ResourcesTesting and fixing
Automated scans vs. testing by people
Quick answer
A scanner finds the errors a tool can detect, like missing alt text, low contrast and form fields with no label. It can't tell whether a person can book or buy with a keyboard or a screen reader. You need both: scans to cover every template, and people to test the rest.
Facts last checked
What the evidence says
Two public studies measured what scanning tools find. They count different things.
Deque, 202157% of the issues logged in Deque's own audits were found by its automated tests. This counts issue volume, not WCAG criteria.
Automated tests found an issue under 16 of the 50 WCAG 2.1 Level A and AA criteria in Deque's data.
At least one of 10 tools found 101 of the 143 barriers planted on a UK government test page. All 10 missed the other 42.
Source: Deque, 2021; UK Government Digital Service, 2017.
- All 10 tools combined: 71%
- Tool that found the most, counting errors and warnings: 37%
- Tool that found the least: 17%
Source: UK Government Digital Service, 2017. One purpose-built test page.
What a scan finds, what a person finds
A scan can find
- Images with no alt attribute
- Text with low contrast
- Form fields with no label
- Links and buttons with no name
- A page with no language or no title set
- Zoom that has been turned off
A person has to check
- Whether alt text describes the image
- Whether every control works with a keyboard
- Whether focus is visible and moves in a sensible order
- Whether a screen reader announces errors and updates
- Whether link text makes sense, not just "click here"
- Whether captions are there and are accurate
Same element, two questions
| On the page | A scan asks | A person asks |
|---|---|---|
| Image | Is there an alt attribute? | Does the text describe the image? |
| Form field | Does it have a label? | Is it the right label? Are errors announced? |
| Button | Does it have a name? | Can you reach and use it with a keyboard? |
| Link | Does it have text? | Does the text say where the link goes? |
| Video element | Can't decide. Flags it for review. | Are there captions? Are they accurate? |
Source: axe-core rule descriptions; Deque coverage report, 2021
Needs review
Some checks end without an answer. The tool flags the item, and a person has to decide.
Contrast
Text over a gradient
When the background is a gradient, or another element overlaps the text, the tool can't always work out the contrast.
Captions
Video captions
The axe-core rule for captions on video elements only ever returns needs review. A person has to play the video.
Skip link
A way past the menu
The rule that looks for a way to skip repeated blocks also returns needs review only.
What we do with themaxe-core, a scanning engine we use, says these items "require further testing". They aren't errors, and they aren't cleared. Each one is a task for a person. Our team checks them, and only confirmed problems go into your issue log.
How we combine them
Scan every template
Automated scans run on every page template in scope.
Review each flag
A person checks every item the tools mark as needs review.
Keyboard test
A person goes through your key journeys, like booking or checkout, with a keyboard only.
Screen reader test
The same journeys again, with screen readers on a computer and a phone.
Retest after fixes
A person tests each fixed issue again.
What a clean scan means
A clean scan tells you
- The tool detected none of the errors it can check for
- On the pages it scanned
- On the day it ran
It can't tell you
- Whether the site meets WCAG 2.2 Level AA
- Whether a person can finish a booking or a checkout
- What changed after the scan
W3C, the body that publishes WCAG, says "no tool alone can determine if a site meets accessibility standards. Knowledgeable human evaluation is required to determine if a site is accessible."
About our free SnapshotIt's an automated check of up to 10 key pages, so it finds only some issues. It's a starting point, not an audit.
Sources
- UK Government Digital Service, What we found when we tested tools on the world's least accessible webpage (2017)
- Deque, The Automated Accessibility Coverage Report (2021)
- Deque, announcement of the coverage study (2021)
- Deque, axe-core rule descriptions (2026)
- Deque, axe-core API documentation (accessed 2026)
- Deque University, axe 4.10 rule: color-contrast (accessed 2026)
- WebAIM, WAVE Help (accessed 2026)
- GOV.UK Service Manual, Testing for accessibility (2024)
- W3C Web Accessibility Initiative, Evaluating Web Accessibility Overview (2026)
We provide technical accessibility services, not legal advice.
More guides
The most common accessibility barriers
The six errors WebAIM's yearly scan detects most often on home pages, what each fix looks like in the code, and the barriers a scan misses.
Do accessibility overlays fix a website?
What an accessibility overlay widget is, what a script can and can't change on your site, and what the public record shows.
WCAG explained: principles, levels and versions
What WCAG is, what Levels A, AA and AAA mean, and which version the rules name, shown in charts and tables.
Start with a free snapshot.
Tell us your website. We'll run an automated check of up to 10 key pages and walk you through what we found. It's an automated check that finds only some issues, not an audit.
Get a free snapshot