can copy. To place the code on a site, how to add schema markup to a website covers the routes. If you want the BlogPosting block generated from your own page details, the Article schema generator builds it for you.

Before and after: one image URL everywhere

Hover over or tap a row to highlight it. On small screens the table scrolls sideways.

SignalBeforeAfter
Article imageThe article markup has no image property.The article markup includes a representative image.
Page imageThe WebPage does not identify a main image.primaryImageOfPage identifies the main image.
Open GraphNo og:image tag exists.og:image uses the same absolute URL.
Visible imageThe image may exist only as a CSS background.The image is available through an HTML img src.
Image URLsMetadata may point to different files.Every signal points to the same image URL.

What the image itself needs

Google's image guidance asks for a relevant, representative image. A site logo is a generic choice that Google advises against, and so is artwork heavy with text, because cropping a thumbnail can make the words unreadable. Extremely narrow or extremely wide images are also poor candidates. Use high resolution when you can.

Google lists these formats for images referenced in an img src attribute: BMP, GIF, JPEG, PNG, WebP, SVG and AVIF. Google says it can find images in an <img src> attribute and does not index CSS background images. So, the preferred image should appear on the page in a standard <img> element, and what the markup describes should match what a visitor sees there.

The image URL also has to work for crawlers. Google's general structured data guidelines ask that markup be crawlable, indexable and consistent with visible content, and the same reasoning applies to the file the markup points at. A typo, an expired CDN link or a long redirect chain leaves every signal pointing at nothing.

Use one stable, absolute URL, with the scheme and domain included. Relative paths and addresses that change when you rebuild the site are the easiest way to end up with three signals that no longer match.

An illustrated crawler shape reaches a picture tile through an unlocked gate while a padlock and a detour loop sit to the side
Figure 2. A crawler has to be able to request the image file for the markup to point at anything.

Why Google can still choose another image

Even with all three signals aligned, Google's image selection stays automated. The metadata is a source Google reads, and Google's systems still make the final pick. Google does not promise a particular thumbnail, ranking, rich result or Discover appearance from this markup.

Discover has its own guidance. The March 2 update log names it, and Google's Discover documentation describes its own image recommendations, so read that page for current Discover requirements before you plan around a specific image size.

If the wrong image is still showing, work through the likely causes in order: different URLs across the three signals, an image blocked from crawlers, an image that only exists as a CSS background, or a logo or text-heavy graphic used in place of the page's main image. Fix those first, then wait for Google to recrawl the page and reprocess it. Allow for that delay before you judge the result.

Make the image work for Bing, Apple and Yandex

One consistent image across schema.org and Open Graph serves every reader of this markup at once. Bing supports Schema.org, JSON-LD and Open Graph, and its structured data documentation says Open Graph image metadata may be used by Bing News to display images in news results. Apple uses og:image for rich link previews in Messages; Ensuring Beautiful Rich Links recommends a page-specific image with no text in it, and an icon as a fallback.

Yandex's structured data validator checks Schema.org and Open Graph, and its graphics guidance stresses crawlable image files, meaningful filenames, nearby text and useful alt text. Do not promise a particular image in any AI answer. The guide on whether AI systems use schema markup has the full comparison.

Check it and allow time to recrawl

Work through the page in this order. Choose the image that best represents the page and place it in the visible content with an <img> element. Use one absolute URL, then add that URL to the main entity's image, to primaryImageOfPage when the page has a distinct main image, and to og:image.

Next, confirm that Googlebot, Bingbot, Applebot and the social crawlers you care about can request the image. Look for a robots.txt block, a login wall, a firewall rule, a hotlink restriction or a noindex response header on the image file. Validate the JSON-LD, remembering that comments, trailing commas and duplicate keys make it invalid, and open the rendered HTML to see the tags the browser receives.

Then wait. Recrawling and reprocessing take time, and the markup influences selection without guaranteeing a specific thumbnail, ranking, rich result, Discover appearance or AI citation.

Questions people ask

What did Google change on March 2, 2026?

Google added a section on specifying a preferred image with metadata to its image SEO best practices and Discover documentation. The update log says Google uses both schema.org markup and the og:image meta tag as sources for thumbnails in Search and Discover.

What is the difference between image and primaryImageOfPage?

image describes an image associated with an entity such as a BlogPosting. primaryImageOfPage sits on the WebPage and indicates the main image on the page. Use the same URL in both when they refer to the same picture.

Can I use my logo as the preferred image?

Google's guidance advises against generic images such as a site logo. Choose an image that represents what the page is about, and use the logo for brand markup instead.

Does Bing use og:image?

Bing says it supports Open Graph, and its documentation says Open Graph image metadata may be used by Bing News in news results. Bing also says annotations do not guarantee a visually rich result.

Quick check: signal or guarantee?

Choose an answer to see feedback. Nothing is scored or saved.

Question 1 of 3

Which statement matches Google's documentation?

Question 2 of 3

Which image is the better preferred image for a recipe page?

Question 3 of 3

Where should the preferred image URL appear?

Terms used in this update

Select a term to see its definition.

primaryImageOfPage. A schema.org property indicating the main image on a WebPage.