Image Tagging for SEO: Which Tags Google Actually Reads (2026)

Your site carries 2,000 images, the files are named DSC_0481.jpg, the alt attributes are empty, and Google Images ranks almost none of them. Fixing that is what image tagging for SEO means, and most guides on it never tell you which tags Google can see.

"Image tag" means three different things

Only one of the three things people call an image tag is visible to Google: the text on the page. Most confusion about image tagging for SEO comes from one word doing three jobs.

• On-page tags. Text in your HTML: the alt attribute, the filename in the image URL, the caption, structured data. Google reads all of these.
• Embedded metadata. Keywords written into the file itself, in IPTC or XMP fields. Lightroom and stock agencies search these; Google does not rank on them.
• Catalog tags. Keywords stored in a photo tool's index so your team can find shots. Invisible to Google entirely.

Blending these layers is how site owners end up keywording 2,000 files in Lightroom and wondering why search never changed. Spend your SEO effort on the first layer, and use the third to produce it faster.

The image tags Google reads

Alt text leads the list: Google's image SEO guide calls it "the most important attribute" for giving Google metadata about an image. Google also reads the photo itself with computer vision, alongside the alt text and the page around it, so your words should describe what is really in the frame.

• Alt text. The attribute most sites skip. Write one sentence describing what the photo shows, subject first: "Matte black ceramic mug on a walnut desk." It doubles as screen-reader text, and Baymard Institute found 55% of e-commerce sites fail to communicate the content of informational images through alt text.
• Filename. Google says a filename gives "very light clues about the subject matter of the image," so black-ceramic-mug-walnut-desk.jpg beats IMG_2841.jpg. Set it at upload, since renaming changes the URL; our photo file naming conventions cover the pattern.
• Captions and surrounding text. Google pulls subject matter from the page, "including captions and image titles," so a photo inside relevant text gets context a bare gallery grid lacks. A title attribute is no substitute for alt text: it shows a hover tooltip, and screen readers handle it inconsistently.
• Structured data. Product, article, and recipe schema tell Google which image represents the page and can put your images in rich results, "including a prominent badge in Google Images." The schema's image property is required for that badge.
• An image sitemap. It lists images Google "might not have otherwise discovered," like photos loaded late or hidden behind galleries, and the image URLs can point at a CDN on another domain. It aids discovery rather than rank, but an image Google never finds never ranks.
• A real <img> element. "Google doesn't index CSS images," so a photo set as a CSS background-image gets none of the text above.

Note. One natural description beats five keyword variants. Alt text stuffed with synonyms reads as spam to Google and as noise to a screen-reader user. If it sounds like a search query instead of a description, rewrite it.

The tags that do nothing for your ranking

Three popular kinds of image tagging do nothing for rank, though two of them still have other jobs.

• Embedded IPTC keywords. IPTC keywords are not among the embedded fields Google extracts. Google's image metadata documentation lists Creator, Credit Line, Copyright Notice, Digital Source Type (which flags AI-generated or composite images), and licensing fields, used to show credits and a "Licensable" badge in Google Images. Keep those accurate; keywords still serve stock submission and desktop search.
• Geotagged EXIF. Writing GPS coordinates into files for "local SEO" is a persistent myth. Nothing in Google's image documentation asks for it.
• Catalog tags. Whatever your photo tool stores in its own index never renders in your HTML, so Google never sees it.

Catalog tags still earn their keep: they are how the visible text gets written at volume.

Automating image tags for SEO at volume

Volume, not the formula, is what stalls image SEO, and Google's guide says as much: "If your site has thousands of images, you might want to consider automating the naming of the images." At about a minute per photo by hand, 2,000 images means 2,000 alt texts, 2,000 filenames, and a project that dies in week two.

Past the 1,000-photo wall, typing it all yourself stops being a plan, so let an AI vision pass read each photo once and write three SEO outputs:

• Alt text. A complete sentence fit to publish, the editorial-grade alt text a page can use as written.
• A filename. A descriptive name that follows the focal subject, the single dominant element of the shot: black-ceramic-mug-walnut-desk.jpg.
• Catalog tags. Searchable labels like "ceramic mug" and "walnut desk," so the frame turns up again later.

Tagrly writes all three for every photo it reads, plus a description and a focal subject. Its hosted copy of each photo carries the descriptive name plus a short ID, so similar shots never collide (black-ceramic-mug-walnut-desk-<id>.jpg), while your own Drive or Dropbox files keep their names.

The tags pay off when you go looking. In Tagrly's own testing on a 5,000-photo set, focal-subject tagging surfaced the correct top match in roughly 9 of 10 searches, against roughly 4 of 10 for the same library under a generic object detector, which is how AI photo search finds the right frame.

How to tag images for SEO at scale

Tagging images for SEO at scale comes down to two things: how many images you have, and where the originals live.

• By hand, under about 300 images. Use the sentence formula and fix filenames at upload. Our free AI alt text generator drafts a line for one photo with no sign-up, and our alt text for product images guide covers product shots.
• A CMS plugin, if photos only live in your platform. WordPress plugins and Shopify apps write alt text where the images sit. Quality varies widely, so read the output on ten of your own photos before letting one loose.
• A connected catalog, if the originals live in Drive, Dropbox, or on your computer. It reads those folders, or photos you upload, and writes alt text, a filename, and tags for every frame.

Tagrly is one example of that third path: it reads Google Drive or Dropbox with read-only access, or takes uploads from your computer, and keeps a hosted copy of each photo with its tags and alt text. The text reaches your site in one of three ways:

• WordPress. Tagrly's plugin imports a photo into the Media Library with its alt text filled in, under the hosted copy's descriptive filename. For images already in the library, its write-back button fills only empty alt fields and never overwrites alt text you wrote.
• Shopify or any other platform. Tagrly has no Shopify app, so you copy each photo's alt text from its detail panel, or a developer pulls alt_text and suggested_filename through the API.
• Hand-built pages. A collection's Export block gives ready-to-paste <img> HTML with the alt attributes filled in, and its zip download names every file descriptively.

Tip. Judge output before you commit to any tool. Tagrly's free tier reads your first 500 photos with the same full reading every plan gets, no credit card, so you can test it on your own images and read the alt text it would hand your site.

Whichever path you pick, review the generated text on your top 20 pages by traffic yourself. A vision model can see "navy linen sofa," but it cannot know your product names.

The bottom line

Image tagging for SEO is narrower than the phrase suggests: Google reads alt text, filenames, surrounding text, structured data, and sitemaps, and nothing embedded in the file or stored in a photo tool moves rank on its own. Write one plain sentence per image and name the files like a human. Under a few hundred images, do it by hand this week. Past that, point an AI at the library and judge its output on your first 500 photos, free, before you commit.