Six Steps to Better Images: A Best Practices Checklist

Aug 24, 2026

Here's our workflow for better images — six steps, in order, applied to every image as you upload it! Print it, pin it near your desk, or paste it into whatever document your team uses for publishing. It takes just a few seconds per image once it's habit.

A women working on a computer creating alt text for accessibility

Step 1 — Rename the file descriptively before uploading
Do this first, because it's the step you can't easily redo later.

IMG_4471.jpg tells nobody anything. blue-ceramic-mug-handmade.jpg tells search engines what the picture shows, and it's the fallback a screen reader may read aloud if alt text is ever missing.

[ ] File renamed before upload, not after
[ ] Words separated by hyphens, not underscores or spaces
[ ] Lowercase, no special characters
[ ] Describes the image, not the campaign (spring-sale-banner-3-final-FINAL.jpg helps no one)
[ ] Reasonably short — three to five words is plenty
Why it's first: once an image is uploaded and embedded across your site, renaming it means broken links. Two seconds now, or a cleanup project later.

Step 2 — Ask: does this image carry information, or is it decoration?
Everything downstream depends on this answer, so answer it deliberately.

The test: if this image vanished, would the reader lose anything?

Carries information — product photos, charts, diagrams, screenshots, headshots, anything with text in it, anything you'd describe out loud to someone on the phone
Decoration — dividers, background textures, corner flourishes, abstract shapes, an icon sitting beside a text label that already says the same thing

[ ] Asked the question consciously rather than defaulting
[ ] Checked whether nearby text or a caption already conveys the same thing (if so, the image is likely decorative)
[ ] When genuinely unsure, treated it as informative — over-describing is a smaller error than leaving someone without

Step 3 — If it carries information, write one sentence describing what it conveys here
That last word does the work. Not what's in the frame — what the image is doing in this context.

The same headshot might be "Sarah Chen, founder" on your team page and "A person working at a home desk" on a post about remote work. Same file, different job.

[ ] Roughly one sentence — long enough to be useful, short enough to listen to
[ ] Describes the image's purpose on this page, not every visible detail
[ ] Doesn't start with "image of" or "photo of" — screen readers already announce it's an image
[ ] No keyword stuffing (it sounds like gibberish read aloud and stopped helping rankings long ago)
[ ] Ends with a period, which gives screen readers a natural pause
[ ] Doesn't duplicate a caption word for word — a screen reader user would hear it twice
[ ] Complex images (charts, infographics, maps) have a fuller description nearby, since a sentence won't cover them

Step 4 — If it's decoration, leave the field empty
Empty. Not a space, not "decorative image," not a description. In Microsoft Office products, check "Decorative" and it skips that image as an Alt Text requirement.

Empty alt text tells a screen reader to skip the image entirely. That's the goal — silence is better than an interruption that adds nothing.

[ ] Alternative Text field left genuinely blank
[ ] Didn't type "decorative," "spacer," or "n/a"
[ ] Didn't put a single space in the field
[ ] Confirmed the platform produces empty alt rather than falling back to the filename or title

Some platforms auto-populate alt text from the filename or title if you leave it blank. Check yours once — it's worth knowing which behaviour you're working with.

Step 5 — If there's text in the image, include that text
No exceptions. Text inside an image isn't text — it's a picture of letters. Screen readers can't read it, search engines can't index it, translation tools skip it, and it turns to mush when someone enlarges the page.

[ ] Every word visible in the image appears in the alt text
[ ] Quote cards, memes, and screenshots have their text transcribed
[ ] Event details — date, time, address, price, registration link — also appear as real text in the post, not only in the graphic
[ ] Infographics have a text summary nearby, not just alt text
[ ] For memes, the alt text conveys the joke, not just the picture

That third box is the one that matters most. If your entire event announcement is a JPG, then for a meaningful number of people, you didn't announce an event.

Step 6 — Read it aloud. If it sounds odd, rewrite it
This catches nearly everything the other five steps miss. Alt text is heard, not read. Saying it out loud surfaces the awkward phrasing, the keyword stuffing, the redundancy, and the descriptions that technically apply but don't actually help.

[ ] Read the alt text out loud, actually out loud
[ ] It sounds like a person describing a picture, not a search query
[ ] Someone who couldn't see the page would understand why the image is there
[ ] It isn't so long you'd lose patience hearing it
[ ] Any AI-suggested alt text was reviewed and rewritten, not accepted as-is

Automated suggestions have gotten decent at describing what's in frame. They remain bad at knowing what an image is doing in your post — that part is yours.

The quick checklist:
- The whole thing on one line each
- Rename the file before uploading
- Ask whether it carries information or decorates
- Describe it in a sentence, in context
- Leave it empty if decorative
- Include the text if there's text in the image
- Read it aloud and rewrite what sounds odd
- Fix your old images, too!
  
  If the backlog is bigger than an afternoon
  For a lot of organizations it is — years of uploads, thousands of images, a document library nobody wants to open. That's an ordinary place to be, and it's not a good use of your team's month.

We provide accessibility audits, ongoing accessibility support, and content remediation, including working through image libraries at scale and training your team so the next few years of uploads don't rebuild the same backlog.