A bad Google thumbnail can make a good page look cheap.

You publish a useful service page or article, then Google displays your logo, a tiny staff headshot, or a decorative stock photo beside it. The image may be technically valid, but it doesn’t help a searcher understand the page or trust the result.

Small businesses have often treated this as Google’s choice and moved on. That is no longer a good excuse. In March 2026, Google added explicit guidance explaining that it uses both schema.org metadata and the og:image tag as sources when selecting thumbnails for Search and Discover. The change is recorded in Google Search Central’s documentation update log.

You still can’t force Google to show a particular image. You can, however, send a much clearer signal. This guide shows you how to do it without slowing your site, breaking social previews, or turning a straightforward SEO task into a development project.

What Google actually changed

Google did not announce a new ranking factor. It clarified how its automated systems choose a preview image for a page.

According to Google’s image SEO best practices, you can identify a preferred image in two main ways:

  • Add primaryImageOfPage to WebPage schema, or add an image property to the page’s main entity, such as a BlogPosting.
  • Add an og:image meta tag in the page’s HTML head.

Google may use these sources when selecting a thumbnail for a standard text result or a Discover preview. The final selection remains automated and considers multiple sources, so metadata is an influence, not a command.

That distinction matters. If your agency promises that one line of code will guarantee a certain thumbnail, the promise goes beyond Google’s documentation. The useful goal is consistency: give Google one clear, crawlable, representative image everywhere it looks.

Why a small business should care

Search results are visual inventory. A useful thumbnail can quickly show the finished kitchen, repaired roof, machined part, patient outcome, or person behind the service. A generic logo tells the searcher far less.

This also matters beyond a normal results page. Google says Discover can surface indexed content related to a person’s interests, and content needs no special tags or structured data merely to be eligible. Appearance is not guaranteed. Google’s Discover guidance recommends large, relevant, high-quality images because they are more likely to generate visits from Discover.

Don’t build a traffic forecast around Discover. Google describes Discover traffic as less predictable than keyword-driven search and recommends treating it as supplemental. The practical business case is simpler: you already paid to create the page. Give that page a fair chance to look credible wherever Google previews it.

Choose the right preferred image

Start with the page’s job, not the brand guidelines.

A roofing service page should show a roof, crew, inspection, or completed job. A bookkeeping article about cash-flow forecasting should use a real chart, worksheet, or owner reviewing numbers. An orthodontist’s treatment page could show the treatment setting or equipment, subject to patient privacy rules. The chosen image should help someone predict what they will get after clicking.

Google’s preferred-image guidance recommends an image that is relevant and representative of the page. It also says to avoid generic images such as a site logo, images with extreme aspect ratios, and images containing text. High resolution is preferred. Those recommendations appear in Google’s preferred-image section.

Use this five-question test:

  1. Would the image still make sense if the headline disappeared?
  2. Is it specific to this page rather than reused across 20 pages?
  3. Can the subject survive a tight mobile crop?
  4. Is the important subject near the center, with breathing room around it?
  5. Do you own the image or have a license that covers website and promotional use?

If the answer to any of the first four is no, pick another image. For licensing, check the actual agreement rather than assuming an image found online is available to use.

Prepare one strong image file

For Discover, Google recommends an image at least 1,200 pixels wide, more than 300,000 total pixels, and close to a 16:9 aspect ratio. It also says large previews should be enabled with max-image-preview:large or AMP. Those specifications are in Google’s Discover image recommendations.

A 1,280 by 720 image is a practical starting point. It meets the width, pixel-count, and aspect-ratio guidance while fitting common content-management workflows. Export it as WebP, AVIF, JPEG, or PNG based on what your system handles reliably. Google lists all four among its supported image formats, along with BMP, GIF, SVG, and others.

Compress the file before upload. Don’t take a 6 MB camera image and ask every visitor to download it just because it looks sharp. Google’s image guidance says images are often the largest contributor to overall page size and recommends current optimization and responsive-image techniques. You can confirm the result with PageSpeed Insights.

Give the file a short descriptive name, such as commercial-roof-inspection.webp, instead of IMG_4827.webp. Google says filenames provide light clues about subject matter and advises placing images near relevant text on a relevant page. Its image SEO documentation also identifies alt text as an important source of image metadata and an accessibility aid.

Write alt text for the person who cannot see the image. “Technician inspecting seams on a white commercial roof” is useful. “Roofing company roofer roof repair best roofer Ohio” is keyword stuffing, not a description.

Add the image metadata

Most small business sites should use both Open Graph and schema markup. This keeps social sharing and search metadata aligned and avoids creating two competing preferences.

Option 1: Add og:image

Place this in the HTML <head> and use an absolute URL:

<meta property="og:image"
      content="https://example.com/images/commercial-roof-inspection.webp">

The Open Graph protocol defines og:image as the image URL representing the object and treats the core URL, title, type, and image properties as basic metadata. You can review the syntax at the official Open Graph protocol site.

WordPress SEO plugins often generate this field from a page’s social image or featured image. Squarespace, Wix, Webflow, Shopify, and other hosted systems also expose social-sharing images in their page or collection settings. The label varies, so inspect the rendered HTML after saving instead of trusting the editor screen.

Option 2: Add schema.org markup

For a service or general page, use primaryImageOfPage on a WebPage object:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "WebPage",
  "url": "https://example.com/commercial-roof-inspection/",
  "primaryImageOfPage": "https://example.com/images/commercial-roof-inspection.webp"
}
</script>

For an article, connect the image to the BlogPosting:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "BlogPosting",
  "mainEntityOfPage": "https://example.com/roof-inspection-checklist/",
  "headline": "Commercial Roof Inspection Checklist",
  "image": "https://example.com/images/commercial-roof-inspection.webp"
}
</script>

Google supports Article, NewsArticle, and BlogPosting objects for article markup and recommends adding the properties that apply to the page. Its Article structured-data guide also advises validating markup with the Rich Results Test and checking deployed pages through URL Inspection.

Your og:image, schema image, visible hero image, and social-sharing setting should point to the same asset unless you have a measured reason to separate them. Alignment makes maintenance easier. It also prevents the common mess where Facebook shows one image, a schema validator sees another, and the page displays a third.

Allow a large preview

Check the page’s robots meta tag. For the large-image treatment Google recommends for Discover, include:

<meta name="robots" content="max-image-preview:large">

Google defines max-image-preview as a control for the maximum size of an image preview in search results. The large value permits a preview up to the image viewport’s width. The full behavior and other values are documented in Google’s robots meta tag reference.

Before adding a new tag, inspect what your site already outputs. Two plugins or templates can create competing robots directives. Consolidate them into one valid instruction, and make sure the page is not accidentally marked noindex.

Test the complete page, not just the code

Publishing valid markup is necessary, but it is not the finish line. Run a real page through this sequence:

  • Open the image URL in a private browser window. It should load without a login, hotlink block, or 403 response.
  • View the page source and confirm the exact absolute image URL appears in og:image and the relevant schema object.
  • Run the URL through Google’s Rich Results Test, then inspect the live URL in Search Console. Google’s structured-data workflow recommends both checks before requesting a recrawl.
  • Confirm the visible page includes the image in a standard HTML <img> element. Google says it discovers images from an <img> element’s src attribute but does not index CSS background images, as explained in its image discovery guidance.
  • Test the mobile crop and page speed. A technically correct 1,200-pixel image still fails if the subject is cut off or the file makes the page painful to load.

After deployment, request indexing for a few important URLs rather than changing hundreds of pages blind. Watch the rendered result, then roll the pattern into your templates.

A practical rollout plan

Don’t begin with every image in the media library. Start with the 10 pages that produce the most revenue or qualified leads. That usually means core service pages, strong comparison pages, high-intent articles, and recent case studies.

Assign one preferred image to each page. Standardize the dimensions and file-naming pattern. Update the page template so editors choose the preferred image once and the system outputs it to the visible page, Open Graph metadata, and schema automatically.

Then document ownership. Marketing should select the asset and write the alt text. A developer or site manager should control template output, compression, crawl access, and validation. One person should perform a quarterly spot check because redesigns, plugin updates, CDN migrations, and social-image changes can quietly break the pattern.

This isn’t glamorous SEO work. It is the kind that compounds because every future page inherits the fix.

What not to expect

Preferred-image metadata won’t improve weak content, guarantee Discover traffic, or force a thumbnail into every search result. Google explicitly says image selection is automated, and Discover eligibility does not guarantee appearance. It also warns against misleading previews, clickbait, and sensationalism in its Discover content guidance.

Treat the image as packaging for a useful page. The page still needs to answer the question, show credible proof, load quickly, and make the next step obvious.

If your top pages show the wrong images, missing metadata, or inconsistent schema, get started with Your Web Team. We’ll audit the template, fix the image signals, and make sure the change works across the site instead of patching pages one at a time.