Menu
In July 2026 we crawled the 10,000 highest-ranked domains on the public Tranco popularity ranking and analyzed every image on each reachable homepage — 306,634 images in total. One in nine has no alt attribute at all: invisible to screen readers, mute to the AI engines that now read the same field. Here is the full data.
■ Non-empty alt text (207,687) ■ alt="" — deliberately marked decorative, valid markup (64,263) ■ No alt attribute at all — the defect class (34,684)
Alt text has been an accessibility requirement since the 1990s. What changed recently is who reads it: AI assistants and AI-augmented search engines parse image descriptions when deciding what a page contains, which products match a query, and which sources to cite. An image without an alt attribute is skipped twice — once by assistive technology, once by the machines that increasingly route traffic.
We wanted to know how the web’s most successful sites are doing on the one field both audiences depend on. So we took the public Tranco ranking of the world’s most popular domains and audited the top 10,000.
11.3% of images carry no alt attribute at all. That is 34,684 images across the web’s front pages. We deliberately did not count alt="" against a site: an empty alt attribute is valid markup that tells a screen reader an image is decorative, and 21.0% of images use it. Only the complete absence of the attribute — which leaves assistive technology guessing and AI engines with nothing to read — was counted as a defect.
This is not a fringe problem: 37.1% of the sites we could audit ship at least one attribute-less image on their homepage. These are among the most popular sites on the internet, most of them run by organizations with dedicated engineering teams.
| Per-site descriptive-alt coverage (5,306 sites) | Share of sites |
|---|---|
| Every homepage image described (100%) | 24.3% |
| Median site | 77.1% of images described |
| Less than half of images described | 30.6% |
| No images described at all (0%) | 8.9% |
The distribution is the story: a quarter of the web’s top sites treat image descriptions as table stakes and cover everything. Nearly a third describe fewer than half their images. And almost one site in eleven — at the very top of the web — describes none.
image-alt-study/1.0. No further crawling beyond the homepage.<img> tag. All image statistics use those 5,306 sites. The remainder are CDN and infrastructure domains, bot-blocked sites, or non-HTML endpoints — they are excluded, never counted as failures.<img> tag. Three classes: non-empty alt, empty alt="" (valid, decorative), attribute absent (defect). JavaScript was not executed, so images injected client-side are not counted — coverage of the modern JS-heavy web is therefore conservative on total volume.Screen-reader users have paid the cost of missing alt text for thirty years. The new development is commercial: AI search engines and assistants read alt text when deciding what to recommend and cite. The gap this study measures is no longer only an accessibility debt — it is a visibility debt, sitting on the front pages of the world’s biggest sites.
We are not the first to measure missing alt text, and the established numbers deserve to be named. The WebAIM Million — the long-running accessibility census of the top one million home pages — puts missing alternative text among the most common failures on the web, with roughly 16.2% of home-page images lacking it in its 2026 edition. HTTP Archive’s accessibility analysis of many millions of pages lands in a similar band, around 13%.
| Study | Sample | Images missing alt |
|---|---|---|
| WebAIM Million 2026 | Top 1,000,000 home pages | 16.2% |
| HTTP Archive 2025 | ~17M pages | ≈13% |
| This study (July 2026) | Top 10,000 homepages | 11.3% |
Our figure is the lowest of the three — because our sample is the hardest. We deliberately audited only the elite: the ten thousand highest-ranked domains on Earth, most of them run by organizations with dedicated engineering and accessibility budgets. If any slice of the web should have this solved, it is this one. It does not: even at the very top, one image in nine is still shipped with nothing for a screen reader or an AI engine to read. The three studies together tell one consistent story — coverage improves with site quality, and the problem still does not go away.
A reasonable assumption: the bigger the site, the better the alt-text discipline. The data says otherwise. Splitting the audited sites by Tranco rank (computable by anyone from the published CSV):
| Rank tier | Sites | ≥1 image missing alt | Every image described | Median coverage |
|---|---|---|---|---|
| Top 1–100 | 47 | 31.9% | 14.9% | 50.0% |
| 101–1,000 | 414 | 34.1% | 25.1% | 71.4% |
| 1,001–5,000 | 2,058 | 35.5% | 23.8% | 74.7% |
| 5,001–10,000 | 2,787 | 38.7% | 24.8% | 80.9% |
The very top of the web is, by two of three measures, the worst tier: the median top-100 homepage describes only half its images, and fewer than one in six describes them all. The likely mechanics: the biggest pages are also the most complex — dozens of templates, components and teams shipping images into one homepage — and coverage decays wherever no single owner is accountable for it. Alt text is not a budget problem. It is an ownership problem.
If you run a WordPress or WooCommerce site: your media library almost certainly follows the distribution above. Start with an inventory — the free image SEO audit crawls any URL and counts described, decorative and attribute-less images the same way this study does. For the backlog itself, bulk generation is the only realistic route past a few hundred images: the ImageSEO plugin writes descriptive alt text for an entire library and keeps writing it for new uploads, so the gap does not reopen.
If you build sites: the defect class in this study — the attribute missing entirely — is almost always a template bug, not a content failure. Audit the components that render images: hero blocks, lazy-loading wrappers, icon systems and CMS field mappings that silently drop alt. Make the attribute required at the template level: render alt="" explicitly for decorative images and fail the build when the attribute is absent. A one-line linter rule prevents the exact gap this study measures.
If you write the descriptions: describe what the image shows in one natural sentence — the subject, the action, the context that matters on that page. Skip “image of” and keyword lists; screen readers already announce an image, and stuffed alt text reads as noise to the people who depend on it. Keep decorative flourishes at alt="" so they stay silent. Around 80–125 characters fits most screen-reader habits, but a clear description that runs longer beats a vague one that fits.
Only the complete absence of the alt attribute on an <img> tag. An empty alt="" is valid HTML that marks an image as decorative, so the 21.0% of images using it were counted separately — never as defects. That makes 11.3% a deliberately conservative floor.
Different sample, same direction. WebAIM audits the top one million home pages; we audited only the top ten thousand — the web’s best-funded slice, where coverage is predictably better. Methodologies also differ in what they count. The finding both studies share: a meaningful share of images at every level of the web still ships with nothing for assistive technology to read.
alt="" bad for accessibility or SEO?No — for genuinely decorative images it is the correct markup. It tells screen readers to skip the image silently instead of announcing a filename. It becomes a problem only when it is applied to images that carry meaning, which is a content decision no automated count can fully judge — one reason this study counts it separately.
Yes. Assistants and AI-augmented search engines parse the text fields of a page — alt text included — when deciding what the page contains, which products match a query and which sources to cite. An image with no attribute contributes nothing to that reading. That is the shift that turns a thirty-year-old accessibility requirement into a visibility requirement.
Run the free image SEO audit on any URL — it classifies every image on the page into the same three classes used here. For a single image, the free alt-text generator writes a description with no signup.
Yes — the dataset is licensed CC BY 4.0. Download the per-domain CSV, reuse it with attribution, and link this page as the source. The underlying Tranco list ZJG2G is versioned, so the crawl is independently reproducible.
Check your own site in minutes.
Run the free image SEO audit on any URL, or let the ImageSEO WordPress plugin write the missing descriptions — it has generated 55,000+ AI image descriptions for 9,500+ websites.
Cite this study: ImageSEO, “The Alt Text Gap” (July 2026). Please link to this page; the dataset is free to reuse with attribution. Questions: contact@imageseo.io.
