> ## Documentation Index > Fetch the complete documentation index at: https://docs.tessera.edstratumlabs.ai/llms.txt > Use this file to discover all available pages before exploring further. # Tessera Access > How accessibility scores, checks, fixes, and accessible formats work. Tessera Access is accessibility built into authoring and learning, not a separate audit run afterward. It checks lessons as they're written, checks uploaded documents when they're scanned (automatically the first time someone opens a file's report), and blocks a lesson from publishing when it doesn't meet the institution's policy. ## Scores and grades Every accessibility report — a lesson, a file, a course, or an institution — carries a score from 0 to 100: ``` score = max(0, 100 − Σ penalty(severity) × min(count, 3)) ``` Penalties by severity: **critical 30**, **serious 15**, **moderate 8**, **minor 3**. Three or more of the same issue count as three, so one severe, recurring problem doesn't sink the score past the point of being useful feedback. The score maps to a grade: | Grade | Score | | -------- | ------------ | | Perfect | 90 and above | | Good | 75 and above | | Moderate | 50 and above | | Low | 25 and above | | Very low | below 25 | A course's or institution's score is the average of its parts' scores. ## What's checked ### Lessons and assignment instructions Tessera's lesson content is a list of typed blocks, so these checks run directly against the block model — instantly in the editor, and again at the publish gate: | Check | WCAG 2.2 | | --------------------------------------------------------------------------- | --------------------------------- | | Image without alt text (or a placeholder like "image1.png") | 1.1.1 Non-text content (A) | | Heading level skips one or more levels | 1.3.1 Info and relationships (A) | | Empty heading | 2.4.6 Headings and labels (AA) | | Link text that doesn't say where it goes ("click here," a bare URL) | 2.4.4 Link purpose in context (A) | | Table without a marked header row | 1.3.1 Info and relationships (A) | | Video without captions or a transcript | 1.2.2 Captions, prerecorded (A) | | Information given by color alone ("the items shown in red") | 1.4.1 Use of color (A) | | Knowledge check missing a question, two answers, or a marked correct answer | 3.3.2 Labels or instructions (A) | | Dense reading level (an approximate grade level over 14) | 3.1.5 Reading level (AAA) | | Very long paragraph (over 200 words) | 1.4.8 Visual presentation (AAA) | | Empty lesson (no blocks at all) | 2.4.2 Page titled (A) | ### PDF files Checked in the Worker against the document's real structure, not just its visible text: | Check | WCAG 2.2 | | ----------------------------------------------------- | -------------------------------- | | No extractable text (the PDF appears scanned) | 1.1.1 Non-text content (A) | | No usable tag structure | 1.3.1 Info and relationships (A) | | No document title set | 2.4.2 Page titled (A) | | No document language set | 3.1.1 Language of page (A) | | Images without matching alt text in the tag structure | 1.1.1 Non-text content (A) | | No headings detected | 1.3.1 Info and relationships (A) | | Structure order differs from the visual reading order | 1.3.2 Meaningful sequence (A) | ### Word documents | Check | WCAG 2.2 | | ----------------------------------------------------- | -------------------------------- | | No document language set | 3.1.1 Language of page (A) | | A long document with no heading styles | 1.3.1 Info and relationships (A) | | No document title set | 2.4.2 Page titled (A) | | Images without a meaningful description | 1.1.1 Non-text content (A) | | Tables without a marked header row | 1.3.1 Info and relationships (A) | | Text under 9 pt | 1.4.4 Resize text (AA) | | Text and background contrast below the required ratio | 1.4.3 Contrast, minimum (AA) | ### PowerPoint files | Check | WCAG 2.2 | | ----------------------------------------------------- | ------------------------------------------------ | | Slides with no title placeholder text | 2.4.2 Page titled (A) | | Images without a meaningful description | 1.1.1 Non-text content (A) | | Most slides have no speaker notes | 1.2.1 Audio-only and video-only, prerecorded (A) | | Slide text under 18 pt | 1.4.4 Resize text (AA) | | Text and background contrast below the required ratio | 1.4.3 Contrast, minimum (AA) | ## Fixes Every fix — alt text, a marked table header row, corrected title or language metadata, or OCR — saves a **new version** of the document. The original is never overwritten; every version is kept. AI-suggested fixes (an alt text description, a plain-language rewrite, a clearer link name) are drafts, the same rule that applies to every other AI output in Tessera: a person reviews and keeps a suggestion before it takes effect. An unreviewed AI draft blocks publishing, but it does not itself count against the accessibility score — the score reflects the content's actual state, not whether a draft is pending review. Alt text can be written into Word and PowerPoint files. For PDFs, Tessera doesn't yet write alt text into the file or extract its images for AI suggestions; the fixes it applies to PDFs are title, language, and OCR. The reading version and e-book carry a document's text and headings, not its images. ## Accessible formats Every PDF, Word, and PowerPoint file gets a menu of accessible formats, generated on first request and cached per version: * **Reading version** — the extracted content rendered as a plain, accessible page. * **E-book** (ePub 3). * **Audio** — a synthesized read-through. * **Searchable version** — for a scanned PDF with no extractable text, OCR produces a searchable copy so screen readers and search both work. See [OCR for scanned PDFs](#ocr-for-scanned-pdfs) below. ## OCR for scanned PDFs A scanned PDF has no text layer for a screen reader to read or for search to index. Running OCR on it is both a **fix** — it saves a new, searchable version of the file — and an **accessible format** in its own right, since the reading version, e-book, and audio of a scan are built from the OCR text. ## Publishing policy Administrators set two things that the server enforces, not just the editor screen: * A **minimum accessibility score** a lesson must meet to publish. * A list of **blocking severities** — for example, treating any critical or serious issue as a publishing block regardless of the overall score. An instructor sees exactly why a lesson can't publish: the score, the policy's threshold, and which issues are blocking.