Google recently updated its image SEO best practices and Discover guidelines to spell out exactly how it selects thumbnail images and how you can influence that selection. (Search Engine Land covered the changes here.)
This isn’t new functionality. Google has used schema markup and og:image for thumbnails for a long time. What’s new is that Google is being specific about the mechanics, particularly for Discover, where the thumbnail isn’t just helpful. It’s the entire first impression.
Most sites aren’t managing this deliberately. They’re letting their CMS pick a default og:image, skipping the relevant schema properties, and hoping for the best.
Why This Is a Discover Story
Google’s documentation says these methods apply to both standard search results and Discover. That’s true. But the practical impact is almost entirely on the Discover side.
In regular search, thumbnails are small. They don’t always appear. The title and snippet are still doing most of the persuasion. A suboptimal thumbnail in search results is a minor missed opportunity.
Discover is different. Discover is a feed. There’s no query. Users aren’t searching for anything. Google surfaces content based on interests and browsing behavior, and the result looks like a social media feed, not a search results page. The image isn’t supporting the content. It is the content, at least from a first-impression standpoint. A bad thumbnail in Discover tanks your CTR before anyone reads a word.
Does This Apply to You?
Discover traffic doesn’t flow to every type of site. It tends to favor publishers, news-adjacent content, blogs with strong topical authority, recipe sites, and lifestyle or travel content. If you’re running a local service business, a B2B SaaS site, or a thin product catalog, Discover probably isn’t a meaningful part of your traffic mix.
Check your Google Search Console under the Discover tab. If there’s data there, keep reading. If there’s nothing, the Discover-specific guidance below isn’t a priority for you, though the general thumbnail hygiene still applies.
How Google Selects Thumbnails
Google says its image selection is fully automated, pulling from multiple sources on the page. You can influence that decision through three methods:
Schema markup using primaryImageOfPage. This is probably the most direct signal. You attach it to the page-level schema, and it tells Google which image you want representing the page.
An image attached to the page’s main entity. If your schema uses mainEntity or mainEntityOfPage, and that entity has an image property, Google will consider it.
The og:image meta tag. Most sites already have this because it’s what Facebook, LinkedIn, and other platforms use for link previews. But “having it” and “having it set to the right image” are two different things.
Google considers all three. It’s not picking one method over the others.
What Google Says to Avoid
The documentation is specific:
Don’t use your site logo as the thumbnail image. This happens a lot on pages where the CMS defaults to the logo when no featured image is set.
Don’t use images with text baked in. Promotional banners, sale graphics, images with headlines overlaid. Google doesn’t want those as thumbnails.
Don’t use extreme aspect ratios. Too narrow, too wide, or oddly cropped.
Don’t use low-resolution images.
Discover’s Specific Image Requirements
For Discover, Google’s requirements are more precise. Images should be at least 1,200 pixels wide, high resolution with a minimum file size of 300K, and ideally in a 16:9 aspect ratio.
Google will auto-crop images for Discover if you don’t crop them yourself. That can go wrong fast, especially with vertical images where the important content sits at the top or bottom and gets cut in a landscape crop.
There’s also a technical requirement that trips people up: your robots meta tag needs to include max-image-preview:large. Without it, Google won’t use large images in Discover at all. A lot of sites don’t have this set.
What Most Sites Get Wrong
The issue isn’t that sites lack og:image tags or schema markup. Most CMS platforms handle that automatically. The issue is that nobody reviews what those defaults actually produce.
The og:image is auto-generated and never checked. WordPress will typically use the featured image. That’s fine if you chose the featured image with thumbnails in mind. Most people choose it based on how it looks on the page, which is a different context entirely.
Schema markup exists but doesn’t include primaryImageOfPage. This property is underused. Most sites running schema have Article or WebPage types but never specify which image should represent the page. Adding it is straightforward and gives Google a direct signal.
Featured images are chosen for the page, not the feed. An image that works as a hero banner on your article might look terrible cropped to 16:9 in a Discover card.
Nobody checks what the thumbnails actually look like. You can see what Google is pulling in Search Console, and you can test your og:image with social media preview tools. Most people never look.
What to Do Now
This is a short audit. Do it once per template type rather than page by page.
Check your og:image tags. Are they intentional, or whatever the CMS generated? Do they look good at thumbnail size?
Check your schema for primaryImageOfPage. If it’s not there, add it. One property, clear signal.
Check your robots meta tag for max-image-preview:large. If it’s missing, add it. This is table stakes for Discover.
Look at your images at 16:9. Take your most important pages and see what those images look like cropped to landscape. If the answer is bad, you need different images or deliberate cropping.
Check Discover in GSC. Look at which pages are showing up and what images Google is pulling. If the thumbnails are wrong, now you know how to fix it.
For sites where Discover is a real traffic channel, the difference between a good thumbnail and a bad one is measurable. Google just told you exactly how to control it.
