Image SEO Checklist: 32 Steps for Better Visibility and Speed

The 32-step image SEO checklist

Use this checklist in priority order. A technically perfect decorative image creates little search value, while a unique diagram can still fail if it is blocked, hidden in CSS or delivered too slowly.

Make every important image crawlable and indexable

Place search-relevant images in an HTML <img> element. Google states that it does not index CSS background images, so a product photograph, chart or primary editorial visual should not exist only as a background declaration. Responsive markup should retain a valid fallback src.

Do not block essential image directories through robots controls, authentication or accidental CDN rules. Test the final rendered page, not just its source. JavaScript galleries, consent layers and client-side transformations can leave crawlers with a placeholder or an unusable URL.

Keep the same image at a consistent URL when its underlying asset has not changed. Google’s April 2025 documentation update specifically recommended consistent image URLs to support crawling and caching. If a migration changes hosts or paths, map old assets carefully, update references and inspect server logs for persistent failures.

An image sitemap is useful when images are difficult to discover through normal crawling, including some JavaScript galleries or CDN-hosted assets. It improves discovery, but does not guarantee indexing or rankings.

Write alt text, filenames and context for meaning

Alt text should communicate the image’s content or function to someone who cannot see it. Describe what is visible and relevant to the page. Do not force every target phrase into the attribute. Google explicitly warns that keyword stuffing in alt text can be treated as spam.

Useful example: A comparison page showing two compressed formats might use alt text such as, “Side-by-side crop comparing JPEG and WebP compression around fine hair.” “Best image SEO image optimization WebP JPEG” is not useful alt text.

Use an empty alt attribute for a purely decorative image so assistive technology can ignore it. If an image acts as a link or button, describe its function. For complex charts, use concise alt text and explain the findings in visible adjacent copy. This also makes the evidence available to text-based retrieval systems.

Choose readable filenames such as product-name-front.webp rather than IMG00491.webp, but do not churn existing URLs merely to rename files. Captions and nearby headings can clarify entities, locations, products and relationships more fully than a filename can.

Choose formats, dimensions and responsive variants

Google supports JPEG, PNG, GIF, WebP, SVG, BMP and AVIF. Match the extension to the actual encoded format. Choose by image characteristics rather than fashion: JPEG, WebP or AVIF can suit photographs; PNG can preserve transparency or sharp interface details; SVG can suit logos, diagrams and simple vector artwork.

Export from a clean, high-resolution source, then create variants near the rendered widths your layouts require. A phone should not download a desktop-sized photograph only to display it in a narrow card. Use srcset and sizes so the browser can select an appropriate candidate, while keeping a dependable src.

Compression is a quality decision as well as a byte decision. Inspect faces, gradients, text, product edges and fine texture at realistic display sizes. Excessive compression can reduce usefulness and trust even when a performance audit looks better.

Set intrinsic width and height values or an equivalent reserved aspect ratio. The 2025 HTTP Archive dataset found that 62% of mobile pages and 65% of desktop pages lacked explicit dimensions on at least one image, creating avoidable layout shift risk.

Optimize the LCP image without delaying it

Images were the Largest Contentful Paint element on 85.3% of desktop pages and 76% of mobile pages in the 2025 Web Almanac dataset. Image optimization is therefore often an LCP project, not merely an Image search project.

Identify the actual LCP element by template and device. If it is the hero image, include it in the initial HTML, avoid lazy loading it and make its request discoverable early. The same dataset found that about 16% of pages lazy-loaded their LCP image, a pattern that can delay primary content.

Lazy-load below-the-fold galleries and secondary thumbnails when doing so reduces unnecessary transfers. Reserve their dimensions so content does not jump as images arrive. Configure durable caching for versioned assets and avoid generating a new URL on every render.

Do not optimize solely against a synthetic score. Compare field LCP, image bytes, request timing and conversion behavior across page templates. A lower-quality hero that harms product evaluation may be a poor business trade even if it saves bytes.

Select primary images, structured data and licensing metadata

Choose a representative, high-resolution primary image that still works when cropped. Avoid extreme aspect ratios, small subjects surrounded by empty space and graphics dominated by text. Google identifies primaryImageOfPage, entity image properties and og:image as signals that can help systems select preferred previews.

Structured data must agree with visible page content. Product, recipe, article and other supported types can associate images with a specific entity, but markup does not make an irrelevant image relevant.

Publish creator, credit, copyright, license and acquisition information when rights management matters. Google supports both structured data and embedded IPTC metadata for image licensing information and potential Licensable badges. Verify that license URLs work and lead to genuine terms or acquisition instructions.

For duplicate or syndicated pages, maintain page-level canonical discipline. A canonical declaration consolidates page signals, but should not be treated as a substitute for stable, accessible image URLs.

Design images for search results, AI answers and visual discovery

A useful visual should answer a subquery on its own. Examples include an annotated process, original chart, before-and-after comparison, product angle, local landmark view or diagnostic flowchart. Surround it with a concise statement of what it shows, the entities involved and the conclusion a reader should draw.

This answer-first context helps conventional indexing and may make the page easier for AI systems to retrieve and summarize. Google Images can also contribute visibility across web Search, Discover and visual search experiences. However, no documented image technique guarantees inclusion in Google AI Overviews, AI Mode, Bing or Copilot, or ChatGPT.

Practitioners have observed AI Mode selecting and cropping prominent page images. Treat that as volatile behavior, not a ranking rule. Keep critical subjects near the useful center, test common crops and ensure that the visible article text contains any essential fact shown in a chart.

Create natural link demand with original datasets, embeddable diagrams and comparison assets. Provide attribution guidance without requiring manipulative links. Reclaim legitimate unlinked mentions and use link-intersect research to identify publishers that cite competing visual resources.

Diagnose missing rankings, slow images and bad previews

Use the failure symptom to choose the next test instead of changing every attribute at once.

Inspect representative URLs from every template. Segment logs by image directory, response code, crawler activity and CDN host. A spike in 404 or 403 responses after a deployment can explain lost visibility faster than an alt text rewrite.

Measure image SEO as a search and business channel

In Google Search Console, select Search type: Image and review impressions, clicks, CTR, queries, landing pages, countries, devices and average position. Compare page groups rather than relying on a sitewide average. Product pages, editorial diagrams and stock-heavy articles have different opportunities.

Pair search metrics with field performance and business outcomes. Useful KPIs include image search clicks, qualified sessions, assisted conversions, image bytes, LCP, CLS, crawl errors, indexed landing pages, licensing inquiries and earned links to visual assets.

Annotate migrations, compression changes, new primary images and markup deployments. Test controlled page groups where possible. A higher image CTR does not prove higher ordinary web rankings, and stronger page rankings may increase image impressions without any image-specific change.

Build a hub-and-spoke structure around visual intent. A central guide can link to format comparisons, compression tests, alt text examples, responsive implementation and troubleshooting pages. Consolidate overlapping assets, refresh obsolete screenshots and redirect retired supporting pages when a stronger replacement serves the same intent.

What is proven, what is consensus and what remains uncertain

Proven or officially documented

• Google uses page content and metadata to understand images.
• Important images should use crawlable HTML image elements rather than CSS backgrounds.
• Image sitemaps can help Google discover images, including CDN-hosted files.
• Alt text should be descriptive, and keyword stuffing can be spam.
• Search Console can report Image search performance separately.

Strong practitioner consensus

• Original diagrams, comparisons and product visuals usually create more search and link value than generic stock images.
• Stable delivery, contextual text, structured data and good performance work best as a combined system.
• Template-level auditing produces more reliable improvements than editing random images one at a time.

Uncertain or volatile

• No fixed formula determines which page image an AI answer system will display or cite.
• Image rankings do not reliably cause ordinary page rankings to rise. Community reports are anecdotal and should be tested independently.
• Format choice alone is not a ranking advantage. The practical outcome depends on quality, bytes, compatibility and delivery.

Implement in three passes: remove crawl and rendering barriers, fix performance and accessibility at template level, then invest in original visuals for pages with demonstrated query demand or commercial value.