An Archive Everyone Can Read: Accessibility by Design in UK Document Capture

Accessibility begins at the scanner

A scanned image PDF is a picture of a page. For a screen reader user, it is often silent. For a keyboard-only user, it may be a dead end. For a colleague with low vision, it may be a document that cannot be enlarged without turning into a blur. The capture team may believe it has created a digital record, but it has created a digital image. The difference matters.

Accessibility is not a repository feature that can be bolted on at the end. It is shaped by decisions taken during capture: resolution, OCR settings, reading order, metadata fields, file format and quality control. A capture project that ignores those decisions creates an archive that only some employees can use.

The OCR quality question

Optical character recognition is usually discussed as a search feature. It is also an accessibility feature. Without a text layer, a screen reader has nothing to read. Without accurate text, search results mislead and pronunciation becomes unreliable.

Capture teams should set minimum standards by document type rather than relying on a default. Text documents generally need a resolution high enough to preserve character shapes; 300 dots per inch is a common baseline, but coloured paper, small print and stamps may need more attention. Deskewing, blank-page removal and speckle reduction help, but over-processing can damage faint text. Test samples rather than assuming.

Accuracy should be measured at word level as well as character level. A character error rate of two per cent may sound acceptable until it affects supplier names, invoice numbers or patient references. For high-value records, build a verification step. For lower-risk records, sample and monitor. The aim is not perfect OCR everywhere; it is fit-for-purpose text that supports reading, search and navigation.

Structure: headings, lists and tables

A screen reader user often navigates by headings, lists and tables. If OCR outputs one continuous block of text, that navigation collapses. The document may be searchable but not usable.

Where possible, capture workflows should preserve structure. Layout analysis can identify headings, columns and tables. Tagged PDF or accessible HTML can expose that structure to assistive technology. Forms need labels. Tables need header rows. Multi-column letters need a reading order that follows the intended sequence, not the scanner's guess.

Scanned paper is difficult. A typed letter with a logo, address block and two columns may never convert perfectly. The practical response is to prioritise. High-use records, such as policies, procedures and frequently requested client documents, may need a reviewed accessible version alongside the scanned original. Low-use historical records may be acceptable as searchable images with a clear description. The archive should state what level of accessibility each record offers.

Metadata that works

Filenames such as IMG_20261108_001.pdf are not metadata. They tell a person almost nothing and a screen reader even less. Accessible metadata gives a record a meaningful title, date, document type, reference, owner and retention period. It also gives the repository something to announce in search results.

Metadata must be exposed to assistive technology, not merely stored. Form fields need labels. Filters need keyboard access. Status should not be communicated by colour alone. Search results should announce the title, date and type in a logical order. These are repository design issues, but capture decides what metadata exists in the first place.

Build metadata capture into the scanning process. Barcode cover sheets, prompted fields and validation rules are more reliable than retrospective tagging. A document that arrives without a title will usually remain untitled.

Low vision and dyslexia

Low vision users need zoom, contrast and reflow. A scanned image can be enlarged, but it does not reflow to fit the screen. A text layer with responsive HTML gives the reader more control. Where the original layout matters, provide both: the image for fidelity and the text for reading.

Dyslexia-friendly presentation is not about rewriting records. It is about reducing unnecessary friction. Clear fonts in the repository interface, consistent navigation, adequate line spacing and plain-language descriptions help. Justified text in wrapper pages can make reading harder. Long documents benefit from summaries and clear section headings, even when the original record remains unchanged.

These adjustments are not cosmetic. They determine whether a colleague can complete a task independently or has to ask for help.

UK duties in practice

The Equality Act 2010 requires employers to make reasonable adjustments for disabled employees, and the public sector equality duty applies to public bodies. The Public Sector Bodies (Websites and Mobile Applications) Accessibility Regulations 2018 require public sector websites and apps to meet accessibility standards, typically WCAG 2.2 AA. Private SMEs are not exempt from workplace duties simply because the regulations are aimed at public bodies.

This is not legal advice, but it is a practical warning: an inaccessible document repository can become an inclusion problem, a procurement problem and a reputational problem. Ask suppliers for accessibility conformance reports and evidence of testing. Do not accept the word accessible without a standard, a date and a scope.

Test with the people who use it

Accessibility testing is often reduced to a checklist. A better approach is task-based testing with screen reader users, keyboard-only staff, people with low vision and dyslexic colleagues. Ask them to find a supplier invoice from March, read the total, download a copy and share it with a colleague. Observe where they stop. Fix the cause. Repeat.

Make accessibility a gate in every capture project, not a review at the end. The archive that works for everyone is usually the archive with better search, cleaner metadata and fewer duplicate scans. Inclusion and good records management tend to reinforce each other.