Nixeny Dijital - Mersin / Accessibility 2026
The home pages of 3,382 business websites in Mersin were put through automated accessibility checks on behalf of disabled users. 68.7% of the sites carry at least one structural barrier that does not slow a screen reader user down but stops them. One question: can everyone use this address?
What these numbers measure, and what they do not
Automated checks catch only part of the barriers real users meet — by common estimates about a third of them. A site passing every check here may still be unusable with a screen reader. The sentence “that site does not comply with WCAG” does not follow from this data.
Colour contrast, keyboard navigation order, focus visibility, keyboard traps, reading order, animation duration and the real screen-reader experience were not measured. All of them need the rendered page and usually a person; rather than being estimated, they were left out of scope.
The alt-text rate is calculated only over sites with images, the label rate only over sites with forms. A page with no images is neither a pass nor a fail on the alt-text question — it is unanswered. Six separate denominators are used, and each rate says which one it is.
Forms, tables and videos on inner pages are not seen. The rates are therefore a LOWER BOUND: the real state may be worse, it cannot be better. Sites using decorative markup correctly are not counted as faults and are not in these numbers.
Not a list of technical faults but people: a blind customer using a screen reader, a user navigating by keyboard alone, an older customer with low vision. Every number below says whether one of those users can reach a business.
Glossary · every term explained once
- Accessibility
- That a website can also be used by people with visual, hearing, motor or cognitive disabilities. The subject of this report is not a list of faults but people: a customer using a screen reader, a user navigating by keyboard alone, an older customer with low vision.
- Automated check
- A test that inspects the page's code by machine. By common estimates it can catch only about a third of the barriers real users meet. A site passing every check here may still be unusable with a screen reader.
- Screen reader
- Software that reads the page aloud. The user does not listen to the page from start to finish; they jump between headings, landmarks and link lists. Without those structures, all that is left is listening to everything one item at a time.
- Denominator
- This report has six main denominators: 3,382 sites with a readable home page, 2,886 with images, 1,805 with forms, 3,377 rendered in a browser, 3,169 with links and 278 with video. A page with no images is neither a pass nor a fail on the alt-text question — it is unanswered.
- Alt text
- A short description saying in words what an image shows. A user who cannot see “sees” the image through that text. An image without alt text either does not exist in a screen reader or is read out as a file name.
- Decorative markup
- alt="", role="presentation" and aria-hidden are the CORRECT way to hide ornament from assistive technology. They are not counted as faults in this study; doing otherwise would penalise the sites getting it right.
- Form label
- The linked text that tells a machine what a form field is asking for. A placeholder alone is not enough: it does not stand in for a label with a screen reader and it disappears as soon as the user starts typing. A user reaching an unlabelled field hears only “editable text”.
- Landmark
- Markers such as main, nav, header and footer that say which part of the page is what. Screen reader users navigate by jumping between them; without a main content landmark there is no way to go straight to the content.
- Skip link
- A link at the very start of the page that passes the menu and goes straight to the content. For a keyboard user it is the difference between passing dozens of menu items on every page and not.
- Vague link
- A link whose text alone does not say where it goes: “click here”, “more”, “read on”. Screen reader users listen to a page's links as a list; twenty items with the same text means twenty identical options.
- Screen-reader barrier
- Each of four structural failures that do not slow a user down but stop them: the page language being undefined, images missing alt text, form fields being unlabelled, and there being neither a main content landmark nor an H1 heading.
- Accessibility score
- A weighted 0-100 composite of seven components: text alternatives (20), form labels (20), structure and navigation (20), mobile readability (15), link text (10), media and motion (10), accessibility statement (5). A component that cannot be measured is removed from both numerator and denominator; below 60% coverage no score is published.
- Accessibility statement
- A page where a site sets out its own accessibility state, the gaps it knows about and a way to report a problem. It is not a certificate of compliance; it is a commitment and a channel.
61.3% of form fields do not say what they want
When someone using a screen reader reaches a form field, they hear that field’s label: “full name”, “phone”, “your message”. Without a label the only thing they hear is “editable text” — nothing says which piece of information is wanted.
The home pages of Mersin’s business sites carry 11,254 form fields that need a label. 6,902 of them (61.3%) have none. 75.2% of the sites with a form carry at least one unlabelled field.
The most common cause is the placeholder: the grey hint text written inside the field. It does not stand in for a label with a screen reader and it disappears the moment the user starts typing. 62.9% of the sites with a form rely on it alone.
These two rows count fields, not sites. Converting the same number to sites gives a different rate, and the two cannot stand in for each other.
The site-level counterpart of the same finding. The “unnamed button” row has a different denominator: only sites with a button are counted.
10,651 images are invisible to a user who cannot see
Alt text is the short description that says in words what an image shows. A user who cannot see “sees” the image through that text. Without it the image either does not exist or is read out in a screen reader as a file name — “IMG underscore 4471 dot jpg”.
Of the 67,550 images that need alt text, 10,651 have none. 37.6% of the sites with images carry at least one such image.
That number does not include sites hiding decorative images correctly. alt="", role="presentation" and aria-hidden are the correct way to hide ornament from assistive technology and are not counted as faults in this study. 56.4% of the sites do it.
The denominator is the images that need alt text. Images correctly marked as decorative are outside it.
The first row is not a fault but correct practice, and it sits inside the same figure on purpose: without it, every empty alt in the corpus is assumed to be an oversight.
45.1% of sites have no main heading
Someone using a screen reader does not listen to the page from start to finish. They navigate by jumping between headings, landmarks and link lists. Headings are a screen reader’s table of contents: without an H1 what the page is about is lost, and where the order skips, its structure is.
45.1% of the sites have no H1 heading at all and 46.2% skip a heading level. 63.1% define no main content landmark (main) — meaning there is no way to pass the menu and go straight to the content. The number of sites offering a skip link is 467 (13.8%).
The cheapest line on the list is also one of the most striking: 446 sites never define the page language at all. That is a single HTML attribute, and without it a screen reader tries to read Turkish text with English pronunciation.
Ordered not worst-to-best but starting with the ones that remove a navigation method entirely. The last two rows count correct practice, not a gap.
All 278 videos are without captions
Screen reader users listen to a page’s links as a list. If twenty items in that list carry the same text, there are twenty identical options. 69.0% of the sites carry at least one link with no text — usually social media icons without a label; 25.4% carry link text like “click here” or “more” that says nothing on its own.
On the media side there is one number and it has no exception: on all 278 sites with video, the videos have no captions. For a deaf user that content does not exist.
Three checks, one denominator. They are not mutually exclusive: one site can carry both an empty link and a vague one.
Three different denominators: captions only from sites with video, frame titles only from those with an embedded frame, autoplay from every readable page.
These four measurements are taken from the page as rendered in a real browser — not from the code itself. For a user with low vision the most direct barrier is the first row.
68.7% of sites carry at least one stopping barrier
Not all of the checks above weigh the same. Four of them do not slow a user down but stop them: with the page language undefined the text is read with the wrong pronunciation, with images missing alt text the content is lost, with a form field unlabelled the form cannot be filled in, and with neither a main content landmark nor an H1 there is no way to find your bearings on the page.
At least one of those four barriers is present on 68.7% of the sites. Those sites represent 2,711 businesses. The number of sites carrying no barrier at all is 1,058 — 31.3%.
These rows do not sum: one site can carry all four barriers. The total is given as the buckets below.
Each of the hundred squares stands for about 34 sites. The buckets are mutually exclusive and sum to the denominator.
The median site scores 71.5; 11 publish a statement
Seven components give a weighted 0-100 score: text alternatives (20), form labels (20), structure and navigation (20), mobile readability (15), link text (10), media and motion (10) and an accessibility statement (5). The median is 71.5, with quartiles between 62.8 and 82.4.
Looking at the components shows where the distribution comes from. The median for mobile readability is 100.0 and for link text 92.6 — those are things the theme brings ready-made. Form labels sit at 57.5 and the accessibility statement at 0.0: those are the things a person has to sit down and do.
Five buckets summing to the denominator. The distribution gathers in the 60-79 band: half the sites are there.
Each of the hundred squares stands for about 31 sites.
The component's weight is written beside it. The statement component's median is zero and it carries the smallest weight: so that a single absence does not swamp the six that describe how the page actually behaves.
A statement is not a certificate of compliance. It is a page where a site sets out its own accessibility state, the gaps it knows about and where somebody wanting to report a problem should write — that is, a commitment and a channel. The number of business sites in Mersin publishing one is 11.
A second measurement points the same way
A report that scores other people’s work owes its reader a way to check it. Google Lighthouse’s own accessibility score is not an input to this study’s score — its weights are a closed box that moves between versions. It is published alongside.
The Lighthouse median is 83.0 and this study’s median 71.5. The Spearman rank correlation between the two measurements is 0.642 (Pearson 0.627, n = 3,072). They are not copies of each other — they measure different things with different weights — but they put the same sites in close to the same order.
The gap between them carries information too. The Lighthouse score is 90 or above on 945 sites (28.0%); some of those sites still carry unlabelled form fields or uncaptioned video. Both measurements are automated and neither stands in for real use.
The dot shows the median, the bar the interquartile range. The two measurements sit on the same 0-100 scale but do not measure the same thing: Lighthouse uses its own check set, this study its seven components.
| Spearman rank correlation | 0.642 |
|---|---|
| Pearson correlation | 0.627 |
| Comparison base | 3,072 |
| Lighthouse score 90+ | 945 (28.0%) |
What decides it is not the trade but the theme installed
The spread between sectors is narrower than the spread between platforms. The medians of the 30 sectors run from 59.9 to 82.3; across platforms Wix scores 85.0 while Ticimax scores 59.2.
That says the problem comes largely from the theme’s defaults — and the good news sits exactly there: one theme fix can change thousands of pages at once. So does the bad news: a business usually did not produce these gaps in its own site by its own decision, and cannot fix them by its own decision either.
The two denominators on a row are not the same. The Lighthouse column is an independent measurement and is not an input to this study's score.
| Platform | Sites | Median score | Lighthouse | Barriered | No H1 |
|---|---|---|---|---|---|
| Wix | 121 | 85.0 | 96.0 | 15.7% | 26.5% |
| Next.js | 146 | 83.8 | 91.0 | 39.7% | 28.8% |
| WordPress | 559 | 76.9 | 87.0 | 58.4% | 40.5% |
| IdeaSoftcomparison | 37 | 75.2 | 77.0 | 86.5% | 83.8% |
| WooCommerce | 386 | 74.6 | 85.0 | 62.3% | 51.4% |
| Custom build | 1,912 | 68.8 | 79.0 | 76.3% | 45.9% |
| Ticimaxcomparison | 48 | 59.2 | 71.0 | 97.9% | 72.9% |
Each tick is one sector's median score. Tick height carries no data; it varies on a three-step cycle only to keep values that fall next to each other countable.
Geography does not decide this measurement either: the range is narrow and districts with few sites do not enter the comparison.
This is not a WCAG compliance audit
Every line in this report measures the result of an automated check. None of them says a site is or is not accessible, and none of them means a business is excluding disabled people.
The sentence “one in 6,902 form fields on this page is unlabelled” is true and checkable. The sentence “this site does not comply with WCAG” does not follow from this data: automated checks catch only part of the barriers real users meet — by common estimates about a third. A site passing every check here may still be unusable with a screen reader.
What was not measured matters as much as what was: colour contrast, keyboard navigation order, focus visibility, keyboard traps, reading order and animation duration. All of them need the rendered page and usually a person. Rather than being estimated they were left out of scope, and this report makes no claim about them.
Finally: this is not a measurement of intent. The findings are presented in aggregate and no business is named. Most of the gaps come not from a business’s decision but from the defaults of the theme it uses.
- That the structural barriers automated checks catch are common on local business sites.
- That a screen reader user will meet concrete problems: not being able to fill in a form, not being able to navigate by heading.
- That the difference comes largely from the defaults of the theme and platform used.
- That most of the problems are cheap to fix: adding a label, writing an alt text, defining the page language.
- That two independent measurements — this study's score and the Lighthouse score — point the same way.
- “That site does not comply with WCAG” — this is not a compliance audit.
- “X% of sites are accessible” — passing automated checks is not the same as being accessible.
- Anything about colour contrast, keyboard navigation or focus order — those were not measured.
- “That business is excluding disabled people” — the findings are aggregate, intent was not measured, and no business is named.
- That a gap absent from the home page is absent from the whole site — only the home page was examined.
Limitations
- Only the home page was measured.
- Inner pages, dropdown menus and content arriving later via JavaScript are out of scope. If a site's contact form is not on its home page, that form was never seen in this study.
- Colour contrast and the keyboard experience were not measured.
- Contrast, focus order, keyboard traps and reading order are outside the scope of this study. No claim is made about them — favourable or otherwise.
- No real screen-reader test was run.
- No usability testing was carried out with any disabled user. This scan points to the next step but cannot replace it: trying a handful of sites together with someone who uses a screen reader shows in one sitting what a scan of 3,000 sites cannot.
- Pattern-based detections can produce false positives.
- Measurements such as vague link text and file-name alt text are done by pattern matching. A link legitimately reading “more” can fall into the same bucket.
- A single cross-section.
- The sites were measured as they stood at the time of the audit. One theme update can change part of these numbers at a stroke — which is, in fact, what most of the findings say.
- This is not a legal assessment.
- A binding technical accessibility standard for private-sector websites in Türkiye is not the subject of this study, and no legal assessment is made here. What is measured is the conditions under which disabled users can actually use a site.
- The sector and district breakdowns carry a small sample.
- Only segments with 30 or more audited sites were taken into the comparison. Rows below that threshold stay in the table but are not carried into a headline.
Citation rules
These six rules are not a stylistic preference. In this report, framing a number wrongly leaves a business facing either a legal claim or an accusation of intent — and this data can say neither.
- 01Do not say “does not comply with WCAG”; say “this automated check failed”.
- 02State everywhere that automated checks catch only part of the real barriers.
- 03Decorative markup (alt="", aria-hidden) is not counted as a fault.
- 04Every rate is given with its own denominator; a site with no images is not in the alt-text rate.
- 05Site names are not used to expose anyone; the findings are presented in aggregate.
- 06Because colour contrast and the keyboard experience were not measured, no claim is made about them.
The method, briefly
- 01UniverseThe dataset of the Mersin / State of Digital 2026 study: 4,167 audited domains, 3,395 reachable sites, 3,382 sites with a readable home page. Those sites represent 3,928 businesses. No new request was sent to any site for this report.
- 02The denominator ruleEvery check was calculated only over the sites that could answer it: alt text over 2,886 sites with images, form labels over 1,805 with forms, link text over 3,169 with links, the mobile measurements over 3,377 rendered sites, captions over 278 with video. No site that could not be measured was counted as zero.
- 03Nine check familiesAlt text, form labels, heading structure, landmarks, link text, table headers, media captions, embedded frame titles and mobile readability. Decorative markup (alt="", role="presentation", aria-hidden) was not counted as a fault in any check.
- 04Definition of a barrierFour structural gaps were counted as a “screen-reader barrier”: language undefined, image alt text missing, form field unlabelled, and neither a main content landmark nor an H1. These are gaps that stop rather than slow. How many of them each site carries was counted separately.
- 05ScoreA weighted 0-100 composite of seven components. A component that cannot be measured is removed from both numerator and denominator; below 60% coverage no score is published. 3,087 sites were scored, 308 could not be and were not counted as zero. The Lighthouse score is not part of the score.
- 06Independent verificationGoogle Lighthouse's own accessibility score was published alongside and the rank correlation between the two measurements computed (Spearman 0.642, Pearson 0.627, n = 3,072). Lighthouse is not an input to the score; two independent measurements pointing the same way make both more reliable than either alone.
| Sites with a readable home page | 3,382 |
|---|---|
| Sites with images | 2,886 |
| Sites with a form | 1,805 |
| Sites rendered in a browser | 3,377 |
| Sites with links | 3,169 |
| Sites with video | 278 |
| Scorable sites | 3,087 |
Sources
- 01Nixeny Dijital · Mersin / State of Digital 2026The same universe, the same fieldwork date. The 4,167 domains, 3,395 reachable sites and 3,382 home-page caches this report uses come from that study. No new request was sent to any site for this report; the measurements were made on the existing HTML and browser output.nixeny.com
- 02WCAG 2.2 · Web Content Accessibility GuidelinesThe checks are informed by the AA criteria of this guideline. This is NOT a statement or audit of conformance: automated checks cover only the machine-testable part of the criteria, and passing them does not mean conforming.www.w3.org
- 03WAI-ARIA Authoring PracticesThe reference for correct practice on landmarks, labelling and decorative markup. Not counting decorative images hidden with aria-hidden and alt="" as faults rests on this source.www.w3.org
- 04Google Lighthouse · AccessibilityAn independent verification measurement. It is NOT AN INPUT to this study's score; its weights are a closed box that can change between versions. It is published alongside and the correlation between them given.developer.chrome.com
- 05Measurement and detection patternsNine check families: alt text, form labels, heading structure, landmarks, link text, table headers, media captions, embedded frame titles and mobile readability. The mobile measurements were taken from pages rendered in a real browser with Playwright. The pattern list and the verification logs are shared on request.
This report’s numbers were not compared with an accessibility measurement from another province or country. A “world average” produced with a different check set and a different denominator looks, placed side by side, as though the two numbers measured the same thing. What it shares with Mersin / State of Digital 2026 is the universe and the fieldwork date.
Colophon
Conducted and published by Nixeny Dijital, Mersin. The measurement, the scoring and the report design were carried out by Nixeny. The universe and the site list are shared with the Mersin / State of Digital 2026 study. No new request was sent to any site for this report.
If a business wants to know its own site’s row in this study, writing in the domain name is enough. The check list, the detection patterns and the verification logs are also shared on request: info@nixeny.com