The model sees more than a file name
Comparing the image with its description distinguishes useful ALT text from an empty label based solely on the file name.
sala-1.webp · hotel function room
The photo shows a specific conference room in a theatre-style layout, with a screen, stage and access to natural daylight.
Description based on the image and source
AI suggests specific ALT text after confirming the space’s name in the hotel’s approved library.
- description matches the image content
- venue name sourced only from the hotel
- status requiring editorial approval
Image descriptions should be accurate and useful
The model can recognise elements visible in a photo, but should not guess the name of a room, room category or location.
Detects gaps
Finds images without ALT text, identical descriptions and technical text based on file names.
Analyses content
Recognises the visible space, layout, facilities and character of the image.
Suggests descriptions
Combines image analysis with approved names and prepares text for approval.
How is an image audit created?
Inventory
The module collects images, page addresses, existing ALT text and basic metadata.
Visual analysis
The model describes what can genuinely be recognised in each image.
Comparison with sources
It checks the existing ALT text and approved names in the hotel’s additional materials.
Suggestions for approval
It prepares improved descriptions and a list of images requiring manual identification.
The image and hotel knowledge play different roles
Visual analysis describes visible content. The proper name of a space or service comes exclusively from an approved source.
What does AI use?
- images and their page addresses
- existing ALT text and file names
- content library with metadata
- optional knowledge from the Context Interview
What does the team receive?
- list of missing and incorrect ALT descriptions
- comparison of the description with the image content
- new descriptions for approval
People confirm names that cannot be seen
The model should not guess room names or categories. Uncertain identifications are flagged, and an editor selects the final text.
An editor approves every ALT suggestion, confirms proper names and can exclude selected images from the analysis.
The module does not modify images, publish descriptions, assess copyright or create place names based on appearance alone.
Better descriptions without guessing what is shown
Images without useful descriptions are shown with their address and page context.
The suggestion reflects elements actually present in the photo.
The editor receives a description to review instead of creating one from scratch.
The website supplies images; the library supplies the right names
The web integration finds images and existing descriptions, while the Additional Sources Index can provide approved content metadata.
Related process elements
Frequently asked questions
Can AI recognise a specific hotel function room?
It can recognise the type of space, but the proper name should be confirmed in approved materials or by the team.
Should every ALT text include the hotel name?
No. The description should primarily reflect the image’s purpose and content, without artificially repeating the brand.
Does the module change descriptions on the website?
No. It prepares suggestions for approval and handover to Technical optimisation.
Does it also analyse decorative images?
It can detect them, but the team decides whether they should have a description or remain ignored by assistive technologies.