What Is Digital Asset Management (DAM)?

A digital asset is any file a business uses and wants to reuse under control: a product photo, the logo, a campaign video, a presentation template, a product data sheet in PDF. The software is called a DAM system or DAM platform.

According to the German-language Wikipedia, the first systems were called image databases, because at first they only handled photos. In the English-language Google Books corpus, the frequency of the phrase "digital asset management" grew more than a hundredfold between 1995 and 2000, according to our own Ngram Viewer query.

Two variants have their own names: media asset management (MAM), when the focus is video and audio, and brand asset management (BAM) or brand portal, when the system serves logos and brand material to dealers and agencies.

• Early 1990s. The IPTC (International Press Telecommunications Council) creates the IPTC-IIM schema, so that details such as creator, caption and rights travel inside the image file. Before that, every system had its own format.
• 1992 to 1994. The first single-user programs appear, according to the German-language Wikipedia: Hasselblad's Pixolo, Canto Cumulus and, a little later, FotoStation from the Norwegian company FotoWare. They later become client-server systems.
• 2000s. Systems move into the browser, with the data held on a central server.
• 2012. XMP, the format developed by Adobe in which IPTC metadata is stored today, becomes the standard ISO 16684-1. The current second edition dates from 2019.
• 2017 to 2021. The IPTC Photo Metadata Standard adds star ratings and a link to the usage rights (2017), image regions (2019) and an alt text field for accessibility (2021).
• 2021. In December, version 1.0 of the C2PA specification is published. It is the technical basis of Content Credentials, which record where a file comes from and how it was edited.
• 2022. In June, the IPTC creates the value trainedAlgorithmicMedia, now labelled "Created using Generative AI". In November, the "Product Shown in the Image" field, which already took a GTIN, gets a generic identifier.
• 2023. In October, the IPTC adds the "Data Mining" field, which states whether data mining of the image is prohibited or permitted.
• 2024. The EU adopts the Artificial Intelligence Act, Regulation (EU) 2024/1689. Its Article 50 requires synthetic content to be marked in a machine-readable way.
• 2025. In November, IPTC 2025.1 adds four fields for AI-generated images: the prompt, its writer, the AI system used and its version.
• 2026. Article 50 applies from 2 August. Providers whose systems were on the market before that date have until 2 December to add the marking. In October, C2PA 2.5 is published.

A DAM system takes in files, adds information to them, controls how they are used and delivers them in different versions. The path of one product photo shows how.

Intake. Files arrive by upload, bulk import or through an API from other systems: the photo studio, the design agency, the PIM or the CMS. The DAM keeps the full-resolution original and can flag duplicates.

Description. An image carries two layers of metadata. The first travels inside the file: the camera's technical EXIF data and IPTC fields such as creator, copyright holder, caption or alt text, stored in the XMP format. The second exists only in the DAM: campaign, SKU, approved channels, licence expiry date. A controlled vocabulary, a fixed list of terms, keeps everyone tagging the same way. Tags that an AI suggests from the image content save time, but someone has to check them.

Versions and approval. Every change creates a new version and keeps the old one. A workflow sets who reviews and who approves before the file becomes available.

Rights. The DAM knows where each photo comes from, under which licence and until when. When a licence expires, it can pull the file from the channels or alert the owner.

Output. From one original, the DAM produces the versions each channel needs: a square crop for social media, a 1,200-pixel JPEG for the marketplace, WebP or AVIF in several sizes for the store. Delivery can run through a CDN, and every page uses the same URL for each image.

Archive. Files that are no longer in use move to an archive with their metadata instead of being deleted.

Images on the web. Google's image SEO guidelines, last updated in March 2026, list points a DAM can turn into fixed rules: If alt text is a required field in the DAM for each language, the store, the blog and the featured image of each article receive it ready-made.

• Formats that Google Search supports: BMP, GIF, JPEG, PNG, WebP, SVG and AVIF, with a file extension that matches the file type.
• Short, descriptive filenames instead of image1.jpg, translated when the image is localized.
• Alt text that describes the image. Google uses it together with computer vision and the page content to understand what the image shows.
• Several resolutions of the same file, served with the srcset attribute or the picture element, so a phone does not download the large version.

Source: own compilation from vendor pages (accessed 08.11.2026). Media asset management for video and audio is covered under the differences.

Worked example, with no data from any company: an online kitchenware store with 3,000 products, six photos per product, three languages (English, Spanish and German) and four channels (store, marketplace, Google Shopping and social media).

The DAM holds 18,000 originals (3,000 times 6). If each original is delivered in four versions, that makes 72,000 derived files nobody uploads by hand. With one alt text per language, the store needs 54,000 texts (18,000 times 3). That is why the field is required at upload and not filled in later in the store.

One of those photos, a 28 cm frying pan on a white background, looks like this in the DAM:

• Filename: nonstick-frying-pan-28-cm-front.jpg, with translated names for the Spanish and German versions.
• IPTC fields inside the file: creator (the photo studio), copyright holder (the store), alt text in three languages.
• DAM fields: product SKU and GTIN, the studio's licence with no end date, approved channels (all four), autumn campaign.
• Versions: 2,000 pixels for the zoom view, 1,200 for the marketplace, 600 as WebP for the category page and a square crop for social media.

In the PIM, the pan's record stores the link to the DAM file by SKU. If the studio delivers a better photo, it is replaced once in the DAM, and the store, the marketplace and the feed show it after the next update.

Source: own formulas; no published benchmarks with a stated method exist.

Source: own compilation.

Source: own compilation; DAM per the Gartner glossary; cloud storage and MAM per the English-language Wikipedia.

In the Keyword Planner figures gathered for this entry (Google, November 2026), Bynder has about 6,600 brand searches a month in the United States and Brandfolder about 4,400, the highest among the DAM vendors checked. The table is alphabetical and is not a ranking.

Source: vendor pages, accessed 08.11.2026; alphabetical order.

A DAM organizes what it receives. Files that arrive with no creator, no licence or a name like IMG_4711.jpg get stored anyway, and filling in that information is still work for people, which can cost more than the licence.

The second limit is the boundary with other systems. If product photos can be uploaded to the DAM and also straight to the store or the PIM, two versions of the same image appear and nobody knows which one is right. Before rollout, decide which system leads for each kind of file.

Automatic tags make their own mistakes. An AI recognizes a frying pan but does not know whether it is the 24 cm or the 28 cm model, or whether the photo may be used in Germany. Metadata with legal or commercial consequences remains human work.

Whether a photo or video may be used depends on its licence, which is why a DAM records the rights holder, terms and expiry date. Photos of identifiable people also raise privacy questions; in the EU, the GDPR applies. For businesses selling into the EU, Article 4 of Directive (EU) 2019/790 allows text and data mining unless the rights holder has reserved it in an appropriate manner, such as machine-readable means for content made available online. The IPTC "Data Mining" field can record that choice inside the image, though the IPTC notes that the applicable law overrides the field value. For AI-generated images, Article 50(2) of the EU AI Act, Regulation (EU) 2024/1689, requires the provider of the AI system to mark them in a machine-readable format, by 2 December 2026 for systems placed on the market before 2 August 2026. A DAM can keep that marking and the Content Credentials when it creates new versions instead of stripping them; see also the entry on AI content labelling.

• Count what you have: number of files, formats, where they sit and how many duplicates there are.
• Define the metadata schema: which fields are required (creator, licence, expiry, alt text per language, SKU) and which controlled vocabularies apply.
• Record the rights of each file before migrating it. Anything without a clear licence stays out of the channels.
• Set a rule for filenames and folders, with descriptive words in the publishing language.
• Start with one campaign or one product category as a pilot and measure completeness before and after.
• Connect the DAM to the PIM, the CMS and the store so they pull images from it and not from local copies.
• Check the metrics from the section above every month.

• Upload each photo once at full resolution and let the DAM create the versions.
• Link product images by SKU or GTIN, not by filename.
• Write alt text in every language you publish in, not only in English.
• Store creator and rights inside the file as well (IPTC), so they travel with the image.
• Keep provenance data and Content Credentials when exporting.

• Accepting AI-suggested tags without review.
• Recording rights only in the filename or in a separate spreadsheet.
• Uploading the same image in five sizes instead of generating versions from one original.
• Leaving alt text until the end and typing it into the store, outside the DAM.
• Migrating the whole old archive unfiltered and filling the system with duplicates.

This entry was written in November 2026 with the help of AI. It started from an analysis of the top ten search results in the United States, Spain and Germany and of the answers from four AI assistants, to cover what they leave out: the IPTC and XMP metadata standards, images for the web, content provenance with C2PA and the link to the PIM. Every fact was checked against the original source.