A product page can display the right diagram while the asset inventory points to an old file. A sitemap can list an image that now requires a login. A content delivery network can serve several exports of one illustration under unrelated URLs.
Those are different faults. An image sitemap addresses discovery information. It cannot settle which drawing the business intends to publish or remove a delivery restriction.
Start an image discovery audit with the relationship between a page and its assets. Record what the page uses, where each image lives, and what a public request receives. Then compare that record with the sitemap. The worksheet below is a proposed audit method, illustrated with fictional URLs. It is not a report about any live asset library.
Define the outcome before collecting URLs
Give the audit a bounded question: do these public product pages reference the intended public images, and do the image sitemap entries describe those relationships?
Keep four states separate in the report. Discovery means a search engine encounters a URL. Crawling means it requests the resource. Indexing concerns whether the resource enters its index. Search visibility concerns whether it appears for a particular search.
A local request proves what that request received. It does not prove Google crawled the resource. A submitted sitemap proves neither indexing nor visibility. Google's sitemap overview states that sitemap inclusion does not guarantee crawling or indexing.
Write the acceptance criteria around things the team can inspect. Examples include an intended image reference, a working public response, a recorded host owner, and a correct sitemap entry. Keep future search appearance outside that acceptance checklist.
Use the broader technical SEO audit checklist for site-wide review. This audit follows the image relationship through the publishing and delivery systems.
Map pages to images and hosting domains
Begin with a small representative set. Include a product page, a technical article, and a page that reuses an existing asset. Expand only after the team agrees on what each worksheet field means.
The following inventory uses reserved example domains. None of these URLs represents a checked website.
| Page URL | Image URL | Hosting domain | Intended role |
|---|---|---|---|
https://example.com/products/inspection/ |
https://cdn.example.com/inspection-flow.png |
cdn.example.com |
Current process diagram |
https://example.com/guides/rework/ |
https://cdn.example.com/inspection-flow.png |
cdn.example.com |
Reused process diagram |
https://example.com/products/sensor/ |
https://media.example.net/sensor-front.webp |
media.example.net |
Public product photograph |
https://example.com/guides/legacy-line/ |
https://cdn.example.com/old-line.png |
cdn.example.com |
Asset awaiting owner decision |
Keep one row per page-image relationship. A unique image list alone hides reuse. A page list alone hides shared hosting dependencies.
Add the asset's source record if the content management system provides one. A media identifier helps distinguish one file reused twice from two separate uploads. Record the final requested URL separately from any authoring field. The address an editor selects may differ from the address a publishing template emits.
Do not infer ownership from a familiar hostname. Assign someone to confirm who controls the host and who can investigate its delivery settings.
Compare the inventory with a minimal image sitemap
Google permits a separate image sitemap or image extensions in an existing sitemap. Its image sitemap documentation defines image:image and image:loc as required image tags. The page's loc identifies the containing page; image:loc identifies its image.
This fictional example maps two pages to their images. It uses the documented namespaces and supported image tags. It is a teaching example, not an instruction to change a production sitemap immediately.
<?xml version="1.0" encoding="UTF-8"?>
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9"
xmlns:image="http://www.google.com/schemas/sitemap-image/1.1">
<url>
<loc>https://example.com/products/inspection/</loc>
<image:image>
<image:loc>https://cdn.example.com/inspection-flow.png</image:loc>
</image:image>
</url>
<url>
<loc>https://example.com/products/sensor/</loc>
<image:image>
<image:loc>https://media.example.net/sensor-front.webp</image:loc>
</image:image>
</url>
</urlset>
Google's documentation no longer includes the deprecated caption, geographic location, title, and license image tags. Keep this example limited to the current documented fields. Put the image's editorial explanation on its page.
Compare entries with the inventory in both directions. Which intended public image is absent? Which listed image no longer belongs to that page? Assign each mismatch an owner and a proposed decision before generating a replacement file.
Check access on the image's own host
Cross-domain image URLs are allowed. Google's image sitemap guidance calls for verification of both domains in Search Console and warns against robots rules that prevent crawling of desired content.
Record verification as an evidence field. “CDN configured” does not establish which property is verified or who can see its reports. When the team cannot verify a third-party host, record that limitation and request the provider's documented options. Do not claim control the team lacks.
Inspect the image host's robots rules as well as the page host's rules. The worksheet should identify the host, relevant path, checked date, and observed restriction. A permissive page host says nothing about a separate image host.
For public access checks, use a fresh request without an authenticated editor session. Record the response status and whether the response contains the intended image. A success status carrying an access-denied document still fails the intended-public-image check.
Google's technical requirements explain that login-protected pages are not publicly crawlable and that technical eligibility does not guarantee indexing. Apply that access distinction when assessing the asset delivery route. Keep the private library private; decide whether a separate approved public export is needed.
Treat temporary URLs as a publishing decision
A fictional image URL might include a token that expires after an editor preview. Record that expiry as a suspected delivery dependency, then confirm its behavior with the responsible developer. Do not assume every query parameter means expiration.
Ask three concrete questions. Does the URL work without the editor's session? Does it remain usable after the documented expiry window? Can the publishing system provide an approved public URL for the intended asset?
Do not strip tokens from a URL and treat the resulting address as valid. The token may control access. Do not copy confidential signatures into an audit circulated beyond the permitted team. Store only the evidence the task requires.
Separate the decision to publish an asset from the work to make it reachable. If the file contains restricted drawings, the correct outcome may be exclusion from the public library. A discovery audit does not authorize changing the access policy.
For a public asset, prefer a deliberate published reference over an accidental preview link. Google's image SEO guidance recommends consistent URLs for reused images. Use that principle when defining the asset's public address.
Reconcile duplicates without deleting evidence
Two different URLs may contain the same image. Two similar drawings may describe different product revisions. Keep those cases separate until an owner checks them.
Add a duplicate group field to the worksheet. For the fictional inspection flow, the group might contain a current CDN export and an older upload on the main domain. Record which pages reference each URL before proposing consolidation.
If the files match, ask whether one published URL should serve both pages. If they differ, compare labels, conditions, revision notes, and intended audiences. An image that looks almost identical can still express a different operating boundary.
Do not remove a file because it lacks a sitemap entry. It may support a page outside the initial sample. Likewise, do not retain an outdated drawing merely because it receives requests. The technical evidence informs the editorial decision; it does not make it.
Use the headless CMS guide when mapping ownership across authoring and delivery systems. Keep the specific image decision in this audit record.
Use a worksheet that ends in an assigned action
Keep observations, decisions, and verification separate. “Old image” is an observation someone still needs to prove. “Replace the reference” is a proposed action. “Page now references the approved replacement” is a checkable result.
| Field | What to record |
|---|---|
| Containing page | Exact public page URL |
| Referenced image | Exact image URL found on the page |
| Asset identity | Source media record or reviewed filename |
| Public access | Request date, response, and content observed |
| Host access | Relevant robots restriction and host owner |
| Verification | Search Console property evidence or known limitation |
| Sitemap relation | Present, missing, outdated, or intentionally excluded |
| Duplicate group | Related asset URLs and confirmed differences |
| Decision | Keep, replace, investigate, or exclude |
| Owner | Person responsible for the next step |
| Completion evidence | Rechecked page, image, and sitemap relationship |
For removed images, record the page decision first. Does the page need a replacement, a revised explanation, or no image? Then check whether the sitemap still names the removed asset.
For a CDN move, record the old URL, intended new URL, affected pages, and observed behavior at both addresses. Use the site migration checklist for the wider release plan. This worksheet adds the image relationships that a page-only inventory can miss.
Recheck the relationships after the change
Use the original sample as the first acceptance set. Reopen each containing page and confirm the intended image reference. Request the image independently. Compare the resulting address with the sitemap entry.
Check reused assets on every sampled page. One template may adopt the new host while another retains an old reference. Record each result separately so a passing product page does not conceal a failing guide.
Keep the completion note precise: the reviewed pages reference the approved public image URLs, and the inspected sitemap represents those relationships. Record any remaining host verification or access questions by name.
Search performance belongs in a later observation period. Set its scope, dates, and available evidence separately. Do not label a delivery fix unsuccessful because an immediate manual search does not show the image. Do not call it a visibility success because the sitemap submission succeeds.
Leave the next reviewer a dated inventory and specific unresolved rows. The asset library needs a traceable public path.
Sources
- Google Search Central: image sitemaps. Supported image tags, cross-domain hosting, verification, and deprecated fields. Checked 2026.10.28.
- Google Search Central: image SEO best practices. Consistent references for reused images. Checked 2026.10.28.
- Google Search Central: sitemap overview. Sitemap purpose and crawling/indexing limits. Checked 2026.10.28.
- Google Search Central: technical requirements. Public access and indexing eligibility. Checked 2026.10.28.