Semalt series

Images and Video: The Traffic Most Sites Throw Away Without Noticing

File names, alternative text, modern formats, video sitemaps and transcripts. Small work, disproportionate return — and almost nobody does it.

Updated: 2026-08-18 9 min read 2.068 words
A photographer organising images on a computer

Key takeaways

  • Image search is a real traffic source for visual businesses, and most sites disqualify themselves through file names and missing alt text.
  • Alt text exists for people who cannot see the image. Written properly for that purpose, it also does everything SEO needs.
  • Video without a transcript is invisible to search. The transcript is usually the highest-return twenty minutes in the whole project.
  • Media is also where most performance problems live, so this work pays twice.

Every site has images and an increasing number have video, and on the majority of them both are published exactly as exported: IMG_4471.jpg, no alt text, four times larger than displayed, and a video embedded with no accompanying text at all. It is not neglect exactly — nobody ever decided to do it badly. It simply falls between the person who takes the photographs and the person who thinks about search.

This article covers closing that gap using the Semalt platform to find what is missing: what actually matters for images, what video requires, and where the returns are concentrated for a business that sells something visual.

Who this matters for

Honesty first: image search is not equally valuable to everyone. For a B2B software company, it is close to irrelevant. For a restaurant, a hotel, a clinic showing before-and-after work, an architect, a furniture shop or anyone selling something people want to look at before buying, it is a genuine acquisition channel that most competitors ignore.

4 fieldsFile name, alt text, dimensions, format — the entire basic checklist
20 minTo add a transcript, usually the best-value work on a video page
2 winsMedia work improves both discovery and page performance

Working parameters from the checklist below.

The four things that matter for images

  1. File name

    Describe the subject in words, separated by hyphens, before uploading. cadeira-carvalho-macico.jpg rather than IMG_4471.jpg. It costs nothing at the point of upload and cannot be retrofitted cheaply once a thousand files exist.

  2. Alt text

    Write what the image shows, for someone who cannot see it. Specific and factual, not a keyword list. If a person listening to a screen reader would understand what is in the picture, the alt text is correct.

  3. Dimensions and format

    Serve the image at the size it is displayed, in a modern format, with width and height declared. This is a performance requirement that happens to also matter for search.

  4. Surrounding context

    An image is interpreted partly through the text around it. A photograph placed under a relevant heading with a descriptive caption is understood far better than the same file dropped into a gallery with no words nearby.

The alt text rule that resolves every edge case. Write it for a person using a screen reader, not for a crawler. That single framing gets you the right answer every time: describe photographs, leave decorative images empty, describe the destination for images used as links, and never stuff keywords — because a person listening to keyword-stuffed alt text has a genuinely unpleasant experience, which tells you everything.

We cover this in detail in Internal Links: The Free Lever Almost Every Site Leaves Unused.

Alt text: the cases people get wrong

Image typeCorrect alt text
Informative photographDescribe what it shows, specifically
Purely decorativeEmpty (alt="") — not omitted, empty
Image used as a linkDescribe the destination, not the picture
Chart or diagramState the finding, not "chart"
Product photographProduct, key attribute, view — no model codes
Text rendered as an imageThe text itself, verbatim — and reconsider the design

The second row is worth noting because it is counterintuitive. A decorative image needs an explicitly empty alt attribute, not a missing one. Missing means a screen reader may read the file name aloud; empty means it correctly skips it.

Video: what actually gets found

Video has a structural problem for search: the content is inside a file that cannot be read. Everything that makes a video discoverable is text placed around it, and most sites publish none.

Three things matter, in order of return.

A transcript on the page. This is the highest-value item and the most commonly skipped. It gives the page substantial relevant text, makes the content accessible to people who cannot or will not watch, and lets someone find the specific part they need. Automatic transcription then fifteen minutes of correction produces something entirely adequate.

Text around the video. A descriptive heading, a summary of what it covers, and the key points in writing. A page containing only an embedded player is, from a search perspective, a blank page.

VideoObject markup and a video sitemap. These make the video eligible for video-specific results, with a thumbnail and duration. Worth doing where video is central to the business, skippable where it is occasional.

A page with a video and no transcript is a blank page with a rectangle on it. Twenty minutes of transcription turns it into the most detailed page on the topic you have.Why transcripts outperform almost everything else in media SEO

Where to host video

A recurring question with a straightforward answer that depends on what you want.

Embed from a video platform when

  • You want reach on the platform itself as a channel
  • You lack bandwidth or infrastructure for self-hosting
  • The video is promotional and reach matters more than exclusivity

Self-host when

  • You want the page itself to rank, not the platform's page
  • The video is core product content you do not want surrounded by competitors' suggestions
  • You can manage the delivery and performance cost properly

The practical compromise most businesses land on: host on a platform for reach, embed on your own page with a full transcript and supporting text, so the page can rank on its own merits regardless of where the file lives. The embed is then a convenience rather than the entire content.

The performance dividend

Media work is unusual in that it pays twice. The same optimisation that makes images findable — correct dimensions, modern formats, declared sizes — is also the single largest lever on page performance for most sites.

Two specifics worth stating. Lazy-load everything below the fold and never the main image at the top, since that is the element the performance measurement is waiting for. And be careful with video embeds: a player script loaded on page render can cost more than every image on the page combined. Loading the player only when someone clicks a thumbnail is a well-known pattern and it frequently transforms a page's measured performance.

There is a full walkthrough in Speed That Counts.

Check what your CMS is actually serving. Many platforms generate multiple sizes on upload and then serve the original anyway, because the template asks for the full-size version. The images appear optimised in the media library and arrive at the browser at four thousand pixels. Look at what is actually delivered rather than what was uploaded — this single check explains a large share of unexplained slowness.

Practical notes for Portuguese sites

Two details specific to working here. File names and alt text should be in the language of the page and of the searcher — Portuguese for the Portuguese version — which sounds obvious and is frequently not done, because the photographs arrive named in English from a supplier or a stock library and nobody renames them.

Accents in file names are best avoided: use the unaccented form in the file name and the correct accented text in the alt attribute and the caption. The file name is a technical identifier where accented characters cause encoding problems; the alt text is human-facing and should be written properly.

For businesses photographing their own work — restaurants, clinics, workshops, hotels — there is a real and underused opportunity here. Original photography of the actual place, properly named and described, competes for image searches against stock photographs that everyone else is using. That is one of the few areas where a small local business has a structural advantage over a larger competitor.

Galleries, portfolios and the archive problem

Visual businesses tend to accumulate large galleries, and galleries have a specific failure pattern worth naming: hundreds of images on one page, none of them individually reachable, all of them lazy-loaded, with no text anywhere. It looks impressive and it is close to invisible.

See also: Structured Data in Practice.

Three adjustments make a portfolio work as a discovery asset rather than a decorative one.

Give projects their own pages. A gallery of two hundred images has one URL competing for everything. Twenty project pages, each with several images, a description of what the work involved and the constraints it addressed, gives you twenty pages that can each rank for something specific. The gallery then becomes the hub linking to them.

Write the context nobody else writes. The photographs show what it looks like. The text should say what it cost, how long it took, what the difficulty was, what the client asked for. That is what people actually search for before commissioning something, and it is precisely what competitors leave out because it feels like giving away information.

Do not paginate the archive into oblivion. Older work reachable only through page nine of a gallery is, structurally, work that does not exist. Group by category, by year or by type so that everything is a couple of clicks from somewhere meaningful.

One caution about batch-processing an existing archive: resist retrofitting alt text to four thousand images with generated descriptions. Generic descriptions of the "modern interior with natural light" variety add nothing for either readers or search, and the effort is better spent writing genuinely specific text for the two hundred images on pages that matter commercially.

Finding what is missing

A crawl reports the state of every image on the site at once: missing alt attributes, oversized files, images without dimensions, broken references. That list is the work queue, and it is usually longer than expected.

Prioritise it the same way as everything else: images on commercial pages first, then high-traffic pages, then the archive. Retrofitting alt text across four thousand images is a project nobody finishes; doing it for the two hundred images on pages that generate enquiries is an afternoon and captures most of the value.

Where the returns are

For a visual business, this is one of the few areas where a modest amount of unglamorous work produces disproportionate results, because so few competitors do it at all. File names and alt text on commercial pages, correct sizing and formats, and a transcript under every video.

None of it is difficult. All of it is skipped by default, which is precisely why it remains worth doing.

If you have never audited the media on your own site, the crawl takes minutes. Open the dashboard, run it, and look at how many images on your commercial pages have no alt text at all.

Frequently asked questions

What should alt text actually say?

Describe what the image shows, for someone who cannot see it. Be specific and factual rather than listing keywords. Decorative images take an explicitly empty alt attribute rather than a missing one. Images used as links describe the destination rather than the picture. If a person using a screen reader would understand the image from your text, it is correct.

Do file names still matter?

Yes, and they are effectively free at the point of upload while being expensive to fix later. A descriptive, hyphenated name in the language of the page is a small, consistent signal about the subject. Renaming a thousand existing files retroactively is rarely worth it — start doing it correctly now and fix only the images on commercially important pages.

Should we host video ourselves or embed it?

Embed from a platform if you want reach on that platform and lack infrastructure; self-host if you want your own page to rank rather than the platform's. The common compromise is to host on a platform for reach and embed on your page with a full transcript and supporting text, so the page ranks on its own content regardless of where the file lives.

Is a transcript really necessary?

It is the single highest-return item on any video page. Without it, the page contains no readable content and is effectively blank to a search engine. Automatic transcription plus fifteen minutes of correction produces something adequate, makes the content accessible to people who cannot watch, and lets visitors find the specific part they need.

Try it

Open your Semalt dashboard

Audits, rank tracking, competitor data and reporting in one place. Sign in and you will be looking at real numbers for your own domain within minutes.

Sign in to Semalt

Or browse the service overview at semalt.com.