Alt text is the written description attached to an image's HTML code that tells screen readers, search crawlers, and AI answer engines what an image shows and why it's there. Get it right and a blind user understands your product photo, Google can surface your image in search results, and an AI Overview or ChatGPT answer can cite what your page shows without ever rendering it. Get it wrong, skip it, or stuff it with keywords, and all three audiences lose something. This guide covers what alt text actually is, how to write it well for a person and a machine at the same time, and what's changed about it now that AI systems read images differently than search engines always have.
What Is Alt Text?
Alt text, short for "alternative text," is a line of code inside an image's <img> tag that describes the image in words. It lives in the alt attribute, so a photo of a beagle in a snowy field might carry the markup alt="Beagle standing in a frosty field on a cold morning". A site visitor never sees that text rendered on the page. A screen reader speaks it aloud. A browser displays it in place of the image if the file fails to load. A search engine or AI crawler reads it as the clearest signal available for what the image contains, because none of them can look at a picture the way a person can.
Alt text is also called an "alt attribute" or "alt tag," and all three terms describe the same thing. The distinction worth keeping straight is between alt text and a caption. A caption is the line of visible text underneath an image, usually a photo credit or a clarifying note. Alt text is invisible to a sighted visitor and lives only in the code, built for a different audience: assistive technology and machines, not the person already looking at the picture.
Why Alt Text Still Matters in 2026
Alt text does three separate jobs at once, and each one has gotten more consequential rather than less. The accessibility case now has a legal deadline attached to it, the SEO case is unchanged in principle but sharper in execution, and a newer use case, feeding AI answer engines, barely existed when alt text advice was first written.
Alt Text for Accessibility
Screen readers announce alt text out loud, and for a blind or low-vision visitor it's often the only information they get about an image. WCAG, the Web Content Accessibility Guidelines that most accessibility laws point back to, requires it under Success Criterion 1.1.1: every piece of non-text content needs a text alternative that serves the same purpose. That rule isn't new, but the enforcement window is tightening. Title II of the Americans with Disabilities Act sets an April 2026 compliance deadline for public entities serving larger populations, with a 2027 deadline for smaller ones, and the EU's Accessibility Act has been in force since mid-2025. On a site serving a regulated audience, government, education, healthcare, finance, alt text has stopped being a best practice and become a compliance line item.
Beyond the legal angle, the practical test still works just as well as it always has. Try a free screen reader extension on your own site for ten minutes. Missing or lazy alt text becomes obvious fast, usually as a filename read aloud or silence where an image should be.
Alt Text for Image SEO
Search engines still can't fully read an image, so they lean on alt text, the filename, and the surrounding page copy to work out what it shows and whether it deserves to rank. Google's own image SEO guidance is direct about the failure modes: don't leave alt text off, don't stuff it with keywords, and don't write something generic when something specific is available. "Dalmatian puppy playing fetch in the backyard" gives a crawler far more to work with than "puppy" or a list of keywords crammed into one attribute. Alt text also decides whether an image can appear in Google Images or a Lens-style visual search result, which for product photos and how-to diagrams is a real, if secondary, traffic channel. For the full mechanics of naming, compressing, and structuring images beyond alt text, see this site's broader image SEO guide.
Alt Text for AI Overviews and Other Answer Engines
AI Overviews, ChatGPT, Perplexity, and similar systems face the same visual blind spot search engines always have, and for the same reason: running full image recognition on every photo across every page they crawl would be too costly to do at scale, so they default to text. That makes alt text, along with the caption and the paragraph next to an image, one of the clearest signals an answer engine has for understanding and citing what a page shows. A product image with vague or missing alt text is effectively invisible to a system deciding which page to quote in a generated answer. This is less a different way of writing alt text and more a reason the old advice, be specific, be accurate, don't stuff keywords, now pays off with a second audience. Anyone already tracking answer engine optimization as its own discipline should treat image alt text as part of that same checklist, not a separate task bolted onto SEO.
Alt Text for Loading Issues and Slower Connections
When an image fails to load, whatever the browser shows in its place is the alt text, not a blank box. A visitor on a spotty connection, a network that blocks certain image hosts, or a browser set to load images on demand still gets the gist of the page instead of a gap. This overlaps with the same loading-speed concerns covered in a guide to Core Web Vitals, since a large uncompressed image is exactly the kind of asset that stalls a page's Largest Contentful Paint score while it loads. Alt text costs nothing once the description already exists for accessibility and SEO reasons, and it's one more argument against ever leaving the attribute empty on an image that carries real meaning.
What Makes Alt Text Good? An Example First
Good alt text describes what a reader would need to know if they couldn't see the image, in the fewest words that get the job done. Take a photo of a beagle standing in a frosty field: alt="Beagle standing in a frosty field on a cold morning" names the subject and the setting in one clean phrase.
A short label works too, when the image doesn't carry much meaning beyond identifying the subject. A statue photo only needs alt="Dignity of Earth and Sky Statue", because the name is the information a reader needs.
Complex images need more room, and that's fine, alt text isn't capped at one word. A chart or map showing a specific claim, like an ocean current pattern, should describe what the visual actually communicates. alt="Clear evidence: Atlantic currents carry the Gulf Stream" tells a screen reader user the point of the graphic instead of just naming its subject.
To check whether a live page has alt text at all, or what it currently says, right-click any image and inspect its HTML, or run the page through a site-wide crawl that flags every image missing the attribute.
How to Write Alt Text Step by Step
Most CMS platforms give images a dedicated alt text field, so writing alt text rarely means touching raw HTML directly. The code itself is simple regardless: <img src="file.jpg" alt="description here" width="" height="">. What takes practice is deciding what the description should actually say. Five habits cover almost every case.
1. Identify what kind of image it is first
Before writing anything, decide whether the image is informative, functional, decorative, or complex, because each type gets a different treatment. An informative image, a product photo, a person, a scene, needs a description of what it shows. A functional image, a button or an icon acting as a link, needs a description of what happens when it's clicked, not what it looks like, so a magnifying-glass search icon gets alt="Search", not alt="magnifying glass icon". A decorative image, one that adds nothing to the page's meaning, gets an empty attribute rather than a forced description, covered in more detail further down. A complex image, a chart, infographic, or diagram, needs enough alt text to convey its point, sometimes paired with a fuller explanation in the surrounding body copy since a single attribute can't hold an entire dataset.
2. Describe what matters, skip what's obvious
A useful way to test a draft is to imagine describing the picture to someone over the phone, then cutting anything they'd already know from context. A photo sitting inside a review of running shoes doesn't need its brand repeated in the alt text if the brand is already the subject of the paragraph around it. Include what actually adds information: setting, action, and any detail that's genuinely part of what makes the image relevant to the page, including race, gender, or other visible identity markers when the image is making a point about representation.
A description crammed with every visible detail runs too long for an attribute meant to be spoken aloud in one breath. Anything that detailed belongs in a caption or the surrounding body copy, not crammed into the alt attribute. Most alt text should land at a sentence or a short phrase, roughly under 125 characters, long enough to be specific, short enough that a screen reader doesn't turn one image into a monologue. Alt text is not a caption either, so a source credit or citation belongs in a visible caption, not buried in the attribute.
3. Work the keyword in only if it belongs there
If the target keyword is genuinely what the image shows, include it, because that's the same information a crawler needs anyway. A photo of a runner in trail shoes on a muddy path can honestly carry alt="trail running shoes on a muddy forest path" if that's the keyword the surrounding article targets. What doesn't work is forcing a keyword into an image that isn't actually about it.
Long-tail keywords tend to fit alt text more naturally than short head terms anyway, since a longer descriptive phrase reads like an actual sentence instead of a tag.
For example, a page ranking for the broad term "whale shark" has little shot against established marine-life sites, but "whale shark with its mouth open" describes the actual photo and has a real path to ranking.
4. Never stuff it
Keyword stuffing in alt text fails both audiences it's meant to serve, a screen reader user hears a string of noun phrases instead of a sentence, and a search engine reads it as manipulation rather than description.
alt="custom dog tag, custom dog ID tag, customized dog ID" is a list, not a description, and it reads as spam to anyone or anything encountering it. The fix is the same one that works everywhere else in SEO: write the accurate description first, and let the keyword show up because the image is genuinely about it.
For a product photo of a chew toy, alt="heavy duty dog chew toys" is specific and accurate, and Google has surfaced that exact kind of image for the semantically related query "dog toys for heavy chewers" without any extra keyword forcing.
5. Skip "image of" and skip typos
Screen readers already announce that an element is an image, so opening the alt text with "photo of," "picture of," or "image of" just adds words a listener has to sit through before the actual content starts. Go straight to the description. The same discipline applies to spelling: a typo in alt text can garble what a screen reader says aloud, and it's just as visible to a crawler evaluating the rest of the page's quality signals.
Buttons and linked icons are images too, so an image-based "Submit" button still needs alt="Submit" describing the action it performs, not what the graphic looks like. Long, information-dense images like infographics can't fit their full content into one attribute either, so pair a short summary in the alt text with a fuller explanation in the body copy next to the image instead of trying to cram a paragraph into the attribute itself.
Alt Text for Special Cases: Decorative Images, Text in Images, and Video
Decorative images, ones that add nothing to the page's meaning beyond visual flourish, don't need a description, but they still need the attribute present and empty. Leaving alt off an <img> tag entirely can cause some screen readers to fall back to reading the filename aloud, which is worse than saying nothing. Writing alt="" tells the screen reader to skip the image silently, which is the correct outcome for a stock background photo, a spacer graphic, or a repeated icon that's purely decorative. This is also the right call for an image that's a link with a text label sitting right beside it, since the text already carries the meaning.
A screenshot or photo of a block of text is the one case where alt text has to do double duty. Web crawlers can't read the words baked into a JPEG or PNG the way they read HTML text, and a screen reader has no idea a quote or a paragraph is sitting inside the image at all. The fix is to type the image's actual text into the alt attribute, or better, avoid publishing text as an image in the first place and put it on the page as real HTML.
Video doesn't take alt text at all, since there's no single frame to describe, but it needs the same underlying accommodation in a different form: a transcript. A written transcript serves hearing-impaired viewers, viewers who don't speak the video's language and want to translate it, and anyone watching with the sound off, which describes a large share of social video traffic already.
How to Check and Fix Alt Text Across a Whole Site
Checking one image is simple: right-click it, select "Inspect" or "Inspect Element" (Control-click on a Mac), and the alt attribute shows up in the HTML panel. That works fine for spot-checking a single page, but a single URL scan only catches missing or empty alt attributes on that one page, not the rest of the site.
Neither approach scales past a handful of pages, and most sites carrying years of content have thousands of images with alt text that was never written, was written badly, or has gone stale as products and pages changed. A crawl-based site audit that checks markup across every indexed page, the kind covered in a broader on-page SEO checklist, is the only realistic way to find every image missing a description on a site of any size, rather than relying on someone stumbling across the gap manually. Site Auditor inside Search Atlas runs that kind of full-domain crawl and flags content markup issues, including missing or malformed alt attributes, across every page it indexes, then hands each one a prescriptive fix instead of just a red flag.
Should You Use AI to Generate Alt Text?
AI alt text generators have gotten good enough by 2026 that plenty of sites now run one instead of writing every description by hand, and for large image libraries that's often the only realistic option. These tools look at an image and output a description automatically, some plugging directly into a CMS or SEO plugin, others running as a bulk pass across a whole media library. For product catalogs running into the thousands, a generator that gets most descriptions right and flags the rest for review beats a backlog of images with no alt text at all.
The catch is that automated description still needs a human check before it ships, especially on anything customer-facing or legally regulated. An AI model can misidentify a product, miss the point of a chart, or produce a description that's technically accurate but useless, a generic "shoe on white background" instead of the actual model and color a shopper is trying to identify. Treat AI-generated alt text the way a competent editor treats any AI first draft: a fast way to clear a backlog, not a substitute for reviewing what a page's most important images actually need to say.
Alt Text Questions Worth Answering Directly
Is alt text a Google ranking factor? Not directly for the page itself, but it's a confirmed input into how Google understands and ranks images in Google Images and visual search, and it feeds the broader on-page signals Google uses to understand what a page is about.
Does alt text affect AI Overviews and other AI answers? Yes. Answer engines rely on text, including alt text, to understand and cite what an image shows, since running image recognition on every photo across every crawled page isn't cost-effective at their scale.
Do videos need alt text? No, alt text applies to still images. A video needs a written transcript instead, covering the same accessibility ground for viewers who can't hear or see it play.
Is alt text the same as a caption? No. A caption is visible text under an image, usually for a credit or clarifying note. Alt text lives only in the HTML and is read by screen readers and crawlers, not seen by a sighted visitor.
How long should alt text be? Long enough to be specific, short enough to say in one breath, roughly a sentence or a short phrase and generally under 125 characters. Complex images like infographics can run longer if the surrounding body copy also carries part of the explanation.
Some Models of Decent to Effective Alt Text
There are varying levels of alt text quality, and the difference between decent and genuinely good usually comes down to specificity. Here are a few examples moving from bad to better to best:
Bad: alt=”dog”
Better: alt=”brown dog with leash”
Best: alt=”Tan poodle happily playing in the grass with its leash still attached”
Bad: alt=”people with books”
Better: alt=”mother and son doing homework”
Best: alt=”illustration of a black mother helping her son with his homework to demostrate the power of involved parents.”
Bad: alt=”picture of a cup, napkin, and pen”
Better: alt=”a blue coffee mug next to a napkin with writing and pen”
Best: alt=”a blue coffee mug with coffee to the left sitting on a wooden table with a pen opposite and a napkin between with the words set goals, not limits”
Write Alt Text Once, Serve Three Audiences
Alt text takes thirty seconds to write and gets skipped constantly anyway, usually because it looks like a small, invisible detail with no clear payoff. The payoff is just split three ways instead of showing up in one obvious place: a blind visitor who understands the product photo, a search crawler that can index the image accurately, and now an AI Overview or chatbot that can cite what a page shows without ever seeing it. None of those outcomes require creative writing talent or technical skill, only the habit of describing what an image actually shows, in plain language, every time one gets published.
