Did you know that a worksheet can look perfect on a laptop screen while its words remain invisible to search, copy-and-paste and some reading technologies? Two pupils may see the same page yet be unable to use it in the same way. The difference often begins with four everyday technical terms: scan, text, OCR and reading order.
The useful answer is to check the job of each layer. A scanned PDF may hold photographs of printed pages; OCR (optical character recognition) attempts to recognise characters in those pictures; searchable text lets readers locate and copy words; and a tagged, correctly ordered document communicates structure such as headings, lists and the sequence in which content should be read. Searchable does not automatically mean accessible or accurate.
Contents: 1. The homework instruction that two students hear differently · 2. Scan is a capture method, not a guarantee about text · 3. OCR makes an interpretation that must be checked · 4. Searchable text and selectable text are useful but separate checks · 5. Tags describe what the document means structurally · 6. Reading order is not the order that looks convenient to the eye · 7. An example with a school timetable · 8. The four-test diagnosis for parents and students · 9. A worked case that distinguishes three kinds of success · 10. How students can practise without risky tools or private uploads · 11. From first reading to independent transfer · 12. Common mistakes and the exact repair · 13. Frequently asked questions · 14. Why a small vocabulary distinction improves learning
The homework instruction that two students hear differently
Imagine an invented Secondary 1 Science worksheet. The first page shows a table of temperatures and a neat paragraph telling pupils to compare the readings. A teacher photographed the paper and saved it into a PDF. On the screen it resembles the original, but when a pupil searches for “temperature”, the PDF viewer returns no result. Another pupil tries selecting the sentence and gets a rectangle covering the entire page rather than individual words.
The problem is not that the pupil lacks reading comprehension. The file may contain an image of the letters instead of actual characters. That difference matters if a learner needs a screen reader, selectable quotations for notes, quick keyword search or an accessible enlargement of the text. We need vocabulary precise enough to tell the teacher which property of the document is missing.
Compare these reports. “The PDF is broken” is vague. “The pages are visible images, but the text cannot be selected or searched” is a testable observation. “The file needs an OCR pass, followed by a recognition check and a reading-order check” identifies a potential repair. As language becomes more precise, the next action becomes more effective.
Scan is a capture method, not a guarantee about text
A scan is a digital capture of a physical page, commonly made by a scanner or camera. Scanned pixels can preserve page layout and handwriting even when no searchable text exists. A born-digital PDF is created by exporting from an electronic authoring tool, such as a word processor or typesetting program; it may contain live text. But born-digital does not promise correct tags or accessible structure either. The category tells us something about origin, not everything about usability.
One page can mix several content types. A publisher may provide live headings, scanned graphs, image-only handwritten annotations and selectable footnotes in the same file. Therefore, a one-page test is not enough when a document has many different page designs. Students should describe exactly which portion has been checked: for example, “The first printed paragraph can be searched, but labels inside the embedded diagram cannot.”
OCR makes an interpretation that must be checked
Optical character recognition examines image shapes and attempts to turn them into characters. It can transform a photograph of the word “mass” into machine-readable text, but the process may misread letters, spaces, symbols or units. The World Wide Web Consortium’s PDF7 guidance on OCR explains that scanning alone leaves text inaccessible and that recognised content should be checked for completeness and correct order. OCR is a step toward accessibility; it is not the entire job.
Consider a fictional laboratory table whose image shows “1.0 mL”. If imperfect OCR creates “10 mL”, the error changes the amount by a factor of ten. A pupil who copies the OCR output without looking at the actual measurement may write an incorrect conclusion. Even a small character conversion can move an answer from correct to unsafe or misleading. In high-stakes material, refer to verified original information rather than trusting a plausible-looking automated transcription.
The word recognise in OCR describes the computer’s proposed interpretation. It does not mean the result is necessarily verified. A recognition error needs review; a verified transcription means someone has checked the converted text against the source using suitable methods. The verbs detect, recognise, verify and correct describe different actions.
Searchable text and selectable text are useful but separate checks
A PDF can offer a text layer that allows searching even if the words were recognised inaccurately. Try searching for a distinctive phrase, then selecting one sentence, copying it into a plain-text editor and comparing each character with the visible page. Do not assume that because a word was found, every mathematical symbol and line break was preserved. Likewise, copying an apparently complete paragraph does not prove that the page’s two-column layout will be read in the correct sequence.
The distinction is particularly useful during revision. A student may want to search a 30-page chemistry booklet for “diffusion”, while another may need to enlarge text or listen to it. Search supports retrieval; accurate extraction supports quoting and taking notes; a meaningful reading order supports comprehension. The student’s real question is not just “Can I open it?” but “Can I use the parts of this document needed for my task?”
Tags describe what the document means structurally
A tag in an accessible PDF can identify a heading, paragraph, table cell, link or image. Proper heading tags let technologies recognise a document’s hierarchy; the W3C’s PDF9 guidance explains how heading tags communicate structure. Visual appearance alone cannot tell you whether that hierarchy exists inside the file. Large bold lettering can look like a heading while remaining an unmarked text span.
Consider a worksheet with one title and three exercises. If the source document treats everything as ordinary text without reliable tags, an assistive reader may not provide useful heading navigation. If each exercise is accurately tagged as a section, the learner gains a map of the material. Good structure is not decoration: it is a way to make the same intellectual content accessible through more than one reading method.
Another important word is alternative text, or alt text. A diagram may communicate information visually that cannot be deduced from an image filename. Providing a meaningful textual equivalent where appropriate can help a student understand the diagram’s role. An automatic OCR pass does not by itself author a correct explanation of every chart, illustration or photograph.
Reading order is not the order that looks convenient to the eye
A two-column page might appear obvious to a sighted reader: finish the left column, then read the right. But the underlying text objects could be stored in a sequence that jumps left–right after every line. A screen reader or copied text may then produce a confusing mixture of two unrelated paragraphs. That is a reading-order problem, even when every individual word has been recognised correctly.
The W3C’s PDF3 guidance on logical reading order explains why the sequence of tagged content and interactive elements matters. In an accessible review, the tester should check what is read first, next and last. The linked eduKate specialist, Why a Worksheet Can Read in the Wrong Order, develops this sequence test further. This article instead concentrates on telling scan, OCR, searchable text and structure apart.
An example with a school timetable
A fictional school issues a photographed timetable with two columns: “Monday: English, Mathematics” at left and “Tuesday: Science, Art” at right. The visual order is clear, but extracted text reads “Monday Tuesday English Science Mathematics Art”. All six words have been captured, yet the temporal grouping has been lost. The information is not properly usable if a pupil listening to the file cannot tell which subjects belong to which day.
Repairing the file is not necessarily a matter of adding more OCR. The semantic groupings need to be represented and tested, perhaps by recreating the source timetable with proper table structure or by providing an accessible text alternative. Before any repair, preserve the original schedule and verify that the recreated version retains its meaning. The goal is equivalent usable information, not simply a higher count of recognised words.
The four-test diagnosis for parents and students
First perform the search test: search a distinctive printed word. If it cannot be found, note the limitation without assuming the whole document is image-only. Second perform the selection test: try selecting and copying one sentence. Third perform the accuracy test: compare punctuation, numbers, symbols, superscripts and units against the visible original. Fourth perform the sequence test: read or export a two-column passage or table and check that the meaningful order is preserved.
These checks answer different questions. A passing search test does not erase recognition errors. Passing the accuracy test for one sentence does not prove the whole booklet is correctly tagged. A reader using assistive technology may need more comprehensive testing, including keyboard navigation, links and alt text. A teacher can begin with these observable checks and then follow the school’s accessibility process rather than declaring a technical certificate from a quick experiment.
When documenting a problem, write a miniature diagnostic record: file name, page, task, observed behaviour, expected behaviour, impact and attempted check. For instance: “Page 4; I searched for the word ‘evaporation’, but no result appeared even though the word is visible; a colleague reproduced this in the same viewer.” That report is more actionable than “OCR bad.” It also teaches the student to separate observations from explanations.
A worked case that distinguishes three kinds of success
Imagine three fictional versions of the same five-page worksheet. Version A displays all pages, but no words can be selected. Version B lets a reader search all five pages, although “0.5” on page 3 is recognised as “O.5”. Version C has correct extracted numbers and headings but reads a table row in the wrong order. Their failures are not the same. A needs text recognition or a true-text source; B needs recognition correction; C needs structure and sequence repair.
If someone says “Version B is accessible because the search works”, identify the gap. Searchability demonstrates only one function; the mistaken character and any other untested accessibility barriers remain. If someone says “Version C is fine because all letters are correct”, identify the second gap. Characters can be correct and still arrive in an order that changes the meaning. This layered comparison is the heart of the vocabulary lesson.
How students can practise without risky tools or private uploads
A safe classroom exercise uses a teacher-made fictional worksheet with no student names, marks or confidential information. Export one version from a word processor; create an image-based teaching example; compare selection and text extraction. A school-approved OCR application can be used where permissions allow. If a product asks students to upload a file to an unfamiliar online service, they should not upload real school documents or personal data just to complete this exercise.
Ask pupils to predict which copy is easiest to search, which preserves the original appearance, and which permits heading navigation. Test those predictions, then discuss why the results sometimes differ from expectations. Give students one sentence to finish: “I can see the words, but I cannot yet prove that ______.” Answers might include “they are real text”, “the numbers were recognised correctly” or “a screen reader receives the correct order.”
From first reading to independent transfer
On the first day, pupils label examples as scanned image, recognised text or tagged structure. The next day, remove the labels and ask them to design a suitable test for each claim. A week later, transfer the skill to a PDF containing a photograph of a newspaper article or an examination revision sheet. The student should know which limitation to test even though the subject and appearance have changed.
For English learning, turn a vague paragraph into evidence-based language: “The file appeared legible on my phone. Keyword search did not locate the visible heading. A text extraction test is required before I describe it as searchable.” For Science, treat OCR output as a measurement-like record whose accuracy must be checked. For Mathematics, show how confusing 1 and 7 or a negative sign can change a calculation. The contexts vary, but the habit of checking what a word claims remains the same.
Common mistakes and the exact repair
Mistake one is using PDF as a synonym for accessible document. PDF names a file format, not a guarantee about internal text, tags or order. Mistake two is using OCR complete as a synonym for proofread. Recognition still needs a check, particularly for numbers and unusual layouts. Mistake three is using tagged as a synonym for correctly navigable. The sequence and semantics must be tested.
Mistake four is responding to an inaccessible worksheet by asking the pupil to try harder or change the zoom level. Increased zoom can help some readers, but it cannot manufacture searchable text or correct a scrambled reading order. The targeted repair is to recreate the document from a structured source where possible, perform and review OCR when needed, tag meaning correctly and test with relevant assistive technology. Users’ needs must guide that choice.
Frequently asked questions
Does every scanned PDF need OCR?
If the scanned pages contain essential written material and the goal is searchable, machine-readable text, OCR or an equivalent accessible text source is usually needed. But an OCR pass alone does not settle structure, alt text or reading order. Use an appropriate authorised school process.
Is searchable the same as accessible?
No. Searchable text can still contain recognition mistakes, missing descriptions, poor tagging or an incorrect reading sequence. Accessibility requires attention to the actual ways readers navigate and understand the document.
Can a document be born-digital and still difficult to read?
Yes. An exported PDF can retain live text but have badly structured headings, complex graphics without suitable alternatives or problematic content order. The file’s origin is evidence of a workflow, not proof of success.
How can I tell whether OCR changed a number?
Compare extracted numbers with the displayed source, including units, decimal points, minus signs and mathematical notation. Test representative pages, and expand the check when the stakes or error risk are high.
What should I write when asking a teacher for an accessible version?
Describe the task and observation: “I can open page 3, but I cannot select the text, and the reader speaks the table in the wrong order. Could I have a version with searchable text and correct reading order?” That request states a specific barrier and desired outcome.
Why a small vocabulary distinction improves learning
In this case, precise words become an instrument for checking information. Scan describes capture; OCR describes attempted recognition; searchable describes a function; tagged describes structural markup; reading order describes sequence. Students who learn the distinctions can request the right repair, produce clearer resources and understand why digital availability is not the same thing as equal access.
Read the eduKate screen-reader sequence guide for the deeper order problem and the PDF page-label guide for a separate navigation distinction. For connected word learning, use the vocabulary meaning and precision hub. Technical primary references reviewed 26 October 2026: W3C PDF7 OCR, W3C PDF9 headings and W3C PDF3 reading order. W3C techniques describe possible ways to meet accessibility requirements, not a guarantee that a file is compliant.