Modern Image SEO Using Schema Structured Data

Published:
by Wayne Smith

Image SEO isn’t just about ranking or visibility for a single keyword. Images are integral to a webpage’s content and influence its overall visibility. With AI-driven search, images now help systems understand and select relevant content more effectively. While AI evaluates overall page context, it complements – rather than replaces – traditional signals like page titles. Not all schema is equally valuable: generic markup offers minimal benefit, whereas context-rich schema provides a gain of knowledge that enhances both AI understanding and user engagement – key factors influencing search performance.

Modern Image SEO goes beyond traditional optimization by strategically leveraging schema to create blue ocean opportunities, standing out in under-served search areas. It considers the role of images in Shopping platforms, visual search tools, featured thumbnail images, AI-guided search, and user accessibility. Well-optimized images enhance visibility in visual results and AI summaries, driving traffic and improving user experience through clearer, faster content understanding and better accessibility.

Square, PowerPoint-style slide. Header Text: (Modern Image Search Uses). List items depicted: Visual product search with AI recognition, Identify products without knowing their names. Direct purchase options from images, See stores and prices instantly. Schema markup enhances AI understanding, Connects images to structured product data. Integrated into browsers and apps, Built into Chrome, mobile apps, and ecommerce. Improves visibility across AI and search platforms, Enhances rankings, summaries, and rich results.

Google Lens search is built directly into the Chrome browser. With a right-click, I can pull up Google Lens, highlight a product I don't even know the name of, and instantly see stores where I can purchase it – this is today’s image search in action.

Old School Image SEO

In the past, image SEO focused mainly on the image’s URL, file name, alt text, and the surrounding text. Today, however, the landscape has changed. The internet is filled with images -- almost every webpage includes them -- and the placement and relevance of these images can influence SEO. Additionally, Schema markup now allows websites to provide a gain of information, such as copyright and licensing details, which Google considers important. AI has also advanced to the point where it can create images and generate detailed descriptions for them, making visually descriptive schema, alt text, and semantic HTML even more essential.

Much of the information about image SEO is outdated; in short, it’s old school and needs to be updated.

URL and File name of the image

Both the URL and file name remain factors, and it is an SEO best practice to use a file name that clearly describes the image. However, due to the vast number of images online, the URL and file name now play a relatively minor role in modern and AI-based SEO.

Alt tag

For purely decorative images, it is best to leave the alt attribute empty (alt=""), which complies with ARIA web accessibility guidelines and ensures that screen readers skip these elements without impacting how search algorithms understand page content.

W3C – Alternative Text

For images with contextual importance, best practice is to provide clear, descriptive alt text to improve accessibility and help search engines understand the content. AI systems may also prefer more visually descriptive text, as this helps identify specific details even when users do not know the exact terms and are being guided by AI. A JSON-coded Schema description can also be helpful, as many AI systems read Schema data. The Schema description should align with the page content, but does not need to be a verbatim copy of the alt tag. Using a visually descriptive Schema is valuable, especially when such language would be awkward for page readers.

Nearby or adjacent text coded in the HTML

The text associated with the image is used by image search. Old School would generally consider nearby text as the text code in a sibling element on the page. When the text is significantly distant algorithms may lose confidence the text is related. While algorithms have improved, it is still an SEO best practice to keep related text near the image in the page’s code structure to maintain clear contextual association.

How AI Interprets Visual Descriptions

AI’s cost, speed, and accuracy are rapidly improving. While LLMs effectively extract key data points, generating precise images remains challenging. Accurate grammar and clear understanding are essential, as even a misplaced pronoun can cause confusion. Often, less is more: an expert can describe a visual with a single precise term, while someone unfamiliar with the correct terminology might need an entire paragraph.

(OpenAI Prompt Engineering Guide)

While a layperson using AI-assisted search may not know the correct terms, AI can guide them to what they want -- and yes, people may begin searching differently as they discover AI's ability to guide them. AI-assisted search changes how people find materials on the internet and changes the rules for websites to be found. Content for AI can be presented in a more knowledgeable tone, using terms specific to the industry.

(Google Search Generative Experience)

For example, if I search for "the thing with green, yellow, and red lights," AI search and overviews will usually suggest a traffic light as the most common object with those colors. If I then ask a follow-up question like "inside," AI narrows down suggestions to indoor lighting options with similar colors.

An Example of: Crafting Product Descriptions to Enhance AI Understanding

An example AI image discussed below

When I prompt an AI image generator with “traffic light,” it returns a traffic light. But if I use “a thing with green, yellow, and red lights,” results are inconsistent. This shows AI image generation requires precise terminology for accuracy.

Clothing is an area where precise terms matter. Google's Search Generative Experience (SGE) helps users refine vague queries by suggesting expert terminology. AI-assisted search guides users toward these precise terms, which can be supplied on the site using content and schema.

For example, when searching AI Mode for summer dresses, I found a product description, it uses semantic adjacent text:

“The Poised and Perfect Striped Midi Dress is a timeless dream! Featuring a notched neckline, button front, and a flowing midi length, this crisp striped design is effortlessly polished.”

Using this exact description in an AI image generator consistently produced images matching the website's product. Adding context such as “A 19-year-old girl at the beach in a…” helped define the model and background, extending relevancy to searches like “beach dress” or “summer beach dress.” I assume the target market is young women.

However, “at the beach” may not be appropriate for this dress. Creating accurate fashion content requires an expert who knows the correct terminology and suitable occasions. Expertise ensures precision and saves time.

Practical ImageObject Schema Example: Line-by-Line Breakdown

Below is a practical ImageObject Schema example featuring visually descriptive language that clearly establishes the relationship between the image and the page content.

Schema Validation

Validating your ImageObject Schema ensures it is correctly formatted, error-free, and usable by search engines and AI systems. Although Google’s Rich Results Test is the easiest tool for quick validation, note that ImageObject schema does not generate a visible Rich Results snippet in web search. However, validation confirms that your structured data is implemented correctly and can be interpreted by AI systems.

Other recommended validation tools include:

Use the "Test Schema" button above to validate the example using Google’s Rich Results Test.

It is also important to ensure that, beyond having correct JSON formatting, the actual content within your schema is accurate and meaningful. Ranking or visibility benefits of schema markup depend on multiple factors. When your schema content aligns well with page content, it enhances trust and understanding; if it conflicts, it can reduce credibility and hurt performance.

Image Objects are generally contained in a "hasPart" property

The "hasPart" property is being used here. This is meaningful, but it is not required. Its purpose is to indicate that the image is unambiguously part of the page content – it is not just a design or style element, nor is it an unrelated advertisement. The image is, in fact, part of the page’s meaningful content.

It is not required for an ImageObject to be placed under a WebPage schema. However, doing so can help clarify the relationship between the image and the page.

For example, if a product is being sold, the ImageObject can be included within the Product or Offer schema to clearly associate the image with the specific product being sold.

The "@type": "ImageObject"

It is common practice for "@type" to appear first in the JSON schema block, although there is no specific requirement for property order in JSON-LD. However, specifying the type is required to define the schema object as an ImageObject.

The "contentURL" Property

The contentURL property is required. It specifies the direct URL where the image file is located.

The Schema Name and Description Properties

The name and description properties in schema are optional, but they are both useful and flexible. They provide text that AI systems and search engines can interpret to better understand an image’s context. This improved understanding can lead to enhanced visibility in search results and greater inclusion in AI-generated summaries or answers.

Some people still use old-school keyword stuffing in these Schema properties, but it’s important to note that they are not treated the same way as a page title or meta description. Instead, the text in these fields helps AI systems and search engines interpret what the image is about and identify contextually relevant keywords. These properties provide meaning through context rather than acting as direct ranking signals like page titles traditionally do.

(Google Structured Data Guidelines)

Typically, the "name" property is used for the product and manufacturer in commerce schemas. However, in this example, the image serves as a example, so a name like “Schema ImageObject example of a summer dress using visually descriptive language” matches the context of this page. Schema is flexible allows the context for the image can be provided unambiguously. This flexibility provides the potential for "gain of knowledge."

Note: the alt tag, "An example AI image discussed below," provides screen readers with a clean narrative that when read makes sence in context.

The description property is straight forward: It is the description of the image for AI systems to read and interpret.

The "license", "acquireLicensePage", "creditText", and "copyrightNotice" schema properties

These properties can provide a gain of knowledge over pages that do not include them. When image search was first introduced, there were far fewer images online. Today, Google has access to more images than it can display, with many being duplicates from other sites. Providing clear licensing and copyright information helps Google determine which images to prioritize and display in search results.

Including this information is required if you want your images to appear with licensing details in Google Image Search. Specifically:

Google Images License Metadata

The license and acquireLicensePage properties are URLs and can point to the same or different pages. Generally, they link to pages on the publishing website, but this is not a schema requirement. For example, retailers often use images provided by manufacturers, where the license may point to the manufacturer’s terms.

In this example, the image was created by Stability AI. To my knowledge, the copyright status of AI-generated images has not yet been resolved by SCOTUS. Purely mechanical images are not copyrightable under US law, but whether AI-generated images are considered purely mechanical has not been definitively determined. Therefore, I reference Stability AI’s community license agreement for clarity.

These license pages do not require any specific legal wording in schema. For example, if images are covered under US copyright fair use, you may state this on your license page:

No commercial usage permitted. Images can only be used for reviewing the products under US copyright fair use and must include the copyright notice and attribution or credit text when using the image.

The "creator" property

Much like the copyright holder, the creator property identifies who actually created the image. This may or may not be the same as the copyright holder. Including the creator helps clarify authorship in structured data, enhancing transparency for AI and search engines.

The "about" property

The about property generally provides a link to an authoritative website (Wikipedia is widely accepted as authoritative) containing information about the entity, thing, location, person, or event shown in the image.

For example, if you are selling a summer dress, the about property might look like this:

Wikipedia schema entries often include an about property that points to Wikidata, and Wikidata then links to all language versions of that entity or topic. Including both is not required, but if your business is international or multilingual, referencing Wikidata can be beneficial.

Alternatively, manufacturers are authoritative sources for their own products. For example, if the image is of an Apple iPhone 16 Pro:

And yes, if you prompt an AI image generator with "Apple iPhone 16 Pro," it will produce an image of the phone.

Correctly using the about property can provide a “gain of knowledge” for AI systems, especially when the image is an illustration or a logical diagram representing broader concepts. For example, you might reference:

In the case of the above image created by AI for Schema, no about is being used -- the about property is optional -- the schema provides the context in the name and description.

Generic Schema vs. Gain of Knowledge

It should be obvious that simply having an image on a webpage does not improve SEO rankings. However, there is a caveat. Internal systems, such as what has been reported about Navboost, analyze which sites users click on, which pages appear to answer their questions, and how they interact with different types of pages. The user experience these systems evaluate can be positively, negatively, or neutrally affected by images on a page.

Additionally, AI systems and search algorithms attempt to predict how people will engage with new content when they visit a page for the first time. While images improve visual appeal, these predictive algorithms also consider factors like page speed, interactivity, and whether the content effectively answers users’ questions – especially when current search results do not fully address their needs.

Simply adding Schema markup to a webpage does not directly improve rankings. Its true value lies in providing search engines with structured knowledge about your content. This gain of knowledge helps search engines better interpret what your page offers, which can indirectly improve SEO performance by enhancing visibility and click-through rates.

Using the image file name as the Schema name property – as seen with generic schema implementations – adds no meaningful insight for search engines or AI systems. Repeating the file name or alt description within Schema provides no benefit. Schema becomes valuable when it supplies supplementary, additonal context-rich details, with the contentUrl linking this information to the actual content.

Forward-thinking modern SEO is about strategically enhancing content to stand out from the competition or to enter “blue oceans” with minimal competition. To rephrase the old saying, “All roads lead to Rome,” in SEO strategy, all blue oceans should lead to your content. Leveraging Schema for AI understanding creates these blue ocean opportunities. To achieve this, Schema should focus on answering questions such as:

Generic Schema implementations do not go above and beyond – they fail to create the gain of knowledge that leads to blue ocean opportunities.

Forward-thinking SEO recognizes that, for many topics, Google already has more content than it can serve. To remain visible, Schema and content must provide unique, structured insights that go beyond what is currently available, ensuring continued value and avoiding deindexing in future updates.

As users seek more specific answers and AI-guided search queries become increasingly detailed, pages that provide structured, in-depth information are surfaced more prominently, while those lacking such clarity and depth risk losing visibility.

Solution Smith tests SEO and guided AI search the same way it tests software -- methodically and with evidence. If a feature is claimed, it gets tested. Observations begin as anecdotal data points, which are then verified through repeated experiments.

Solution Smith does not rely on Google to confirm or deny findings -- in fact, it’s expected that Google and other search engines won’t publicly disclose the inner workings of their algorithms.