IPTC keywords vs. alt text: what actually affects Google image search

Search "image SEO metadata" and you'll find plenty of advice telling you to fill in a photo's IPTC keyword field before uploading it, as if a longer keyword list inside the file itself will help it rank. Google has never confirmed that this works, in any of its documentation, at any point. That's not the same as proving it does nothing – but after years of this advice circulating, there's still no published statement from Google saying embedded keywords are a ranking signal.

What Google has actually confirmed

Google's own image SEO guidance is specific about what it uses, and IPTC keywords aren't on the list. What is: the file name, the alt text, the text surrounding the image on the page, the image's caption if there's one, and the page's overall topic. All of these are things a crawler reads directly from the rendered page – not fields buried inside a file that only surface if someone downloads the image and inspects its metadata.

This distinction matters more than it sounds. A keyword typed into an IPTC field lives inside the image file itself. Google's crawler indexes the page – the HTML, the visible text, the alt attribute – not the internal metadata of every file linked from it. There's no confirmed mechanism by which a keyword sitting inside a JPEG's metadata block reaches Google's ranking systems at all.

The one place embedded metadata does matter

This isn't "metadata never matters" – that would overcorrect just as much as the original myth. IPTC and XMP copyright and licensing fields have a specific, documented effect: when they're filled in correctly, Google can show a credit line and a "Licensable" badge directly on the image in Google Images search results. That's a real, confirmed use of embedded metadata – it's just a different field, doing a different job (declaring who owns the image and under what terms), not a ranking lever.

The pattern worth remembering: metadata that Google's interface actually surfaces to a searcher (licensing info, shown as a badge) has a confirmed effect. Metadata that never appears anywhere a person or a ranking algorithm can see it (a keyword list buried in IPTC) doesn't have one that Google has ever confirmed.

So why does the keyword-stuffing advice persist

Partly because it used to be closer to true, in an earlier, less sophisticated era of image search. Partly because "add more keywords" is advice that's easy to give and feels like it should work, even without evidence. And partly because it's harmless enough that nobody notices it isn't helping – filling in an IPTC keyword field doesn't hurt anything, so the advice never gets tested against its absence.

What to actually spend the time on instead

If the choice is between typing keywords into a photo's metadata or writing a real alt text description, write the alt text – it's the one with Google's own documentation behind it. Same logic for file names: blue-ceramic-mug.jpg over IMG_4821.jpg is confirmed to help; a keyword field inside the file isn't.

This is also why TagDuck doesn't treat keywords as the star of the show. It generates them – search terms related to a photo have real value elsewhere, like matching a product feed or informing on-page copy – but the description (doubling as alt text) and the file name are doing the confirmed work. Keywords ride along as a bonus, not the main event.