How to write better alt text in seconds with AI

If you've ever run an accessibility audit on your site, this probably looked familiar:

47 images missing alt attributes.
12 images with non-descriptive alt text ("image", "photo1", "DSC_0042.jpg").
8 images with keyword-stuffed alt text.

Alt text is the most-flagged accessibility issue in the world, and one of the most ignored SEO opportunities. Not because it's hard — it's just one extra attribute on your <img> tag — but because writing it for hundreds of images is genuinely tedious. So most teams skip it, batch it for "later," or hand it off to whoever loses at rock-paper-scissors.

This guide covers what alt text actually does, what makes a good one, the mistakes everyone make, and a free tool we built so you can clear the backlog in an afternoon.

What alt text does (and why it's two jobs)

The alt attribute on an <img> element does two completely different things at once.

Job 1 — accessibility. Screen readers (used by people who are blind or have low vision) read the alt text aloud as they navigate a page. If there's no alt text, the screen reader either says "image" or, worse, reads the file path. Imagine browsing your favorite blog and hearing "image. image. image. image." between every paragraph.

Job 2 — SEO. Google can't see your images. Image search, snippets, and rich results all depend on machine-readable signals: the surrounding text, the file name, captions, structured data — and the alt attribute. Good alt text is one of the cheapest SEO wins available, and it's the only one that also makes your site usable by people who can't see.

WCAG 2.2 (the global accessibility standard) and the ADA both require alt text on informative images. If your business operates in the EU, the European Accessibility Act adds teeth to that — non-compliant sites face real fines starting in 2025.

What good alt text looks like

A few quick examples for a coffee-beans photo, three versions of alt text:

• ❌ Bad: alt="beans.jpg" — the file name. Screen reader says "beans.jpg." Useless.
• ❌ Bad: alt="best coffee beans, organic, fair trade, premium coffee, single origin" — keyword stuffing. Google penalizes this; screen reader users get spammed.
• ✅ Good: alt="Top-down view of roasted coffee beans scattered on a wooden surface" — describes the image concisely, uses natural language, and happens to include relevant keywords without forcing them.

The rules of thumb:

1. Describe what's in the image, not what you wish was there. "Roasted coffee beans spilling from a wooden scoop" — not "rich morning brew."
2. Lead with the most important subject. Most screen readers cut off after ~125 characters. Front-load the meaning.
3. Skip the boilerplate. Don't start with "image of" or "photo of" — screen readers already announce that it's an image.
4. Don't keyword stuff. One natural mention of the relevant term is plenty.
5. Decorative images get empty alt. If an image is purely visual flair (a divider line, a background flourish), use alt="" (not "decoration"). That tells screen readers to skip it.
6. Charts and infographics need a longer description. The alt should summarize the takeaway, and the page should include the underlying data nearby.

A good alt text length is around 100 characters. Long enough to be informative, short enough that screen readers don't trail off.

The five mistakes everyone makes

After looking at thousands of sites, the failure patterns are consistent:

1. No alt text at all. The biggest single category. Most CMSes don't enforce it; designers paste images straight in.
2. Filename as alt text. "DSC_0042.jpg" or "Screen Shot 2026-04-12 at 3.42.18 PM.png." The CMS auto-fills it. Nobody fixes it.
3. Generic placeholders. "image", "photo", "picture", "graphic" — technically populated, totally useless.
4. Keyword stuffing. Treating alt as an SEO field instead of an accessibility one. Google's been smart enough to detect and penalize this for years.
5. Redundant captions. When a caption already describes the image, the alt text duplicates it word-for-word. Better: alt text that complements the caption with a different angle.

The combined effect: about 30% of web images have no alt text and another 30% have meaningless alt text. That's a huge amount of unsearchable content, plus a usability barrier for ~3% of your audience.

Where AI helps and where it doesn't

A vision-capable AI model — like Claude or GPT-4 — can look at an image and write a descriptive caption that's pretty close to good alt text out of the box. That solves three of the five problems above (no alt, filename alt, generic placeholders) instantly.

Where AI still needs your eyes:

• Brand-specific or product-specific naming. AI doesn't know that the woman in the photo is your CEO, or that the chair in the lifestyle shot is your "Atlas Lounge — Walnut Edition."
• Editorial context. A photo of a sad dog in a piece about animal shelters needs different alt text than the same photo in a piece about nutrition.
• Charts and data visuals. AI describes what the chart looks like, but you usually want the takeaway: "Customer churn dropped 18% after the onboarding rewrite."
• Cultural and legal sensitivity. A model trained on broad data may write something tone-deaf in context.

The right workflow: AI writes the first draft, you review and edit. That cuts a 30-second-per-image task to 3 seconds.

How to fix a backlog of hundreds of images

If you have a few decades of un-tagged images sitting in WordPress, Shopify, or whatever CMS, the manual route is impossible. You need batching.

We built a free AI alt text generator that takes that part off your plate:

Here's how it works:

1. Drop in up to 25 images at a time (JPG, PNG, WebP, GIF, AVIF).
2. The tool runs each one through Claude Haiku — a vision-capable AI model — and writes concise alt text targeted at the 100-character sweet spot.
3. You review and edit the generated text inline.
4. Export the whole batch as a CSV. Paste it back into your CMS.

It runs in your browser, no signup needed to try it. The images aren't stored — they go to the AI service, get described, and are discarded. The only thing we ask for an email is so you can export the full CSV (we don't keep your images either way).

For an entire site backlog, plan on doing it in passes — 25 images at a time, ~30 seconds per pass — until you're caught up. A 500-image backlog takes about half an hour of focused review. Compare to the 4+ hours it would take to write each one manually.

Beyond the backlog: keep alt text in your workflow

Catching up is good. Not falling behind again is better. Two habits that work:

1. Make alt text mandatory at upload time. Most modern CMSes (WordPress, Webflow, Shopify) let you require alt text before publishing. Turn it on. Yes, your team will complain. They'll also stop shipping un-tagged images.
2. Add alt text to your design review checklist. Before any new page or campaign goes live, run through the images and confirm each has meaningful alt text.

If your team uploads a lot of images, Asset Locker can auto-suggest alt text at upload using the same AI vision model behind this tool — so you start every image with a good draft instead of a blank field. Your team's alt-text quality stops depending on whoever's least bored.

TL;DR

• Alt text is required for accessibility and useful for SEO. Two jobs, one attribute.
• Good alt text is concise, descriptive, leads with the main subject, ~100 characters, no "image of" prefix, no keyword stuffing.
• 60% of web images have missing or junk alt text. The blocker is volume, not difficulty.
• AI can do a solid first draft for most images in seconds. Use the free AI alt text generator to clear a backlog.
• Keep it from coming back: enforce alt text at upload, and review it as part of design QA.

Now go fix those 47 images.