Is automated PDF remediation enough for ADA Title II?

By Nilesh Gadekar, Founder · Last reviewed: November 2026.

What automated remediation does well

Automated remediation and automated checking are related, but different, activities. A remediation tool changes a file; a checker examines it for supported conditions. Both can help a district process a backlog more efficiently. A tool may add an initial tag structure, recognize text in a scan, suggest descriptions or report missing metadata. An inventory service can also identify publicly linked files, approximate their volume and help staff find repeated editions before a production project starts.

The strongest starting point is a well-authored source with straightforward text, consistent headings and few unusual layouts. Simple letters and notices can often move through a repeatable workflow quickly. Automation reduces repetitive effort and makes reruns economical. Verify that the export conveys the right meaning and supports its intended interactions.

Use automation early in drafting, export and triage. Keep the original source and generated output connected so a reviewer can compare them. Even an internal draft may need to work for staff using assistive technology. The useful distinction is between preliminary processing and verified delivery, rather than a rule that internal content never needs review.

What a checker can find, and what a person still judges

An automated check may find that a document lacks tags, an image lacks an alternative-text entry or a form control lacks a programmatic label. Depending on the tool, it may inspect additional structural relationships or flag suspicious patterns. Some newer systems also propose repairs. These are useful findings and suggestions, but the presence of a tag, heading or description does not establish that it expresses the intended information.

A reviewer must consider whether an image description communicates the relevant fact, whether headings divide the document meaningfully and whether a table cell has the right header context. Reading sequence matters when the page includes columns, side notes or attachments. A screen-reader task can expose a confusing result that a visual inspection overlooks. It should complement, rather than replace, structural validation and other applicable checks.

The W3C evaluation-tool guidance explains that tools assist evaluation and that human judgment is necessary for checks that cannot be fully automated. Record the software, version, tasks, coverage and findings; do not assign an invented percentage of issues to tools or people.

Four typical examples to test in your own sample

These are typical failure scenarios for evaluation, not customer cases, measured failure rates or claims about a particular product. A tool may successfully handle any one of them. Use them to inspect representative output.

Budget table without header context

A fictional budget packet has a grid of departments, years and amounts. A reader hears individual values but cannot reliably identify the department or year each value belongs to. The reviewer checks the table’s relationships and follows meaningful financial comparisons. Ask the business office to resolve ambiguous labels; do not let technical repairs invent financial meaning.

Fillable form with unnamed controls

An enrollment form looks complete, but a person moving by keyboard reaches fields announced without useful questions. A test includes locating each required input, understanding its label and completing the intended sequence. The review also checks the returned form’s actual behavior, rather than treating visible text beside a field as proof of a programmatic label.

Newsletter that crosses between columns

A two-column newsletter alternates lines from unrelated articles when read in sequence. The reviewer follows each story, its heading and any associated callout. A layout that looks balanced on the page can still need a different logical sequence. Check the assembled export after source editing, not an earlier file that merely looked similar.

Image description that repeats a filename

A chart’s alternative text reads “IMG_2041.jpg.” An entry exists, yet the reader misses the comparison the chart was meant to communicate. The reviewer asks what information the image adds and obtains an approved description from the content owner. A longer chart explanation or accompanying data table may be needed as well as a short label.

Compare coverage, not just the first processing step

The table describes workflow differences to confirm in a quote. “Often missed” identifies a review concern, not a measured product failure rate. Capabilities, file complexity, sampling coverage and correction terms vary. In the sampled column, a person checks only the selected sample unless the agreement explicitly expands that coverage.

Do overlays or widgets fix PDFs?

A webpage overlay acts on the webpage environment. By itself, it does not rewrite the internal tags, table relationships or field labels of a PDF that a family downloads. Some services separately process documents or provide an HTML alternative. That output should be evaluated on its own merits: identify the changed artifact, confirm the public location and test the relevant task. A new interface around a link is not evidence about the downloaded file.

In January 2025, the FTC announced a proposed order concerning allegedly unsupported marketing claims for an automated web-accessibility product. The settlement included a $1 million payment; the FTC approved the final order in April 2025. The January announcement and case timeline support that narrow factual account. It is not a ruling that every automated tool fails, nor a finding about DistrictProof or a district’s PDFs.

Already using a tool? Bring 25 processed pages

Send us 25 pages your tool has already processed. We review the submitted sample, report the outcome of the screen-reader tasks and identify the findings that remain. Bring a mix of simple text, a real table, a form and a scan if those types appear in your public documents. That makes the review useful for the decisions your team actually faces.

The sample review is free. Corrections are separate: we fix the agreed work at published per-page rates after confirming the written scope. The report identifies reviewed pages, versions and tasks; it does not silently extend a sample result to the unreviewed backlog. Start with public links or arrange secure intake before sending private material.

When to use which approach

Use automated processing for early drafts, inventory and repeatable first-pass work on straightforward documents. Add a planned sample when you need to understand recurring issues or estimate a backlog, while recording what remains unreviewed. For public board packets, budgets, handbooks, enrollment material and forms, purchase review coverage that tests the actual information and tasks. A person should verify the delivered file, not simply inspect a tool’s score.

Keep content approval and publication responsibility visible. Connect the reviewed version with the file posted online, so replacements do not inherit earlier evidence.