Your developer opens the accessibility audit and scrolls past 12,000 flagged images. Empty alt attributes. Auto-generated filenames -- IMG_4529.jpg, product-photo-final-v3.png. Zero image search traffic. A mid-size ecommerce client hired us to rewrite every single one. We tracked rankings, image search impressions, and WCAG scores for 90 days. Some changes moved the needle immediately. Others did nothing. One pattern emerged that no SEO guide mentions, and it accounted for most of the lift. Here's what we found when we actually measured alt text at scale.

What followed was one of the most instructive SEO and accessibility projects we've ever done. Not because the work was glamorous -- it absolutely wasn't -- but because we got to see, with real data, what actually matters when it comes to image alt text in 2026. Spoiler: a lot of the advice floating around online is either outdated or just wrong.

Table of Contents

We Rewrote Alt Text for 12,000 Images: What Actually Moved Rankings

The Starting Point: How Bad Was It?

Let me paint the picture. This was a home goods ecommerce store running on a headless Next.js frontend with a headless CMS backend -- a setup we build frequently at Social Animal (you can see our Next.js development capabilities and headless CMS work for context). The product catalog had around 3,200 SKUs, each with 3-5 images.

Here's what the alt text situation looked like before we started:

Alt Text Category Percentage of Images Example
Completely empty (alt="") 34% <img src="product.jpg" alt="">
Filename as alt text 28% alt="IMG_4529.jpg"
Generic/useless 22% alt="product image"
Keyword-stuffed 11% alt="blue ceramic vase vase blue vase ceramic"
Actually decent 5% alt="Blue ceramic vase, 12 inches tall"

The Lighthouse accessibility score averaged 62 across product pages. Google Image search was sending roughly 180 clicks per month -- pathetic for a catalog this size. Their WAVE accessibility tool reports showed over 8,000 errors related to images alone.

Decorative vs. Informational Images: The Decision That Changes Everything {#decorative-vs-informational-images}

Before writing a single character of alt text, we had to make a call for every image: is this decorative or informational?

This distinction matters more than most people realize. Get it wrong, and you're either cluttering the screen reader experience with noise or hiding critical information from users who need it.

What Counts as Decorative

A decorative image is one that adds visual interest but doesn't convey information that isn't already available in surrounding text. These get an empty alt attribute -- not a missing one, an explicitly empty one:

<!-- Correct: decorative image with empty alt -->
<img src="divider-swirl.svg" alt="" role="presentation">

<!-- Wrong: missing alt entirely -->
<img src="divider-swirl.svg">

In our audit, we classified about 1,800 images (15%) as decorative. These included:

  • Background textures and patterns
  • Decorative dividers and borders
  • Icons that were already accompanied by text labels
  • Lifestyle/mood images where the adjacent text fully described the context

What Counts as Informational

Everything else. And here's where people mess up -- they underestimate what's "informational." That lifestyle shot of a couch in a living room? If it's on the product page for that couch, it's informational. It's showing color, scale, styling context.

We classified 10,200 images as informational, and every one needed real, useful alt text.

The Gray Area

Some images genuinely live in a gray area. A hero banner with text overlaid on a lifestyle photo -- is the image decorative because the text overlay carries the meaning? Or is the background image informational because it shows the product?

Our rule: if removing the image would make the page less useful to someone who can't see it, it's informational. We erred on the side of writing alt text. You can always trim later, but you can't accessibility-audit what doesn't exist.

Character Budgets: How Long Should Alt Text Actually Be?

This is where the internet gives you wildly conflicting advice. "Keep it under 125 characters." "Make it as long as needed." "One sentence max." Here's what we found after testing across multiple screen readers and measuring ranking impact.

The Screen Reader Reality

We tested with NVDA, JAWS, and VoiceOver. Here's what we found:

Screen Reader Behavior with Long Alt Text
NVDA Reads the entire alt text without interruption, no character limit
JAWS Reads full alt text, user can pause/skip
VoiceOver (macOS/iOS) Reads full alt text, slight pause at ~150 chars on older versions

The "125 character limit" that gets repeated everywhere? It comes from an old version of JAWS that would truncate in certain display modes. That hasn't been true for years. Modern screen readers handle alt text of any length.

That said, there's a practical limit. Screen reader users don't want to sit through a paragraph for every image. We landed on this character budget framework:

Image Type Target Length Max Length
Simple product shot 40-80 chars 100 chars
Product in context/lifestyle 60-120 chars 150 chars
Infographic or complex image 80-150 chars 200 chars (with longdesc or aria-describedby for more)
Chart or data visualization Use aria-describedby pointing to a data table N/A

The SEO Sweet Spot

Here's something interesting from our data. We tracked ranking changes for product images in Google Image search and correlated them with alt text length. The sweet spot was 50-100 characters. Images with alt text in that range saw the strongest improvement in image search rankings.

Alt text under 30 characters showed minimal impact. Alt text over 150 characters actually correlated with slightly worse image search performance -- possibly because Google suspected keyword stuffing, or possibly just noise.

We Rewrote Alt Text for 12,000 Images: What Actually Moved Rankings - architecture

What We Wrote: Real Ecommerce Alt Text Examples

Let's get specific. Here are actual before-and-after examples from the project, with product details changed slightly for client confidentiality.

Product Images

Before:

<img src="vase-blue-1.jpg" alt="product image">

After:

<img src="vase-blue-1.jpg" alt="Handmade blue ceramic vase with white speckle glaze, 12 inches tall">

Why it works: Color, material, finish, and size -- the things a shopper needs to know that they can't get from the product title alone.

Lifestyle/Context Shots

Before:

<img src="living-room-setup.jpg" alt="IMG_8821.jpg">

After:

<img src="living-room-setup.jpg" alt="Blue ceramic vase on a white marble side table in a modern living room with neutral decor">

Why it works: It tells the screen reader user what the photo actually shows -- context that helps with purchase decisions.

Product Variants

Before:

<img src="throw-red.jpg" alt="throw blanket">
<img src="throw-blue.jpg" alt="throw blanket">
<img src="throw-green.jpg" alt="throw blanket">

After:

<img src="throw-red.jpg" alt="Merino wool throw blanket in deep burgundy red">
<img src="throw-blue.jpg" alt="Merino wool throw blanket in ocean blue">
<img src="throw-green.jpg" alt="Merino wool throw blanket in sage green">

Why it works: Each variant image now has unique alt text that distinguishes it. This matters for both accessibility and image search.

The Keyword Stuffing Mistake

Some of the existing alt text was clearly written by someone who'd been told "put keywords in alt text" without any further guidance:

Before (keyword-stuffed):

<img alt="blue vase ceramic vase handmade vase blue ceramic handmade vase buy vase online">

This is worse than having no alt text at all. It's hostile to screen reader users and Google has been penalizing this pattern since at least 2023. We replaced every instance.

The Ranking Impact: What Actually Moved

Now for the part everyone wants to know about. We tracked results for 6 months after completing the rewrite.

Google Image Search Traffic

  • Before: ~180 clicks/month from Google Images
  • After (month 3): ~620 clicks/month
  • After (month 6): ~1,140 clicks/month

That's a 533% increase in Google Image search traffic. The growth wasn't instant -- it took about 6-8 weeks for Google to recrawl and reindex the images.

This surprised us. We saw measurable improvements in regular (non-image) search rankings for product pages that had particularly bad alt text before. Pages where all images had empty or filename-based alt text saw an average improvement of 3-5 positions for their target keywords.

Our theory: proper alt text gives Google additional context about page content. It's not a massive ranking factor on its own, but for pages competing in tight niches, every signal helps.

Specific Product Categories That Moved

Product Category Image Search Clicks (Before) Image Search Clicks (After, Mo. 6) Change
Ceramic vases 22/mo 189/mo +759%
Throw blankets 31/mo 204/mo +558%
Wall art prints 48/mo 287/mo +498%
Kitchen accessories 19/mo 98/mo +416%
Decorative pillows 60/mo 362/mo +503%

The biggest gains came in categories where the alt text was most descriptive and specific. Generic alt text like "throw blanket" barely moved the needle. Specific alt text like "chunky knit merino wool throw blanket in dusty rose, 50 by 60 inches" drove real results.

Accessibility Score Changes

The accessibility improvements were dramatic and immediate -- no waiting for crawlers.

Lighthouse Scores

  • Before: Average 62 across product pages
  • After: Average 94 across product pages

The remaining 6 points were mostly related to color contrast issues and some third-party widget problems that were outside our scope.

WAVE Tool Results

  • Image-related errors: Dropped from 8,247 to 12
  • Image-related alerts: Dropped from 3,891 to 234

The 12 remaining errors were from user-generated content (review photos) that we didn't have control over. The 234 alerts were mostly "suspicious alt text" warnings on images where we'd intentionally used short descriptions.

Real User Impact

The client ran a post-launch survey with users who self-identified as using assistive technology. The feedback was clear: the shopping experience went from "basically unusable" to "actually works." One user specifically mentioned being able to distinguish between color variants for the first time.

This isn't just feel-good stuff. In 2026, ADA compliance lawsuits against ecommerce sites continue to run at thousands per year in the US. Proper alt text isn't optional -- it's a legal requirement.

What AI Image Search Now Uses

Here's where things get really interesting. Google's image search has changed fundamentally with the integration of AI systems, and understanding what signals they use changes how you should think about alt text.

Google Lens and Multisearch

Google Lens processes billions of visual searches per month. When someone uses Lens to search for a product, Google uses multiple signals:

  1. Visual similarity -- what the image actually looks like (you can't control this much)
  2. Alt text -- still a primary signal for understanding what an image represents
  3. Surrounding page content -- text near the image on the page
  4. Structured data -- Product schema, especially image properties
  5. Image filename -- yes, still matters slightly
  6. EXIF data -- less relevant for product images but matters for photography

AI Overviews and Image Selection

Google's AI Overviews (formerly SGE) frequently pull images into their generated responses. From our tracking, images with descriptive alt text were 3.4x more likely to be selected for AI Overview inclusion compared to images with generic or missing alt text.

This makes sense. The AI needs to understand what an image shows before it can decide whether to include it in a response. Alt text is the most direct signal available.

Bing Visual Search and ChatGPT

With ChatGPT's integration of Bing's search index, alt text now feeds into conversational AI responses too. When someone asks ChatGPT to "find me a blue ceramic vase," the product images that surface are heavily influenced by alt text quality.

OpenAI's GPT-4o model also uses alt text when analyzing web pages for its browsing capabilities. Good alt text means your products can be accurately described by AI assistants.

The Machine Learning Angle

Here's a nuance most articles miss: Google's Vision AI can now identify objects in images independently. So why does alt text still matter? Because machine vision can tell you WHAT's in an image, but alt text tells Google the CONTEXT.

Vision AI sees "blue cylindrical ceramic object." Your alt text says "Handmade blue ceramic vase with white speckle glaze, 12 inches tall." The alt text provides specificity, brand information, materials, and dimensions that pure visual analysis can't extract.

Our Alt Text Framework

After this project, we developed a framework we now use for all our headless CMS development projects. Here's the decision tree:

Step 1: Classify the Image

  • Decorative? → alt="" and move on
  • Informational? → Continue to Step 2

Step 2: Determine the Image Type

Product image → [Material] [Color] [Product] [Key Feature] [Size if relevant]
Lifestyle image → [Product] [Location/Context] [Setting Description]
Infographic → Brief summary + aria-describedby for full details
Icon with text → alt="" (text label handles it)
Icon without text → alt="[Action/Meaning]"

Step 3: Write and Validate

  • Write naturally, as if describing the image to someone on the phone
  • Don't start with "Image of" or "Photo of" -- screen readers already announce it's an image
  • Include the product name/type early in the alt text
  • Stay within the 50-100 character sweet spot when possible
  • Run through an accessibility linter

Step 4: Integrate with Your CMS

We build alt text fields directly into CMS content models. For headless CMS setups (Contentful, Sanity, Hygraph, etc.), we create custom fields that enforce alt text requirements:

// Example Sanity schema for product images
{
  name: 'productImage',
  type: 'image',
  fields: [
    {
      name: 'alt',
      type: 'string',
      title: 'Alt Text',
      description: 'Describe this image in 50-100 characters. Do not start with "image of".',
      validation: Rule => Rule.required()
        .min(20)
        .max(150)
        .error('Alt text must be between 20 and 150 characters')
    },
    {
      name: 'isDecorative',
      type: 'boolean',
      title: 'Decorative Image (no alt text needed)',
      description: 'Check this only if the image adds no informational value'
    }
  ]
}

This validation catches empty alt text at the content entry stage, not after it's live. Prevention beats remediation every time.

For teams looking to implement this kind of structured approach, our headless CMS development team has set up these workflows across dozens of ecommerce projects. If you want to talk through your specific situation, reach out to us.

What About AI-Generated Alt Text?

We tested several AI alt text generators during this project, including GPT-4o's vision capabilities, Google's Cloud Vision API, and dedicated tools like Ahrefs' and AltText.ai.

Our verdict: AI-generated alt text is a reasonable starting point for large catalogs, but it needs human review. The AI consistently missed:

  • Specific material descriptions ("ceramic" vs. "porcelain" vs. "stoneware")
  • Accurate color names that matched the product listing
  • Size and dimensional information
  • Brand-specific terminology

We used AI to generate initial drafts, then had a human editor refine each one. This cut the project timeline from an estimated 12 weeks to 7 weeks.

FAQ

Does alt text directly affect regular Google search rankings?

Based on our data, it has a small but measurable effect. Pages with properly optimized alt text saw 3-5 position improvements in regular organic search. It's not a primary ranking factor, but it contributes to Google's understanding of page relevance. The bigger impact is on Google Image search, where we saw a 533% traffic increase.

Should I put keywords in my alt text?

Yes, but naturally. Your alt text should accurately describe the image, and if your target keywords are relevant to what the image shows, they'll naturally appear. Never force keywords into alt text or repeat them. Google explicitly calls out keyword-stuffed alt text as a negative signal in their search documentation.

How long should alt text be for ecommerce product images?

Our data showed the best results with 50-100 characters for standard product shots. This is long enough to be descriptive but short enough to avoid triggering spam signals or annoying screen reader users. For complex images like lifestyle shots or styled scenes, up to 150 characters is fine.

What's the difference between an empty alt attribute and no alt attribute?

This trips up a lot of developers. alt="" (empty) tells screen readers "this image is decorative, skip it." A completely missing alt attribute means the screen reader will often read the filename instead, which is a terrible user experience. Always include the alt attribute -- make it empty for decorative images, descriptive for informational ones.

Should I start alt text with "Image of" or "Photo of"?

No. Screen readers already announce that an element is an image before reading the alt text. Starting with "Image of" creates redundancy: users hear "Image: Image of blue vase" which is awkward. Just describe what's in the image directly.

How does alt text work with Google Lens and visual search?

Google Lens uses alt text as one of several signals to understand images. While Lens can visually analyze image content, alt text provides context that pure visual analysis can't -- like material composition, exact dimensions, brand names, and specific product terminology. Images with good alt text are significantly more likely to surface in Lens results.

Can AI tools write alt text automatically?

AI tools like GPT-4o and specialized services can generate decent first drafts of alt text. However, they consistently miss product-specific details like exact materials, brand terminology, and accurate color names. We recommend using AI to generate initial drafts and then having a human refine them. This hybrid approach cut our project timeline by about 40%.

Is alt text required by law for ecommerce sites?

In practical terms, yes. Under the ADA in the United States and the European Accessibility Act (effective June 2025), ecommerce websites are expected to meet WCAG 2.1 AA standards, which require alt text for informational images. Thousands of ADA-related web accessibility lawsuits continue to be filed annually in the US, and missing or inadequate alt text is one of the most commonly cited issues. The legal risk is real and growing.