Image metadata and dimensions

How to use the tool

Drag an image into the frame, or click it to pick a file. Your browser reads the file on the spot: it stays on your computer and nothing reaches the server.

For an image that is already online, paste its address and click “Analyse”. The nemociment.fr server then downloads the image, up to 5 MB, and sends it back to your browser, which analyses it with the same code as a local file. This mode also shows two headers sent by the original site, Content-Length and Cache-Control, handy for checking how your images are cached.

The results open with a preview and a table of properties. Then come the alerts, each with its source, followed by the EXIF fields, the GPS position, the IPTC fields and the XMP data.

What the tool reads in the file

An image format is identified by its first bytes, known as its signature: FF D8 FF for a JPEG, “GIF89a” for a GIF, “RIFF” followed by “WEBP” for a WebP. The tool relies on that signature and never on the file name extension or on the type the browser reports. A PNG renamed to .jpg is still a PNG, and the tool says so: Google states that it is preferable for the file name extension to match the file type.¹

Dimensions come from each format’s own header: the SOF segment of a JPEG, the IHDR chunk of a PNG, the ispe property of an AVIF, the width, height and viewBox attributes of an SVG. iPhone HEIC files share the AVIF structure; the tool recognises them and lists them separately, since most browsers cannot display them.

EXIF data is written by the camera or phone when the picture is taken: make, model, settings, date, orientation and, when location is on, GPS position. IPTC fields come from the press and photo agencies; they hold the title, caption, keywords, creator, credit line and copyright. XMP, an Adobe format written in XML, can carry all of this and more, such as the address of a licence or the digital source type.

The properties table also gives the weight of this metadata, in kilobytes and as a share of the file, block by block. On a small thumbnail, a colour profile and an XMP packet can weigh more than the image itself.

Image file size, by Chrome’s rule

Chrome DevTools flags heavy images in its “Improve image delivery” insight.² Its source code uses a target of 2/12 of a byte per pixel, that is 1/6 of a byte per pixel, the size of a well-compressed AVIF image. The tool applies the same calculation to the file’s dimensions. Above 1/6 of a byte per pixel, and when the possible saving reaches 4,096 bytes, it suggests a modern format (WebP, AVIF) or more compression for a JPEG, PNG or BMP, and more compression for a WebP or AVIF, which are already modern formats.

An animated GIF over 100 KiB follows another rule: Chrome suggests replacing it with a video.² The same Chrome page notes that AVIF and WebP compress better than JPEG and PNG.² SVG files are left out, because an SVG describes shapes and has no pixels to count.

The result is an estimate. Chrome also compares the file with the size at which the page displays it: a large photo squeezed into a small frame wastes bytes the tool cannot measure, since it only sees the image and knows nothing of the page around it.

Formats and attributes Google takes into account

Google supports BMP, GIF, JPEG, PNG, WebP, SVG and AVIF images in the src attribute of an img element. An image set as a background by a CSS stylesheet is not indexed. Google also recommends short, descriptive file names and calls alt text the most important attribute for describing an image.¹

The same advice applies to a page’s share image, the one declared in the og:image tag and picked up by social networks. Once its dimensions and weight are checked here, the tag itself can be written with the meta tags and Open Graph generator, which shows a preview of the share.

GPS position and personal details

A phone whose camera has access to location stores the GPS position of every shot in the EXIF data. If the photo is published as is, anyone can download it and find the place. In December 2012, PetaPixel told the story of John McAfee: a photo published by VICE had kept its EXIF metadata, which showed an iPhone 4S and the exact place where it was taken, in Guatemala. McAfee later admitted that the photo had given away his position.³

The tool shows the position in decimal degrees, with a link to the point on OpenStreetMap. It also lists the fields that say something about the equipment and its owner: camera make and model, body and lens serial numbers, owner name, software, date taken. For a product shot these details hardly matter; they are of no use to visitors and still travel with every copy of the image.

EXIF orientation

A phone held upright often saves the image the way the sensor sees it, with an Orientation field that gives the rotation to apply. According to MDN’s page on the image-orientation property, its initial value is from-image: browsers apply this rotation by default, and this support has been available everywhere since April 2020.⁴

The image therefore shows the right way up, yet its stored dimensions remain those of the sensor. For orientations 5 to 8, width and height are swapped on screen; the tool then gives both, and the displayed dimensions are the ones to use in the width and height attributes of the img element.

The metadata Google reads for Google Images

According to Google’s page on image license metadata, Google reads several IPTC fields: copyright notice, creator, credit line, digital source type, Licensor URL and Web Statement of Rights. The Web Statement of Rights is required for the “Licensable” badge. The same information can be given in the page’s structured data; when the two disagree, Google uses the structured data.⁵

The digital source type (DigitalSourceType) states how the image was produced. The values trainedAlgorithmicMedia, compositeSynthetic, algorithmicMedia and compositeWithTrainedAlgorithmicMedia mark an image created or altered by AI, and the tool raises an alert when it finds them. It also notes the presence of a C2PA manifest, the Content Credentials standard, without checking its signature: Google says a signed C2PA manifest, version 2.1 or later, can feed the “About this image” panel.⁵

Labelling AI-generated content applies to text too. To comply with Article 50 of the AI Act, OpenAI announced an invisible watermark, textGrain, placed in the word choices of ChatGPT text for its users in the European Union; what it means for a website is covered in the article on the ChatGPT watermark and website content.

EXIF data and Google rankings

On September 12, 2024, Search Engine Roundtable reported that Google’s Martin Splitt had said at SMX Advanced London that Google does not use EXIF data for ranking. The article recalls two earlier statements. In 2014, Matt Cutts explained that Google can read this data and reserves the right to use it. In 2019, Gary Illyes said it was not used for ranking and that IPTC was used for copyright.⁶

The value of image metadata lies on the side of privacy and copyright. A forgotten GPS position exposes the photographer; a well-filled creator field is part of what Google reads for Google Images.

Removing metadata while keeping your rights

The “Download a copy without metadata” button redraws the image on a browser canvas, then saves it in the same format when the browser can encode it, at quality 0.9 for JPEG and WebP. Otherwise the tool says so and provides a PNG. The copy loses all its metadata, colour profile included, and becomes a new file whose size and quality change. The button stays disabled for an SVG, which is a text file, and for an animated GIF, PNG or WebP, whose copy would keep only the first frame.

Google points out that removing metadata reduces file size but that doing so may be illegal in some countries, and recommends keeping at least the IPTC creator, credit line and copyright fields.⁵ When your image contains them, the tool reminds you just below the button. To remove only the GPS position and keep the rest, a program such as ExifTool does the job, with the command exiftool -gps:all= photo.jpg.

Frequently asked questions