Alt text, captions & embedded metadata: how they work together
Alt text (the HTML alt attribute) makes an image accessible and acts as a fallback if it fails to load — but it only exists on the one page where you wrote it. Embedded IPTC/XMP metadata inside the file itself travels with the image wherever it goes. This guide explains the difference, why both matter, and how to combine them without jargon or invented statistics.
What is alt text, exactly?
Alt text is the value of the HTML alt attribute on an <img> tag. It's text you write for each image, on each page where it appears: <img src="cat.jpg" alt="Ginger cat asleep on a grey blanket">. It was designed for two specific reasons: it lets screen readers describe the image to blind and low-vision users — this is a requirement under WCAG, the W3C's web accessibility standard — and it acts as fallback text if the image fails to load, whether from a slow connection, a broken link, or a text-only browser.
The W3C's Web Accessibility Initiative publishes a reference tutorial on how to write alt text depending on the image type — informative, decorative, functional, or complex — and it's the most reliable source on the topic, which we lean on throughout this guide: w3.org/WAI/tutorials/images.
- Read aloud by screen readers (VoiceOver, NVDA, JAWS).
- Displays in place of the image if it fails to load.
- Gives search engines textual context about the image.
- Is NOT stored inside the image file — it lives only in the page's HTML.
Why the alt attribute exists
The alt attribute has existed since HTML 2.0 and answers WCAG success criterion 1.1.1, "Non-text Content": all non-text content needs an equivalent text alternative. Without it, a blind screen reader user simply hears "image" with no idea what it shows — a product, a chart, a face, a button.
Alt text and embedded IPTC/XMP metadata both answer the question "what does this image show?" — but one lives in the page, the other lives in the file. Confusing the two, or only doing one, leaves gaps in accessibility, in SEO, or in both.
Where do alt text and metadata actually live?
This is the single most common mix-up in image SEO: assuming alt text is part of the file, or that EXIF/IPTC metadata is visible on the page. It's neither.
Alt text — lives in the page's HTML
alt="White sneaker shown side-on">
Written once, on one specific page. If the same image appears on 5 pages, or on another site after being downloaded, the alt attribute has to be rewritten every time — it does not travel with the file.
IPTC/XMP — lives inside the file's bytes
IPTC:Keywords: sneaker, footwear, product
XMP:dc:subject: sneaker, footwear
Written once into the file with a tool like Exif Injector, and embedded in the image's own bytes. Download the file, email it, reuse it on a different site: those fields are still there — unless the receiving platform strips them.
IPTC IIM is the legacy standard (Description, Caption-Abstract, Keywords); XMP is the modern XML-based container that carries largely the same fields under different names (dc:description, dc:subject). Most tools, including Exif Injector, write both at once for maximum compatibility — see the official standard: iptc.org/standards/photo-metadata.
Why relying on alt text alone is fragile
Alt text only exists where you typed it: in the HTML of one specific page. The moment the image leaves that page, the alt text does not go with it.
Image reused elsewhere
The same product photo published on a second page, in a marketing email, or in a product feed doesn't automatically inherit the alt text written elsewhere — it has to be re-entered.
File downloaded or shared
A visitor downloads the image, a partner pulls it for a catalog, a journalist reuses it in an article: the file travels alone, without your page's HTML — the alt text stays behind.
Republished or scraped
On an aggregator, a reverse-image search engine, or a site that hotlinks the file directly, the new context has no access to your alt text — unless another source of information exists, such as the file's own metadata.
Embedded metadata isn't magic either: many platforms — social networks, some CMSs, some CDNs — strip some or all EXIF/IPTC data on upload, for privacy or to shrink file size. So the right practice isn't "metadata instead of alt text" — it's both together, then checking with an EXIF viewer what actually survives on each target platform.
Alt text vs IPTC/XMP metadata
| Attribute | Alt text | IPTC/XMP metadata |
|---|---|---|
| Where it lives | In each page's HTML | In the image file's own bytes |
| Who reads it | Screen readers, browsers, page crawlers | Photo software, DAM systems, some image search engines |
| Survives download / re-upload | No — stays on the original page | Often yes, depending on the platform |
| Required for accessibility (WCAG) | Yes — success criterion 1.1.1 | No — not read by browser screen readers |
| Bulk-editable across a photo library | Page by page (or via a CMS/product field) | Yes — Exif Injector in bulk across a whole folder |
| Read by Google Images | Yes — a documented direct signal | Can be used as a supplementary signal, per Google |
| Depends on page context | Yes — can/should vary per page | No — describes the image itself, not the page |
Source: W3C WAI — tutoriel images · Google Search Central — Images · IPTC Photo Metadata Standard
How to write good alt text
- Be concise and specific. Describe what matters in the image, not every detail. A commonly used guideline is to stay under roughly 125 characters, since some screen readers truncate longer descriptions.
- Skip "image of" or "picture of". The screen reader already announces it's an image — repeating that is redundant.
- Include a relevant keyword, naturally. Only if it's genuinely part of an honest description of the image — never by stuffing it in.
- Decorative images: empty alt. Use alt="" for icons or purely aesthetic visuals so screen readers skip them entirely.
- Describe function for clickable images. If the image is a link or button, alt text should say where it leads or what it triggers — not just what it depicts.
alt="image1.jpg"
alt="image of shoe"
alt="sneaker shoe running trainer cheap buy now sale"
alt="White running sneaker with thick sole, shown side-on"
alt=""
A visual divider or an icon repeated next to already-readable text doesn't need a description — an empty alt keeps screen readers from reading out noise.
How the fields work together
In Google's own image SEO documentation, several signals matter for understanding an image: alt text, the surrounding page text, the filename — and metadata embedded in the file. Nothing suggests any single signal is enough on its own; they complement each other.
| Field | Where it lives | Primary role |
|---|---|---|
| alt (HTML) | Page HTML | Accessibility (WCAG) + fallback if image fails to load |
| IPTC Description / Caption-Abstract | Image file | Portable description, read by DAM systems and some image search engines |
| XMP dc:description | Image file | Modern XML mirror of the IPTC field above |
| IPTC Keywords / XMP dc:subject | Image file | Subject terms that travel with the file to DAM systems, stock libraries, some search engines |
Official source: developers.google.com/search/docs/appearance/google-images
A realistic workflow: bulk metadata, per-page alt text
The idea isn't to pick one over the other, but to handle each at the scale that fits it: metadata is a file-level property, so handle it in bulk across a whole image library. Alt text depends on page context, so it stays written page by page.
Drop a whole product or editorial library — the AI suggests a title, description and keywords for each image, which you can adjust.
Exif Injector embeds those fields directly into every file — the metadata now travels with the image, even outside your site.
On every page where the image appears, write or generate an alt attribute suited to that specific page's context — not a copy-paste of the file's description.
Re-import the published file into the EXIF Viewer to confirm the target platform didn't strip the IPTC/XMP fields you just wrote.
Tools used in this workflow
Embedded metadata protects your attribution and description even when the image circulates outside your control. Alt text remains the one piece that's mandatory for accessibility, page by page. Neither one substitutes for the other.
Practical checklist: do / don't
Do
- Write different alt text if the same image serves a different purpose on another page.
- Embed Title/Description/Keywords into the file before publishing, in bulk.
- Use alt="" for purely decorative images.
- Verify with the EXIF Viewer what survives upload on each platform.
Don't
- Leave the alt attribute empty on an informative image — that's an accessibility gap.
- Stuff keywords into alt text ("sneaker shoe sport running cheap buy").
- Copy the filename into the alt attribute ("IMG_4821.jpg").
- Assume embedded metadata will automatically survive every platform without checking.
FAQ about alt text and image SEO
Is alt text the same thing as an image title or filename?+—
No. Alt text is the HTML alt attribute read by screen readers and shown if an image fails to load. The filename and the HTML title attribute are separate, and neither reliably reaches assistive technology the way alt text does.
Does Google Images read embedded IPTC or XMP metadata?+—
Google's own image SEO documentation says Google can use metadata embedded in an image file, such as IPTC descriptions and keywords, as one of several signals for understanding image content, alongside the surrounding page text and alt text.
Should alt text include keywords?+—
Only if the keyword is genuinely part of an accurate description. Alt text is written for a person using a screen reader first; keyword stuffing hurts accessibility and can be flagged as manipulative by search engines.
What alt text should I use for a purely decorative image?+—
Use an empty alt attribute, alt="". This tells screen readers to skip the image entirely, which is the WCAG-recommended pattern for icons and background flourishes that carry no information.
If I embed IPTC metadata with Exif Injector, do I still need alt text?+—
Yes. Embedded metadata travels with the file but is not reliably read by screen readers in the browser. Alt text is still required on every page where the image appears to meet WCAG accessibility requirements.
Do platforms like Instagram or Shopify keep embedded metadata when I upload images?+—
It varies by platform. Many social networks strip most metadata, including GPS, on upload for privacy and file-size reasons, while some ecommerce and CMS platforms preserve more of the IPTC and XMP fields. Always verify with an EXIF viewer after upload.
How long should alt text be?+—
There's no hard limit, but a commonly used guideline is to keep it under roughly 125 characters, since some screen readers truncate longer descriptions. Prioritize accuracy over hitting a specific length.
Metadata that travels, alt text that welcomes
Embed titles, descriptions and keywords in bulk with Exif Injector, then generate page-appropriate alt text with the Alt Text Generator.
Sources: W3C WAI · Google Search Central · IPTC.org