P: Previews fail to render for Lightroom cloud assets searchable through Adobe/ChatGPT integration

I am using the Adobe integration within ChatGPT with my Adobe Lightroom cloud catalogue.

ChatGPT can successfully search my Lightroom catalogue and Adobe returns the expected assets, including filenames, asset IDs, file types, dates, orientation information and Adobe rendition URLs.

The Adobe search-result interface can also display thumbnails of the Lightroom images to me.

However, when ChatGPT attempts to obtain a model-readable preview of any Lightroom asset through the Adobe integration, the preview-resolution step fails.

Adobe returns:

• previewUrl: null

• renditionUrl: null

• downloadUrl: null

As a result, ChatGPT can search and identify photographs in my Lightroom catalogue but cannot actually inspect or analyse the image pixels.

File formats tested

This has been reproduced with both:

• Fujifilm .RAF RAW files

• Standard .JPG / image/jpeg files

Therefore, this does not appear to be related specifically to Fujifilm RAW support.

For example, Lightroom successfully returned JPEG assets including:

• DSCF8305.JPG

• DSCF8288.JPG

• DSCF8322.JPG

• DSCF8325.JPG

• DSCF8313.JPG

These are correctly identified as image/jpeg.

The catalogue search found 691 JPEG assets.

However, requesting accessible previews for five of these JPEGs resulted in null preview, rendition and download URLs for all five.

The same behaviour occurs with RAF files.

Authentication troubleshooting

I disconnected/re-authorised my Adobe connection and repeated the test.

After re-authentication:

1. Lightroom catalogue searching continued to work.

2. Fresh Lightroom asset IDs and rendition links were returned.

3. Adobe search-result thumbnails remained visible to me.

4. Resolving those assets into model-readable previews still returned null URLs.

Therefore re-authenticating the Adobe account did not resolve the issue.

Steps to reproduce

1. Connect an Adobe account containing a Lightroom cloud catalogue to ChatGPT.

2. Ask ChatGPT to search Lightroom for photographs.

3. Lightroom catalogue search succeeds.

4. Select one or more returned Lightroom assets.

5. Have the Adobe integration resolve the authenticated Lightroom rendition into a preview that ChatGPT can inspect.

6. The preview resolution returns null URLs.

7. ChatGPT therefore cannot visually analyse the photograph despite the asset appearing successfully in Lightroom search results.

Expected result

A Lightroom asset returned through catalogue search should provide an authenticated/presigned image preview that the ChatGPT Adobe integration can retrieve.

This would allow ChatGPT to visually analyse the selected Lightroom photograph.

Actual result

Catalogue metadata and user-facing thumbnails are available, but the model-readable image preview cannot be obtained.

The Adobe preview-resolution response contains null preview, rendition and download URLs.

Important distinction

The photograph thumbnail is visible to me in the Adobe search-results interface inside ChatGPT.

The problem is specifically that the image cannot subsequently be handed to the model for visual analysis.

This suggests the Lightroom search/thumbnail functionality itself is working, while the authenticated rendition/presigned-preview handoff is failing.

Impact

This prevents workflows such as:

• asking ChatGPT to assess Lightroom photographs for sharpness or composition;

• comparing several Lightroom images;

• selecting the best photographs from a shoot;

• assessing photographs for competitions or stock photography;

• providing editing recommendations based on the actual image;

• performing AI-assisted workflows on photographs discovered directly through the Lightroom catalogue.

The Lightroom catalogue/search integration is otherwise working correctly.

Please investigate the Lightroom asset preview/rendition authentication or presigned-URL handoff used by the Adobe integration with ChatGPT.